Smartphones powered by large language models (LLMs) and Agentic AI promise to reshape personal computing with advanced reasoning, contextual awareness, and autonomous decision-making, even as the fate of overhyped devices such as the Rabbit R1 and Humane AI Pin serve as a cautionary note.

In this week’s edition of Mint Tech Talk:

• OpenAI’s rumoured Agentic AI smartphone

• AI Tool of the Week: ChatGPT Images 2.0

• Everyone wants a piece of Anthropic PBC

In May 2025, Sam Altman (who invested in Humane AI) brought on board legendary iPhone designer Jony Ive to build a new class of AI-first devices at OpenAI. And while OpenAI has made no official comment on this topic, OpenAI could unveil its first smartphone by late 2026 or early 2027, with mass production targeted for 2028, according to Ming-Chi Kuo—an analyst with TF International Securities (HK).

The AI smartphone is expected to blend cloud and on-device LLMs, powered by chips optimised for energy efficiency, memory management, and continuous contextual awareness. Partnerships with MediaTek and Qualcomm are reportedly under consideration, as OpenAI seeks end-to-end control over both the hardware and operating system to deliver a “comprehensive AI agent service”, says Kuo.

OpenAI has built a broad infrastructure stack spanning cloud, silicon, and data centres. The company has also outlined plans for a “superapp” that brings together ChatGPT, Codex, browsing, and broader agentic capabilities into a unified, agent-first experience. An “Agentic AI” phone, and a line-up of similar “agentic” devices, closes the loop on the hardware side.

Still, the question remains: Even if OpenAI does launch a ChatGPT phone, how different will it be from today’s Gen AI smartphones?

Devices by Samsung, Google, and Apple already integrate Gen AI deeply into their chips and operating systems, running on-device models via dedicated neural processing units. Smartphones, such as the Google Pixel 10 Pro with its Tensor G5 chip and Samsung Galaxy S26 powered by the Exynos 2600 on a 2 nm processor, highlight how far this integration has come—fast, efficient, and capable of handling meaningful on-device workloads across text, voice, image, and multimodal tasks.

The market is still early but growing. According to Intel Market Research, Gen AI smartphones were valued at about $95 million in 2025 and are projected to reach roughly $414 million by 2034. In 2025 alone, production reached 185 million units at an average price of $558. Demand is being driven by content creation, productivity, and emerging use cases in healthcare, education, and creative fields, while advances in model compression are pushing these capabilities into mid-range devices.

Yet today’s AI remains largely reactive—it waits for prompts. It can summarise, generate, and assist, but it does not autonomously manage tasks like travel-planning or subscription optimisation end-to-end. A true Agentic AI device, by contrast, would embed intelligence into the operating system itself, shifting interaction from “open an app” to “state a goal”.

Potential buyers?

OpenAI potentially has a captive audience. Currently, it has more than 900 million weekly active users, and that base is expected to approach one billion in the near term. Likely AI-native professionals already using tools like ChatGPT, Gen-Z driving cultural adoption, and enterprises seeking workflow automation. A phone purpose-built for agentic enterprise workflows—one that can autonomously draft, send, and follow up on communications—is an automation tool, not merely a handset.

Cost and trust

Gen AI adds an estimated $120–$200 to device manufacturing, even as privacy concerns remain high, according to Intel Market Research. Layer in EMIs, data plans, and AI subscriptions, and pricing becomes tricky. That is unless, as Google and Apple suggest, AI is bundled and largely hidden from the user. IPO-bound OpenAI, on its part, has not disclosed any details.

But where’s the money?

Meanwhile, The Wall Street Journal reported that OpenAI missed its own targets for new users and revenue, raising concern among company leaders about whether it will be able to support its massive spending on data centres. Chief Financial Officer Sarah Friar has said that she is worried that OpenAI may not be able to pay for future computing contracts if revenue doesn’t grow fast enough.

OpenAI has also ended its exclusivity arrangement with Microsoft Corp., meaning the Azure parent will no longer have sole access to OpenAI’s models and products. The move opens the door for OpenAI to work with rival cloud platforms such as Google Cloud and Amazon Web Services.

Announced on 27 April, the revised agreement also removes Microsoft’s obligation to pay a revenue share on OpenAI products it resells via its cloud.

“The greater predictability in the amended agreement strengthens our joint ability to build and operate AI platforms at scale, while providing both companies the flexibility to pursue new opportunities,” the two companies said in a joint statement.

Subscribe now

AI BITS & BYTES

Everyone wants a piece of Anthropic

Google will invest $10 billion in Anthropic PBC, with another $30 billion potentially to follow, strengthening the relationship between two firms that are at once partners and rivals in the race to build AI.

Anthropic has ramped up its fundraising amid the breakout success of Claude Code, an AI agent that speeds the process of writing computer software. The startup said earlier this week that it nabbed another $5 billion from Amazon, also at a valuation of $350 billion, with the option to inject another $20 billion over time.

Anthropic raised $30 billion in February and investors have since sought to back the firm at a valuation of $800 billion or more. Read more.

AI exposure highest in skilled roles not low-wage jobs

According to Anthropic’s Labor Market Impacts Report, professionals currently most exposed to AI are those with higher education, significant experience and above-average earnings.

The findings in this report turn conventional thinking upside down. There’s a clear tension here: the qualifications and expertise that once helped secure a stable career may now be the very factors placing workers at greater risk of disruption.

Amazon’s 6-point AI playbook for engineers

Amazon.com Inc. is formalising how it builds with AI as part of a broader push to make AI central to its engineering culture, Business Insider reports. The e-commerce firm’s massive retail division, known internally as “Stores”, has formalised its approach into a set of six “AI-native engineering tenets” designed to guide how teams should approach AI development across the organisation.

Meta to unwind $2.5-billion Manus acquisition

Meta Platforms Inc. is planning to unwind its acquisition of AI startup Manus after the Chinese government banned the transaction on national security grounds, according to a new report by The Wall Street Journal, citing sources familiar with the matter.

Any attempt to reverse the deal is expected to be complex, requiring the company to disentangle operations, data, and technology that have already been combined.

Meanwhile, a Bloomberg report noted that Manus staff have been integrated into Meta Platforms, funds have been dispersed, and the startup’s leadership has joined the company’s AI division.

Elon Musk tells his side of the OpenAI story

Elon Musk took the stand for the second day Wednesday in the landmark trial that pits the world’s richest person against Sam Altman, a fellow OpenAI co-founder he accuses of betraying promises to keep the company as a non-profit dedicated to humanity’s benefit.

Musk, who invested about $38 million in OpenAI from December 2015 through May 2017, gave his account of OpenAI’s early years, recounting how he lost confidence that Altman would keep it a non-profit. Questioned by his lawyer Steven Molo, Musk said by late 2022 he was concerned Altman was trying to “steal the charity.”

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

In this week’s edition of Mint Tech Talk:

• Hiring slows but data shows AI isn’t killing jobs

• Humanoid beats human record at half-marathon

• AI Tool of the Week: Skills in Chrome

• ChatGPT Images 2.0 ups the stakes

Meta Platforms Inc.’s Mark Zuckerberg, for instance, is working closely with software development teams after two decades. He has moved his desk into Meta’s AI Lab, sitting alongside Alexander Wang and Mat Friedman—co-leads of Superintelligence Labs—and has been actively writing and reviewing code, according to a report by Business Insider. Meta is even building a photorealistic, AI-powered 3D avatar of Zuckerberg to engage with employees in his stead, Financial Times reports.

Google co-founder Sergey Brin is making a similar return. The Information reports that the Gemini maker is assembling an elite “coding strike team” to close the gap with Anthropic, with Brin directly involved. The ambition is to push towards an “AI takeoff” in which systems increasingly improve and code themselves.

Closer home, in January 2025, co-founder Sridhar Vembu stepped down as CEO of Zoho to become chief scientist, focusing once again on writing and reviewing code.

Set against this, Anthropic co-founder and CEO Dario Amodei has argued that coding itself may eventually disappear, replaced by systems that turn human intent directly into working software.

Anthropic is already building toward that future with tools like Claude Code, Mythos (restricted distribution due to security reasons even as OpenAI Inc.’s GPT-5.4-Cyber, released this month, is a model finetuned for additional cyber capabilities and with fewer capability restrictions for those who are part of its Trusted Access for Cyber programme), and Claude Design.

After triggering a selloff in stocks of SaaS companies earlier this year with Claude Code, Anthropic’s new AI tool Claude Design is threatening software design firms like Adobe and Figma. Anthropic says Claude Design can create polished visual assets ranging from interactive prototypes to pitch decks—all with a simple text prompt.

Given these developments, what looks like a contradiction is, in fact, a shift. AI is reshaping how software gets built. Instead of writing detailed instructions, developers increasingly describe outcomes and let machines generate the rest. That makes AI the new interface between humans and computers—a role once held by operating systems and cloud-computing platforms. Whoever controls this interface, in effect, shapes how software is created.

This is precisely why founders are getting closer to software again—not just to write more code but to understand how this new layer behaves.

Consider Meta’s internal targets. According to documents cited by Business Insider, the company expects that by mid-2026, 65% of engineers in its core product groups will generate more than 75% of their code using AI. In a world where small changes in prompts can produce widely different outcomes, rolling up one’s sleeves is not a throwback. It is a strategy.

At the same time, AI is compressing software work. Tasks that once required entire teams can now be handled by a single developer equipped with the right tools. This is not full automation, but a reduction in the number of people needed between idea and execution. Cost too is becoming a constraint: AI-generated code is not free. Tokens carry a price. Every prompt and iteration consumes compute, raising a basic question of efficiency: When is it cheaper to use machines, and when do humans remain more effective?

The answer will shape how quickly AI transforms traditional development. For now, the technology remains imperfect. AI can generate code quickly, but it is still prone to errors, edge cases, and unpredictable behaviour. That makes human oversight more important, not less. The work is shifting from writing code to guiding and validating it.

So what changes? Routine coding tasks will decline. Large engineering teams may shrink. But higher-order skills—system design, judgment, and the ability to translate ideas into precise, executable instructions—will become more valuable. Coding does not disappear; it moves up the stack. The founders returning to it understand this. The future may need fewer programmers, but it will demand more leaders who can think like one.

Subscribe now

Hiring slows but data shows AI isn’t killing jobs

The bad news for white-collar workers continues. As per layoff.fyi, a little over 81,000 employees have been laid off by 97 tech firms since the beginning of this year.

LinkedIn too has confirmed a 20% decline in hiring since 2022. But it does not blame AI for the ongoing slowdown, as per a TechCrunch report, citing an interview at the Semafor World Economy Summit. LinkedIn’s Chief Global Affairs and Legal Officer Blake Lawit instead said the broader decline in hiring is more closely tied to rising interest rates.

“At LinkedIn…we have an economic graph which is over a billion members. We’ve got companies, jobs, skills. It’s really an amazing real-time view of what’s happening in the labour market. And we’ve looked—because everyone wants to know the answer to this question: Is AI impacting jobs right now? We’ve looked and, honestly, we haven’t seen it,” Lawit said. He argued that if AI was actively replacing workers, the job losses would be highly visible in specific, vulnerable sectors such as customer support, administrative roles, or marketing, which was not the case.

Even Stanford computer science graduates “are struggling to find entry-level jobs” in “a dramatic reversal from three years ago”. Yet, it’s hard to find data that provides evidence linking AI adoption to reduced job postings, according to an article in The Stanford Review.

Meanwhile, Yann LeCun, former AI chief scientist at Meta and founder of AMI Labs, has criticised Anthropic’s Amodei for his remarks that AI will reduce tech jobs by 50%. Lecun noted that AI lab CEOs—including himself, OpenAI CEO Sam Altman, Yoshua Bengio, and “Godfather of AI” Geoffrey Hinton—aren’t necessarily the right authorities to forecast AI impact on jobs and the broader economy. Instead, Lecun noted that economists can be a better judge of how AI will affect the job market.

“Dario is wrong. He knows absolutely nothing about the effects of technological revolutions on the labour market. Don’t listen to him, Sam, Yoshua, Geoff, or me on this topic. Listen to economists who have spent their careers studying this.”

By 2030, Gartner says, CIOs will expect all IT work to involve AI, with 75% handled by humans augmented by AI and the remaining 25% performed autonomously. The research advisory maintains that AI is less about job loss and more about workforce transformation, and expects AI’s overall impact on global employment to remain neutral through 2026 and turn net-positive by 2027 as new roles outpace those displaced.

AI BITS & BYTES

Humanoid beats human record at Beijing half-marathon

A humanoid robot developed by Honor has captured global attention after winning a half-marathon race for robots in Beijing, posting a time that surpasses the human world record for the distance. The event, held in the Beijing Economic-Technological Development Area (E-Town), highlights China’s accelerating progress in artificial intelligence and robotics.

The winning robot completed the 21-km course in 50 minutes and 26 seconds. This time is notably faster than the current human half-marathon world record held by Jacob Kiplimo, who ran the distance in approximately 57 minutes at a Lisbon road race in March 2026.

A separate Honor robot—operated via remote control—crossed the finish line even faster at 48 minutes and 19 seconds. However, under the competition’s weighted scoring system, which prioritised autonomy, the fully autonomous humanoid robot was awarded first place.

ChatGPT Images 2.0 ups the stakes

The internet is buzzing over OpenAI’s rollout of ChatGPT Images 2.0, which it frames as the start of “a new era of image generation”. The update builds on the December launch of ChatGPT Images positioned as a response to Google Gemini’s Nano Banana.

ChatGPT now “thinks” before it creates—planning outputs, pulling in references, and checking for errors before delivering results. The upgrade also brings practical gains: up to 2K resolution, batch generation of eight images, flexible aspect ratios from 3:1 to 1:3, and improved multilingual text rendering. It is not perfect, though.

I prompted it to “Please draw a picture of a Royal Enfield cruiser bike with essential accessories.” What I got was a decent picture but the fuel tank, as you see, belongs more to the Classic Model rather than the Meteor 350.

As ChatGPT Images 2.0 works to improve with time, the move intensifies competition in a crowded field that includes Midjourney, Stable Diffusion, Adobe Firefly and Google Gemini.

GPT 5.5 ‘Spud’ looms, GPT-6 still a meme

Reports and speculation suggest ChatGPT 5.5 (codenamed “Spud”) will be a modest upgrade within the GPT-5 family, aimed at bridging the gap between GPT-5.4 and GPT-6, with a possible release by April end. OpenAI, however, has made no official announcement.

As for GPT-6, the signal is even noisier. In October 2025, CEO Sam Altman leaned into the viral “6-7” meme, joking on X that GPT-6 might be rebranded “GPT-6-7,” after Dictionary.com crowned the absurd phrase its Word of the Year. The comment was tongue-in-cheek but underscored how Big Tech firms make gossip work for them.

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

Technically, leading AI models are converging, with open-weight systems narrowing the gap. But the means of measuring that progress are fraying: benchmarks are saturating, transparency is declining, and independent tests do not always align with developer claims, the report notes. In science, AI is shifting from assisting discrete tasks to replacing entire workflows. In medicine, tools such as ambient scribes are scaling across health systems.

Regulation, meanwhile, is sending mixed signals. The European Union is enforcing its AI Act, the US is leaning towards deregulation, while Japan, South Korea and Italy are advancing national frameworks. Many developing economies, including India, are entering the arena, with AI sovereignty emerging as a unifying theme. Optimism is rising, but so is unease.

Here are some highlights from the study:

➡️ AI isn’t plateauing. Instead, it’s accelerating.

➡️ Over 90% of frontier models now come from industry, many matching or beating humans on advanced tasks. Coding benchmarks neared human parity in a year, while adoption hit 88%, with most students using generative AI.

➡️ The US-China AI gap has narrowed to near parity. China leads in AI publications, citations, and patents, while the US dominates in high-impact patents and output (50 notable models in 2025 vs China’s 30). South Korea tops patents per capita, and China’s share of the 100 most-cited AI papers rose from 33 in 2021 to 41 in 2024. In 2024, China accounted for 17.8% of AI publications in 2024, compared to 11.1% from Europe and 7.6% from India.

➡️ The US leads in AI data centres (5,427)—10 times more than any other country—but the chip supply chain is fragile. Most advanced chips rely on Taiwan’s TSMC, creating a critical single-point dependency despite US expansion.

➡️ AI shows gold-medal math performance, yet they are unable to read time from traditional clocks (with about 50% accuracy). AI Agents improved to about 66% task success but still failed one in three tasks.

➡️ AI models can outperform scientists in some domains, but scaling isn’t linear. They beat chemists on average, yet perform poorly in areas like astrophysics.

➡️ Robots thrive in labs but struggle at home—only 12% success rate. Simulations hit about 89%, highlighting the gap between controlled environments and real-world complexity.

➡️ The US leads AI investment ($285.9 billion) and startups, but talent inflows are falling sharply—down 89% since 2017, including an 80% drop in the last one year.

➡️ India leads in AI skill penetration at 3.0 (nearly triple the global average), followed by the US (2.0) and Germany (1.8), LinkedIn data shows. But gender gaps persist. In India, men list AI skills far more than women (3.1 vs 1.9), with a similar, slightly narrower gap in the US (2.1 vs 1.4).

➡️ AI’s environmental cost is rising fast. AI data centre capacity reached 29.6 GW, on a par with New York’s peak demand, while GPT-4o’s annual water use may exceed the drinking needs of 12 million people.

➡️ Smaller, specialised models often outperform much larger ones. Scientific AI is also more collaborative, unlike industry-led general AI.

➡️ Open-source AI continues to scale, with 5.6 million projects on GitHub and Hugging Face uploads tripling since 2023. India remains a growing contributor, representing 5.2% of all projects.

➡️ AI is reshaping clinical care, cutting documentation time and burnout, but strong evidence is limited. Nearly half of studies rely on exam-style data, and only about 5% use real patient data.

➡️ Education is lagging, even as AI use surges. More than 80% students use AI, but policies are unclear and uneven. Meanwhile, AI skills are growing globally, and more PhDs are staying in academia.

➡️ AI sovereignty is rising, with nations investing in domestic capability. Still, model development is concentrated in the US and China. Open source is widening participation and enabling more diverse models.

➡️ Regional efforts are building language-first AI from scratch, not waiting on global labs. Projects like SEA-LION and AI4Bharat are creating local data pipelines, tokenisers, and benchmarks. These initiatives make linguistic inclusion a core design goal, expanding responsible AI beyond major hubs.

➡️ India saw the sharpest rise in AI nervousness of any country surveyed.

➡️ Responsible AI is lagging capability. Safety reporting remains patchy: incidents rose to 362 (from 233 in 2024), and improving one dimension (for example, safety) can worsen another, like accuracy.

➡️ Experts and the public diverge sharply on AI’s impact—73% vs 23% positive views on jobs. Trust in AI governance is fragmented, with the US among the least confident in its own regulation.

Subscribe now

AI BITS & BYTES

Now Hiring | Philosophers at Big Tech firms

Big Tech appears keen to signal a conscience—hiring philosophers to make the case, even as AI threatens to render such roles obsolete. This week, Google DeepMind brought on Henry Shevlin as a “philosopher”. Currently at Cambridge’s Leverhulme Centre for the Future of Intelligence, Shevlin works across cognitive science, AI ethics, and consciousness, with papers in journals like Journal of Consciousness Studies. His mandate now is to help machines think more responsibly.

At Anthropic, Amanda Askell is already doing just that—embedding ethical reasoning into systems like Claude. With a PhD in philosophy from NYU, Askell’s work spans AI alignment, decision theory, and formal epistemology, following earlier stints shaping safety research at OpenAI.

The irony is hard to miss. Even as tech firms recruit philosophers to teach AI right from wrong, some of their own leaders suggest those same systems could hollow out the humanities. Alex Karp, chief executive of Palantir Technologies, has consistently warned that AI will disrupt white-collar careers, especially in fields like philosophy, literature, law, and academia. Ironically, Karp himself has philosophy degrees from Haverford, Stanford, and a PhD from Frankfurt.

$1.35 million

It’s the average ransomware payout by organisations in India, which has emerged as the epicentre of ransomware activity in the Asia-Pacific, per Check Point. The latest report notes that 65% of affected Indian firms paid hackers during 2025.

Globally, attacks on manufacturers rose 56% from 937 incidents in 2024 to 1,466 in 2025 as hackers increasingly prioritise operational disruption and supply chain leverage over standalone data theft.

Meta may unseat Google as world’s largest ad player

Meta Platforms is expected to surpass Alphabet’s Google to become the world’s leading digital-advertising business—a first for the social-media company. Emarketer projects Meta’s ad revenue at $243.46 billion versus Google’s $239.54 billion.

The research firm’s estimates account for revenue after deducting traffic and other content acquisition costs, such as the money Google shares with its creators. Meta’s ad business is seeing a lift, thanks to new ad offerings, including the short-form video format Reels and the broader boost that AI has provided. Read more.

Is AI blocking creation of entry-level job?

A study by venture firm SignalFire found that hiring of candidates with less than one year of experience fell nearly 50% between 2019 and 2024, as work traditionally assigned to junior employees is being absorbed by AI systems.

IBM, in contrast, is increasing its hiring of Gen Z. “The companies three to five years from now that are going to be the most successful are those companies that doubled down on entry-level hiring in this environment,” Nickle LaMoreaux, IBM’s chief human resources officer, said during an interview with Charter this February.

ALSO READ | AI, Iran war hit campus hiring in India

Meta’s AI reset is a threat for jobs, data and its future

Meta Platforms may have finally shown its hand on AI, but the bigger story isn’t the technology. Its new model, ‘Muse Spark’, is being positioned as the backbone of the company’s next phase, powering its AI app and soon embedding across Instagram, WhatsApp and Facebook.

Unlike previous headline-grabbing bets with the metaverse or smart glasses, this is a subtle shift. But it is also far more invasive. The company says it will use public posts to provide “context from your people, right where you need it”, eventually integrating Instagram Reels, photos and posts directly into responses. In effect, Meta is turning its social graph into training data and distribution at the same time. Read more.

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

This month, two reports—one from OpenAI and another from the Center for Humane Technology (CHT)—outline how to keep humans from being sidelined in this speedy race by meaningfully keeping them in the loop.

In today’s edition of Mint Tech Talk:

• OpenAI’s blueprint for sharing AI dividends

• Some hard truths about AI’s race

• Google Gemma 4’s open challenge

• AI Tool of the Week: Google Vids

• Anthropic’s Mythos, Glasswing and Meta Muse

PREVIOUSLY, ON MINT TECH TALK

In its exploratory proposal titled Industrial Policy for the Intelligence Age, OpenAI Inc. outlines a social blueprint built on two pillars—an open economy and a resilient society—designed to distribute gains while containing risks.

Key Highlights

• A Public Wealth Fund: To counter extreme concentration of wealth, OpenAI proposes a fund giving every citizen a direct stake in AI-driven growth, seeded by long-term assets from AI firms and adopters.

• Tax on automated labour (read, robots too): As AI erodes labour income, the tax base must shift toward capital—via higher capital gains, corporate taxes, and potentially levies on automated labour.

• Four-day week: Productivity gains should translate into time. Incentivised 32-hour workweek pilots would maintain pay while converting efficiency into predictable “time dividends”.

• Right to AI: AI access should function like a public utility. Affordable foundational models, paired with “startup-in-a-box” support, can enable broader participation beyond tech giants.

• AI Trust Stack: A framework for trust in AI outputs, combining verifiable provenance with privacy-preserving audits, built on the principle that safety must scale with capability.

• Playbook for AI models: For systems that can’t be recalled, OpenAI proposes downstream response strategies, akin to public health playbooks, coordinated globally to manage high-risk scenarios.

Some hard truths about AI’s race, and how to fix it

Every technological leap carries twin narratives—utopia and collapse. AI breaks that pattern. Its speed and scale have outpaced both institutions and human instincts. What the CHT calls the “fuel of inevitability”—build it before rivals do—now drives the system. In this race, safety becomes a sidelined cost.

Fixing this isn’t about optimism; it requires rewiring incentives. Here are the seven principles for how AI should be built and governed, according to CHT’s AI roadmap.

In the near-future, AI Copilots will be free

By 2028, more than half of enterprises will stop paying for Assistive AI like copilots and smart advisors. They will, instead, back platforms that commit to delivering workflow outcomes, according to a new Gartner report. This marks a fundamental shift with humans moving from doing work inside procedural software to supervising intelligent systems that execute on their behalf.

The first wave of disruption will hit approval-heavy, time-sensitive workflows, where AI can compress decision latency and shift authority to policy-bound agents, Gartner notes. Over time, execution will move out of traditional interfaces into platforms that control enterprise context and can safely delegate work.

Human roles won’t disappear, but evolve into “agent stewards”, overseeing outcomes rather than performing tasks, according to the Gartner note.

By 2030, Gartner predicts companies that bolt AI onto legacy applications, rather than rebuilding for agentic execution, could see margins shrink by up to 80%.

Gemma 4’s Open Challenge

Gemma 4, positioned as Google’s most capable open model family so far, can process images, video, and even audio in some variants, and its context windows of up to 256K tokens can handle large documents or entire code repositories in a single prompt. Further, its training across more than 140 languages allows developers to build more inclusive, globally relevant applications, according to a Google blog.

Gemma, launched in four sizes—Effective 2B (E2B), Effective 4B (E4B), 26B Mixture of Experts (MoE) and 31B Dense—also enables high-quality offline code generation, effectively turning local machines and smartphones into AI coding assistants. It complements Google’s Gemini models, “giving developers the industry’s most powerful combination of both open and proprietary tools”.

That said, while Gemma’s multimodal model (it was just text-to-text when the first version was released in February 2024) may be pitted against closed models like Claude Code or OpenAI’s Codex, its real competition is from the fast-rising cohort of open-weight challengers, including DeepSeek, Qwen and Mistral.

China’s DeepSeek has already emerged as a favourite among developers for its strong reasoning and coding capabilities. Qwen pairs solid performance with deep integration into Alibaba’s enterprise ecosystem, while Mistral has built a reputation for efficiency and tight, developer-friendly application programming interfaces (APIs).

Other than performance, Gemma is also trying to differentiate on deployment. Unlike many rivals, it is aggressively optimised for on-device, low-resource environments. It is small, open-weight, and designed to run locally on laptops and smartphones. In markets like India, where connectivity, cost, and data sovereignty matter, this is a good advantage.

AI TOOL OF THE WEEK

By AI&Beyond, with Jaspreet Bindra and Anuj Magazine

𝚃𝚑𝚎 𝙰𝙸 𝚏𝚎𝚊𝚝𝚞𝚛𝚎/𝚝𝚘𝚘𝚕 𝚠𝚎 𝚞𝚗𝚕𝚘𝚌𝚔𝚎𝚍 𝚝𝚘𝚍𝚊𝚢: 𝙶𝚘𝚘𝚐𝚕𝚎 𝚅𝚒𝚍𝚜

𝚆𝚑𝚊𝚝 𝚍𝚘𝚎𝚜 𝚒𝚝 𝚜𝚘𝚕𝚟𝚎? 𝙴𝚟𝚎𝚛𝚢 𝚘𝚛𝚐𝚊𝚗𝚒s𝚊𝚝𝚒𝚘𝚗 𝚑𝚊𝚜 𝚝𝚛𝚊𝚒𝚗𝚒𝚗𝚐 𝚊𝚗𝚍 𝚌𝚘𝚖𝚖𝚞𝚗𝚒𝚌𝚊𝚝𝚒𝚘𝚗 𝚗𝚎𝚎𝚍𝚜 𝚝𝚑𝚊𝚝 𝚜𝚑𝚘𝚞𝚕𝚍 𝚋𝚎𝚌𝚘𝚖𝚎 𝚟𝚒𝚍𝚎𝚘, 𝚋𝚞𝚝 𝚘𝚏𝚝𝚎𝚗 𝚍𝚘𝚗’𝚝. 𝚂𝚊𝚏𝚎𝚝𝚢 𝚙𝚛𝚘𝚌𝚎𝚍𝚞𝚛𝚎𝚜 𝚐𝚎𝚝 𝚋𝚞𝚛𝚒𝚎𝚍 𝚒𝚗 𝙿𝙳𝙵𝚜. 𝙲𝚘𝚖𝚙𝚕𝚒𝚊𝚗𝚌𝚎 𝚞𝚙𝚍𝚊𝚝𝚎𝚜 𝚐𝚘 𝚞𝚗𝚛𝚎𝚊𝚍 𝚒𝚗 𝚕𝚘𝚗𝚐 𝚎𝚖𝚊𝚒𝚕𝚜. 𝙾𝚗𝚋𝚘𝚊𝚛𝚍𝚒𝚗𝚐 𝚒𝚜 𝚎𝚡𝚙𝚕𝚊𝚒𝚗𝚎𝚍 𝚍𝚒𝚏𝚏𝚎𝚛𝚎𝚗𝚝𝚕𝚢 𝚋𝚢 𝚎𝚟𝚎𝚛𝚢 𝚖𝚊𝚗𝚊𝚐𝚎𝚛. 𝙿𝚛𝚘𝚏𝚎𝚜𝚜𝚒𝚘𝚗𝚊𝚕 𝚟𝚒𝚍𝚎𝚘 𝚙𝚛𝚘𝚍𝚞𝚌𝚝𝚒𝚘𝚗 𝚒𝚜 𝚜𝚝𝚒𝚕𝚕 𝚜𝚕𝚘𝚠, 𝚎𝚡𝚙𝚎𝚗𝚜𝚒𝚟𝚎, 𝚊𝚗𝚍 𝚋𝚎𝚢𝚘𝚗𝚍 𝚝𝚑𝚎 𝚜𝚔𝚒𝚕𝚕 𝚜𝚎𝚝 𝚘𝚏 𝚖𝚘𝚜𝚝 𝚝𝚎𝚊𝚖𝚜．

𝙶𝚘𝚘𝚐𝚕𝚎 𝚅𝚒𝚍𝚜 𝚑𝚎𝚕𝚙𝚜 𝚌𝚕𝚘𝚜𝚎 𝚝𝚑𝚊𝚝 𝚐𝚊𝚙 𝚋𝚢 𝚖𝚊𝚔𝚒𝚗𝚐 𝚒𝚝 𝚎𝚊𝚜𝚒𝚎𝚛 𝚝𝚘 𝚝𝚞𝚛𝚗 𝚎𝚡𝚒𝚜𝚝𝚒𝚗𝚐 𝚌𝚘𝚗𝚝𝚎𝚗𝚝 𝚒𝚗𝚝𝚘 𝚊 𝚏𝚒𝚛𝚜𝚝-𝚍𝚛𝚊𝚏𝚝 𝚟𝚒𝚍𝚎𝚘. 𝚈𝚘𝚞 𝚌𝚊𝚗 𝚜𝚝𝚊𝚛𝚝 𝚠𝚒𝚝𝚑 𝚊 𝚙𝚛𝚘𝚖𝚙𝚝, 𝚞𝚜𝚎 𝚛𝚎𝚕𝚊𝚝𝚎𝚍 𝚏𝚒𝚕𝚎𝚜 𝚏𝚛𝚘𝚖 𝙶𝚘𝚘𝚐𝚕𝚎 𝙳𝚛𝚒𝚟𝚎 𝚜𝚞𝚌𝚑 𝚊𝚜 𝚊 𝙿𝙳𝙵, 𝚋𝚛𝚒𝚎𝚏, 𝚘𝚛 𝚜𝚕𝚒𝚍𝚎 𝚍𝚎𝚌𝚔, 𝚊𝚗𝚍 𝚅𝚒𝚍𝚜 𝚌𝚊𝚗 𝚑𝚎𝚕𝚙 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎 𝚊 𝚜𝚝𝚘𝚛𝚢𝚋𝚘𝚊𝚛𝚍, 𝚜𝚌𝚛𝚒𝚙𝚝 𝚜𝚞𝚐𝚐𝚎𝚜𝚝𝚒𝚘𝚗𝚜, 𝚜𝚝𝚘𝚌𝚔 𝚖𝚎𝚍𝚒𝚊, 𝚊𝚗𝚍 𝚟𝚘𝚒𝚌𝚎𝚘𝚟𝚎𝚛. 𝚃𝚑𝚎 𝚛𝚎𝚜𝚞𝚕𝚝 𝚒𝚜 𝚊 𝚖𝚞𝚌𝚑 𝚏𝚊𝚜𝚝𝚎𝚛 𝚜𝚝𝚊𝚛𝚝𝚒𝚗𝚐 𝚙𝚘𝚒𝚗𝚝 𝚏𝚘𝚛 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚟𝚒𝚍𝚎𝚘𝚜 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚗𝚎𝚎𝚍𝚒𝚗𝚐 𝚊 𝚏𝚞𝚕𝚕 𝚙𝚛𝚘𝚍𝚞𝚌𝚝𝚒𝚘𝚗 𝚝𝚎𝚊𝚖.

𝙷𝚘𝚠 𝚝𝚘 𝚊𝚌𝚌𝚎𝚜𝚜: 𝚟𝚒𝚍𝚜.𝚐𝚘𝚘𝚐𝚕𝚎.𝚌𝚘𝚖 (𝚒𝚗𝚌𝚕𝚞𝚍𝚎𝚍 𝚒𝚗 𝙶𝚘𝚘𝚐𝚕𝚎 𝚆𝚘𝚛𝚔𝚜𝚙𝚊𝚌𝚎 𝙱𝚞𝚜𝚒𝚗𝚎𝚜𝚜 𝚊𝚗𝚍 𝙴𝚗𝚝𝚎𝚛𝚙𝚛𝚒𝚜𝚎 𝚙𝚕𝚊𝚗𝚜)

𝙶𝚘𝚘𝚐𝚕𝚎 𝚅𝚒𝚍𝚜 𝚌𝚊𝚗 𝚑𝚎𝚕𝚙 𝚢𝚘𝚞:

• 𝚃𝚞𝚛𝚗 𝚍𝚘𝚌𝚞𝚖𝚎𝚗𝚝𝚜 𝚒𝚗𝚝𝚘 𝚝𝚛𝚊𝚒𝚗𝚒𝚗𝚐 𝚟𝚒𝚍𝚎𝚘𝚜: 𝚂𝚝𝚊𝚛𝚝 𝚏𝚛𝚘𝚖 𝚊 𝚙𝚘𝚕𝚒𝚌𝚢, 𝚖𝚊𝚗𝚞𝚊𝚕, 𝚘𝚛 𝚋𝚛𝚒𝚎𝚏 𝚊𝚗𝚍 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎 𝚊 𝚟𝚒𝚍𝚎𝚘 𝚍𝚛𝚊𝚏𝚝 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚋𝚎𝚐𝚒𝚗𝚗𝚒𝚗𝚐 𝚏𝚛𝚘𝚖 𝚜𝚌𝚛𝚊𝚝𝚌𝚑．

• 𝚂𝚌𝚊𝚕𝚎 𝚌𝚘𝚖𝚖𝚞𝚗𝚒𝚌𝚊𝚝𝚒𝚘𝚗𝚜 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚊 𝚌𝚛𝚎𝚠: 𝙲𝚛𝚎𝚊𝚝𝚎 𝚕𝚎𝚊𝚍𝚎𝚛𝚜𝚑𝚒𝚙 𝚞𝚙𝚍𝚊𝚝𝚎𝚜, 𝚌𝚘𝚖𝚙𝚕𝚒𝚊𝚗𝚌𝚎 𝚋𝚛𝚒𝚎𝚏𝚒𝚗𝚐𝚜, 𝚘𝚛 𝚙𝚛𝚘𝚍𝚞𝚌𝚝 𝚠𝚊𝚕𝚔𝚝𝚑𝚛𝚘𝚞𝚐𝚑𝚜 𝚞𝚜𝚒𝚗𝚐 𝚜𝚌𝚛𝚒𝚙𝚝𝚜, 𝚟𝚘𝚒𝚌𝚎𝚘𝚟𝚎𝚛, 𝚊𝚗𝚍 𝚜𝚝𝚘𝚌𝚔 𝚟𝚒𝚜𝚞𝚊𝚕𝚜.

• 𝙲𝚛𝚎𝚊𝚝𝚎 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚍𝚎𝚎𝚙 𝚌𝚛𝚎𝚊𝚝𝚒𝚟𝚎 𝚜𝚔𝚒𝚕𝚕𝚜: 𝚅𝚒𝚍𝚜 𝚑𝚎𝚕𝚙𝚜 𝚠𝚒𝚝𝚑 𝚝𝚑𝚎 𝚜𝚝𝚛𝚞𝚌𝚝𝚞𝚛𝚎 𝚊𝚗𝚍 𝚙𝚛𝚘𝚍𝚞𝚌𝚝𝚒𝚘𝚗 𝚍𝚛𝚊𝚏𝚝, 𝚠𝚑𝚒𝚕𝚎 𝚢𝚘𝚞𝚛 𝚝𝚎𝚊𝚖 𝚜𝚞𝚙𝚙𝚕𝚒𝚎𝚜 𝚝𝚑𝚎 𝚌𝚘𝚗𝚝𝚎𝚗𝚝 𝚊𝚗𝚍 𝚏𝚒𝚗𝚊𝚕 𝚛𝚎𝚟𝚒𝚎𝚠．

𝙴𝚡𝚊𝚖𝚙𝚕𝚎: 𝙰 𝚛𝚎𝚐𝚒𝚘𝚗𝚊𝚕 𝙷𝚁 𝚑𝚎𝚊𝚍 𝚊𝚝 𝚊 𝚖𝚊𝚗𝚞𝚏𝚊𝚌𝚝𝚞𝚛𝚒𝚗𝚐 𝚌𝚘𝚖𝚙𝚊𝚗𝚢 𝚒𝚜 𝚛𝚘𝚕𝚕𝚒𝚗𝚐 𝚘𝚞𝚝 𝚊 𝚛𝚎𝚟𝚒𝚜𝚎𝚍 𝚜𝚊𝚏𝚎𝚝𝚢 𝚙𝚛𝚘𝚝𝚘𝚌𝚘𝚕 𝚊𝚌𝚛𝚘𝚜𝚜 𝟷𝟸 𝚙𝚕𝚊𝚗𝚝 𝚕𝚘𝚌𝚊𝚝𝚒𝚘𝚗𝚜 𝚊𝚗𝚍 𝟸,𝟶𝟶𝟶 𝚠𝚘𝚛𝚔𝚎𝚛𝚜. 𝙲𝚘𝚘𝚛𝚍𝚒𝚗𝚊𝚝𝚒𝚗𝚐 𝚒𝚗-𝚙𝚎𝚛𝚜𝚘𝚗 𝚜𝚎𝚜𝚜𝚒𝚘𝚗𝚜 𝚊𝚌𝚛𝚘𝚜𝚜 𝚜𝚒𝚝𝚎𝚜 𝚠𝚘𝚞𝚕𝚍 𝚝𝚊𝚔𝚎 𝚠𝚎𝚎𝚔𝚜. 𝙰 𝚝𝚎𝚡𝚝 𝚌𝚒𝚛𝚌𝚞𝚕𝚊𝚛 𝚐𝚎𝚝𝚜 𝚒𝚐𝚗𝚘𝚛𝚎𝚍. 𝙰 𝚟𝚎𝚗𝚍𝚘𝚛-𝚙𝚛𝚘𝚍𝚞𝚌𝚎𝚍 𝚟𝚒𝚍𝚎𝚘 𝚠𝚘𝚞𝚕𝚍 𝚝𝚊𝚔𝚎 𝚝𝚘𝚘 𝚕𝚘𝚗𝚐.

𝙸𝚗𝚜𝚝𝚎𝚊𝚍, 𝚜𝚑𝚎 𝚘𝚙𝚎𝚗𝚜 𝙶𝚘𝚘𝚐𝚕𝚎 𝚅𝚒𝚍𝚜, 𝚊𝚝𝚝𝚊𝚌𝚑𝚎𝚜 𝚝𝚑𝚎 𝚞𝚙𝚍𝚊𝚝𝚎𝚍 𝚜𝚊𝚏𝚎𝚝𝚢 𝚖𝚊𝚗𝚞𝚊𝚕 𝚏𝚛𝚘𝚖 𝙳𝚛𝚒𝚟𝚎, 𝚊𝚗𝚍 𝚝𝚢𝚙𝚎𝚜:

“𝙲𝚛𝚎𝚊𝚝𝚎 𝚊 six-𝚜𝚌𝚎𝚗𝚎 𝚜𝚊𝚏𝚎𝚝𝚢 𝚝𝚛𝚊𝚒𝚗𝚒𝚗𝚐 𝚟𝚒𝚍𝚎𝚘 𝚏𝚘𝚛 𝚙𝚕𝚊𝚗𝚝 𝚏𝚕𝚘𝚘𝚛 𝚠𝚘𝚛𝚔𝚎𝚛𝚜. 𝙲𝚘𝚟𝚎𝚛 𝚝𝚑𝚎 𝚛𝚎𝚟𝚒𝚜𝚎𝚍 𝙿𝙿𝙴 𝚙𝚛𝚘𝚝𝚘𝚌𝚘𝚕, 𝚎𝚚𝚞𝚒𝚙𝚖𝚎𝚗𝚝 𝚑𝚊𝚗𝚍𝚕𝚒𝚗𝚐 𝚌𝚑𝚊𝚗𝚐𝚎𝚜, 𝚊𝚗𝚍 𝚎𝚖𝚎𝚛𝚐𝚎𝚗𝚌𝚢 𝚎𝚜𝚌𝚊𝚕𝚊𝚝𝚒𝚘𝚗 𝚜𝚝𝚎𝚙𝚜. 𝚃𝚘𝚗𝚎: 𝚜𝚎𝚛𝚒𝚘𝚞𝚜 𝚊𝚗𝚍 𝚌𝚕𝚎𝚊𝚛.”

𝚅𝚒𝚍𝚜 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎𝚜 𝚊 𝚍𝚛𝚊𝚏𝚝 𝚜𝚝𝚘𝚛𝚢𝚋𝚘𝚊𝚛𝚍 𝚊𝚗𝚍 𝚜𝚌𝚛𝚒𝚙𝚝 𝚜𝚞𝚐𝚐𝚎𝚜𝚝𝚒𝚘𝚗𝚜. 𝚂𝚑𝚎 𝚊𝚍𝚍𝚜 𝚊 𝚛𝚎𝚌𝚘𝚛𝚍𝚎𝚍 𝚒𝚗𝚝𝚛𝚘 𝚏𝚛𝚘𝚖 𝚝𝚑𝚎 𝚙𝚕𝚊𝚗𝚝 𝚑𝚎𝚊𝚍 𝚏𝚘𝚛 𝚌𝚛𝚎𝚍𝚒𝚋𝚒𝚕𝚒𝚝𝚢, 𝚛𝚎𝚟𝚒𝚎𝚠𝚜 𝚝𝚑𝚎 𝚗𝚊𝚛𝚛𝚊𝚝𝚒𝚘𝚗, 𝚊𝚗𝚍 𝚏𝚒𝚗𝚊𝚕𝚒s𝚎𝚜 𝚝𝚑𝚎 𝚟𝚒𝚍𝚎𝚘. 𝚃𝚑𝚎 𝚠𝚑𝚘𝚕𝚎 𝚙𝚛𝚘𝚌𝚎𝚜𝚜 𝚒𝚜 𝚖𝚞𝚌𝚑 𝚏𝚊𝚜𝚝𝚎𝚛 𝚝𝚑𝚊𝚗 𝚝𝚛𝚊𝚍𝚒𝚝𝚒𝚘𝚗𝚊𝚕 𝚙𝚛𝚘𝚍𝚞𝚌𝚝𝚒𝚘𝚗.

𝚆𝚑𝚊𝚝 𝚖𝚊𝚔𝚎𝚜 𝙶𝚘𝚘𝚐𝚕𝚎 𝚅𝚒𝚍𝚜 𝚜𝚙𝚎𝚌𝚒𝚊𝚕?

• 𝙳𝚘𝚌𝚞𝚖𝚎𝚗𝚝-𝚊𝚠𝚊𝚛𝚎 𝚍𝚛𝚊𝚏𝚝𝚒𝚗𝚐: 𝚄𝚜𝚎 𝚛𝚎𝚕𝚊𝚝𝚎𝚍 𝚏𝚒𝚕𝚎𝚜 𝚏𝚛𝚘𝚖 𝙳𝚛𝚒𝚟𝚎 𝚝𝚘 𝚑𝚎𝚕𝚙 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎 𝚝𝚑𝚎 𝚟𝚒𝚍𝚎𝚘 𝚍𝚛𝚊𝚏𝚝．

• 𝙰𝚕𝚕-𝚒𝚗-𝚘𝚗𝚎 𝚠𝚘𝚛𝚔𝚏𝚕𝚘𝚠: 𝚂𝚌𝚛𝚒𝚙𝚝, 𝚟𝚘𝚒𝚌𝚎𝚘𝚟𝚎𝚛, 𝚖𝚎𝚍𝚒𝚊 𝚜𝚞𝚐𝚐𝚎𝚜𝚝𝚒𝚘𝚗𝚜, 𝚊𝚗𝚍 𝚎𝚍𝚒𝚝𝚒𝚗𝚐 𝚑𝚊𝚙𝚙𝚎𝚗 𝚒𝚗 𝚘𝚗𝚎 𝚠𝚘𝚛𝚔𝚜𝚙𝚊𝚌𝚎 𝚝𝚘𝚘𝚕.

• 𝙱𝚞𝚒𝚕𝚝 𝚏𝚘𝚛 𝚠𝚘𝚛𝚔𝚙𝚕𝚊𝚌𝚎 𝚟𝚒𝚍𝚎𝚘𝚜: 𝙱𝚎𝚜𝚝 𝚜𝚞𝚒𝚝𝚎𝚍 𝚏𝚘𝚛 𝚝𝚛𝚊𝚒𝚗𝚒𝚗𝚐, 𝚘𝚗𝚋𝚘𝚊𝚛𝚍𝚒𝚗𝚐, 𝚞𝚙𝚍𝚊𝚝𝚎𝚜, 𝚊𝚗𝚍 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚌𝚘𝚖𝚖𝚞𝚗𝚒𝚌𝚊𝚝𝚒𝚘𝚗 𝚛𝚊𝚝𝚑𝚎𝚛 𝚝𝚑𝚊𝚗 𝚙𝚘𝚕𝚒𝚜𝚑𝚎𝚍 𝚜𝚝𝚞𝚍𝚒𝚘-𝚜𝚝𝚢𝚕𝚎 𝚖𝚊𝚛𝚔𝚎𝚝𝚒𝚗𝚐 𝚟𝚒𝚍𝚎𝚘𝚜.

𝙽𝚘𝚝𝚎: 𝚃𝚑𝚎 𝚝𝚘𝚘𝚕𝚜 𝚊𝚗𝚍 𝚊𝚗𝚊𝚕𝚢𝚜𝚒𝚜 𝚏𝚎𝚊𝚝𝚞𝚛𝚎𝚍 𝚒𝚗 𝚝𝚑𝚒𝚜 𝚜𝚎𝚌𝚝𝚒𝚘𝚗 𝚍𝚎𝚖𝚘𝚗𝚜𝚝𝚛𝚊𝚝𝚎𝚍 𝚌𝚕𝚎𝚊𝚛 𝚟𝚊𝚕𝚞𝚎 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚘𝚞𝚛 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚝𝚎𝚜𝚝𝚒𝚗𝚐. 𝙾𝚞𝚛 𝚛𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗𝚜 𝚊𝚛𝚎 𝚎𝚗𝚝𝚒𝚛𝚎𝚕𝚢 𝚒𝚗𝚍𝚎𝚙𝚎𝚗𝚍𝚎𝚗𝚝 𝚊𝚗𝚍 𝚗𝚘𝚝 𝚒𝚗𝚏𝚕𝚞𝚎𝚗𝚌𝚎𝚍 𝚋𝚢 𝚝𝚑𝚎 𝚝𝚘𝚘𝚕 𝚌𝚛𝚎𝚊𝚝𝚘𝚛𝚜.

AI BITS & BYTES

The New Yorker’s damning portrait of Altman

On 6 April, The New Yorker published a sweeping investigation into Sam Altman, built on more than 100 interviews, unpublished memos from former chief scientist Ilya Sutskever, and private notes from Dario Amodei. The report alleges a long-running pattern of deception at OpenAI, stretching across Altman’s career.

The account traces tensions back to his early startup Loopt, through clashes with partners at Y Combinator, and culminating in the recent OpenAI boardroom drama. You can read the New Yorker article here.

Central to the piece are Sutskever’s memos—compiled from 70 pages of Slack texts and HR documents—which claim that Altman misrepresented safety protocols to the board. Sutskever stepped down from the board in May 2024 after months of internal strain. Amodei’s long-kept private notes reportedly echo those concerns, distilling them into a blunt assessment: “The problem with OpenAI is Sam himself.”

The personal rift remains visible. At the recent India AI Impact Summit, the two were even prompted onstage to hold hands, a gesture they pointedly declined.

You may also want to read:

• The decade-long feud shaping the future of AI

• Microsoft’s new journey after reworked OpenAI deal

  

Anthropic’s Mythos, Glasswing and Meta Muse

Last week, we asked: What’s the mystery around Claude Mythos—Anthropic PBC’s latest AI model has been making waves on social media even before its official launch. Codenamed “Claude Mythos”, the model was accidentally revealed after descriptions of the model were stored in a publicly accessible data cache.

This week, Anthropic announced ‘Project Glasswing‘—its joint initiative with partners including Amazon Web Services, Apple, Microsoft, and Nvidia—to secure critical software using its unreleased AI model, Claude Mythos Preview. This model, claims Anthropic, can outperform most humans in finding and exploiting vulnerabilities. The project aims to deploy this capability defensively such as scanning systems, fixing flaws, and strengthening open-source infrastructure, before such tools spread to malicious actors.

Meanwhile, Meta has released Muse Spark—the first in the Muse family of models developed by Meta Superintelligence Labs. It is a multimodal reasoning model with support for tool-use, visual chain of thought, and multi-agent orchestration.

Muse Spark “is the first product of a ground-up overhaul of our AI efforts”, Meta says. A Contemplating mode allows Muse Spark “to compete with the extreme reasoning modes of frontier models such as Gemini Deep Think and GPT Pro”.

🔗 Using OpenClaw on Claude comes at a price

Intel + Terafab

Intel announced its partnership with Elon Musk’s SpaceX and Tesla on the Terafab project to build semiconductors. Musk launched Terafab last month, calling it the “most epic chip-building exercise in history”.

Terafab, claims Musk, is the only facility in the world to have the entire chip-making lifecycle under one roof, including memory, packaging, testing, and manufacturing of lithography masks. Read more.

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

In this week’s edition of Mint Tech Talk:

• How Generative AI is reshaping advertising

• AI Tool of the Week: Claude Dispatch

• Why netizens are fussing over Claude Mythos

• Googlers turning to Agent Smith to automate tasks

• Is AI to blame for Oracle layoffs?

PREVIOUSLY, ON MINT TECH TALK

Leader of the pack

Nearly four out of five AI-driven visits originate from ChatGPT, which generates about eight times more traffic than the next-largest player Google Gemini, according to SE Ranking. Through much of the first half of 2025, Gemini’s share hovered at 0.0083-0.0105%, even as ChatGPT’s share of global traffic more than doubled from 0.0799% to 0.1801%. By August, ChatGPT was sending 21.7 times more traffic to websites than Google Gemini.

Then, the momentum shifted.

In September 2025, ChatGPT traffic rose 53.8% but Gemini’s grew faster at 63.9%. By January 2026, Gemini’s traffic had more than doubled—around 12 times faster than its growth between January and October 2025—coinciding with the rollout of Gemini 3 models. It also surpassed Perplexity in global AI traffic.

ChatGPT remains the dominant platform but Google Gemini is growing at a faster clip. Advertisers are paying attention because the growth of AI referrals is steep. Adobe has reported a 10 times surge in Generative AI-driven referrals, while Similarweb estimates over a billion monthly visits now originate from AI tools.

AI & Advertising

Gen AI systems are replacing ranked links with synthesised answers, and advertising is moving upstream into those outputs, according to a BCG report.

Google’s AI Overviews provide an early signal. For exploratory and research-oriented queries, shopping and search ads now appear directly inside AI-generated responses, often ahead of organic results. Hence, visibility now relies less on keyword coverage and more on how well product data, creative, and bids align with model-level interpretation.

Microsoft Copilot is following a similar path, reframing search as an ongoing conversation built atop Bing’s existing ad ecosystem. According to BCG, Perplexity is testing a more explicit approach, surfacing sponsored follow-up questions that are clearly labeled and contextually generated, reflecting rising expectations for transparency in AI-generated experiences.

Additionally, conversational advertising is emerging as a distinct line item in the media mix, separate from search and programmatic, as brands reserve dollars to test, learn, and build early capability, notes BCG.

In the AI era, advertising inventory is emerging in three distinct forms:

• In-answer ads, embedded directly within synthesised responses when commercial intent is detected.

• In-conversation ads, appearing alongside dialogue during exploration—even for non-transactional queries.

• Agentic ads, where sponsored options surface as an AI Agent plans, compares, or completes tasks on a user’s behalf.

Agentic AI — Clicks to Outcomes

The advertising game is no longer just about traffic. It’s more about monetisation. In 2026, OpenAI began testing ads inside ChatGPT, reportedly crossing $100 million in annualised revenue within weeks and onboarding hundreds of advertisers, according to Reuters. Brands like Target, Ford, and Adobe have participated via agency partners such as WPP and Omnicom, The Verge and Adweek have reported. These are not experimental budgets but early bets on a new interface for discovery.

What advertisers are buying today is not “AI referrals” in the traditional sense. They are buying proximity to intent—placements that appear when a user asks a question with clear commercial relevance. It’s closer to search advertising than display, but a single prompt can collapse the entire funnel comprising discovery, comparison, and decision.

Agentic AI threatens to break current advertising models completely since such systems don’t just answer questions but also take actions. For instance, if you ask an AI Agent to “book the best hotel in Goa under ₹8,000”, instead of offering links, it could compare options, select one, and complete the transaction. This makes the very idea of “referral traffic” disappear since there are no clicks—simply decisions.

When Agentic AI systems begin to execute tasks end-to-end, marketers will care less about how many users visited their site and more about whether their product was selected by the AI Agent in the first place. This creates a new competitive dynamic. The traditional search offered blue links; even dominating one position left room for others. AI answers, by contrast, often surface just a handful of options—sometimes only one.

There are early hints of how this might play out. The e-commerce platforms already operate on “winner-takes-most” mechanics, such as Amazon’s Buy Box, where a single seller captures the majority of transactions. Agentic AI could extend that logic across the web. Instead of competing for page rank, brands may compete to become the default choice in an algorithmic decision.

None of this will happen overnight

Today’s AI still frequently defers to links, and ad formats remain limited and geographically constrained. In markets like India, monetisation is nascent. Attribution is murky, and user trust remains fragile, especially if paid placements blur into organic answers.

But the trajectory is clear. In the short term, AI usage will drive more referral traffic as users click through for deeper information. In the medium term, better answers will reduce the need to click. And in the long term, as agentic capabilities mature, the click itself may disappear. For advertisers, that is the real inflection point.

The question is no longer how to attract attention, but how to influence decisions made by machines on behalf of users. AI referral traffic, then, will serve as the bridge from an internet built on clicks to one built on choices.

AI TOOL OF THE WEEK

By AI&Beyond, with Jaspreet Bindra and Anuj Magazine

𝚃𝚑𝚎 𝙰𝙸 𝚏𝚎𝚊𝚝𝚞𝚛𝚎 𝚠𝚎 𝚞𝚗𝚕𝚘𝚌𝚔𝚎𝚍 𝚝𝚘𝚍𝚊𝚢: 𝙲𝚕𝚊𝚞𝚍𝚎 𝙳𝚒𝚜𝚙𝚊𝚝𝚌𝚑

𝚆𝚑𝚊𝚝 𝚙𝚛𝚘𝚋𝚕𝚎𝚖 𝚍𝚘𝚎𝚜 𝚒𝚝 𝚜𝚘𝚕𝚟𝚎? 𝙿𝚒𝚌𝚝𝚞𝚛𝚎 𝚝𝚑𝚒𝚜: 𝚊 𝚜𝚎𝚗𝚒𝚘𝚛 𝚖𝚊𝚗𝚊𝚐𝚎𝚛 𝚐𝚎𝚝𝚜 𝚘𝚞𝚝 𝚘𝚏 𝚊 𝚍𝚒𝚏𝚏𝚒𝚌𝚞𝚕𝚝 𝚌𝚕𝚒𝚎𝚗𝚝 𝚖𝚎𝚎𝚝𝚒𝚗𝚐 𝚊𝚗𝚍 𝚐𝚎𝚝𝚜 𝚒𝚗𝚝𝚘 𝚊 𝚌𝚊𝚋, 𝟺𝟻 𝚖𝚒𝚗𝚞𝚝𝚎𝚜 𝚏𝚛𝚘𝚖 𝚝𝚑𝚎 𝚘𝚏𝚏𝚒𝚌𝚎． 𝚂𝚑𝚎 𝚛𝚎𝚖𝚎𝚖𝚋𝚎𝚛𝚜 𝚝𝚑𝚊𝚝 𝚑𝚎𝚛 𝙲𝙵𝙾 𝚗𝚎𝚎𝚍𝚜 𝚝𝚑𝚎 𝚖𝚘𝚗𝚝𝚑𝚕𝚢 𝚖𝚊𝚗𝚊𝚐𝚎𝚖𝚎𝚗𝚝 𝚙𝚊𝚌𝚔—𝚍𝚊𝚝𝚊 𝚙𝚞𝚕𝚕𝚎𝚍 𝚏𝚛𝚘𝚖 𝚜𝚒𝚡 𝚍𝚎𝚙𝚊𝚛𝚝𝚖𝚎𝚗𝚝 𝙴𝚡𝚌𝚎𝚕 𝚏𝚒𝚕𝚎𝚜, 𝚏𝚕𝚊𝚐𝚐𝚎𝚍 𝚏𝚘𝚛 𝚟𝚊𝚛𝚒𝚊𝚗𝚌𝚎𝚜, 𝚏𝚘𝚛𝚖𝚊𝚝𝚝𝚎𝚍 𝚊𝚗𝚍 𝚛𝚎𝚊𝚍𝚢—𝚋𝚢 𝚖𝚘𝚛𝚗𝚒𝚗𝚐. 𝙷𝚎𝚛 𝚕𝚊𝚙𝚝𝚘𝚙 𝚒𝚜 𝚋𝚊𝚌𝚔 𝚊𝚝 𝚑𝚎𝚛 𝚍𝚎𝚜𝚔. 𝙷𝚎𝚛 𝚝𝚎𝚊𝚖 𝚒𝚜 𝚞𝚗𝚊𝚟𝚊𝚒𝚕𝚊𝚋𝚕𝚎. 𝚃𝚑𝚒𝚜 𝚒𝚜 𝚗𝚘𝚝 𝚊 𝚏𝚛𝚒𝚗𝚐𝚎 𝚜𝚒𝚝𝚞𝚊𝚝𝚒𝚘𝚗. 𝚃𝚑𝚒𝚜 𝚒𝚜 𝚃𝚞𝚎𝚜𝚍𝚊𝚢.

𝚄𝚗𝚝𝚒𝚕 𝚗𝚘𝚠, 𝚝𝚑𝚎 𝚘𝚙𝚝𝚒𝚘𝚗𝚜 𝚠𝚎𝚛𝚎 𝚕𝚒𝚖𝚒𝚝𝚎𝚍: W𝚊𝚒𝚝 𝚝𝚒𝚕𝚕 𝚜𝚑𝚎 𝚐𝚘𝚝 𝚋𝚊𝚌𝚔, 𝚌𝚊𝚕𝚕 𝚜𝚘𝚖𝚎𝚘𝚗𝚎, 𝚘𝚛 𝚜𝚔𝚒𝚙 𝚝𝚑𝚎 𝚙𝚛𝚎𝚙 𝚊𝚕𝚝𝚘𝚐𝚎𝚝𝚑𝚎𝚛. 𝙰𝙸 𝚝𝚘𝚘𝚕𝚜 𝚑𝚊𝚟𝚎𝚗’𝚝 𝚜𝚘𝚕𝚟𝚎𝚍 𝚝𝚑𝚒𝚜 𝚋𝚎𝚌𝚊𝚞𝚜𝚎 𝚝𝚑𝚎𝚢 𝚕𝚒𝚟𝚎 𝚒𝚗 𝚊 𝚋𝚛𝚘𝚠𝚜𝚎𝚛 𝚝𝚊𝚋 𝚝𝚑𝚊𝚝 𝚗𝚎𝚎𝚍𝚜 𝚢𝚘𝚞 𝚜𝚒𝚝𝚝𝚒𝚗𝚐 𝚒𝚗 𝚏𝚛𝚘𝚗𝚝 𝚘𝚏 𝚒𝚝．

𝙲𝚕𝚊𝚞𝚍𝚎 𝙳𝚒𝚜𝚙𝚊𝚝𝚌𝚑 𝚌𝚑𝚊𝚗𝚐𝚎𝚜 𝚝𝚑𝚊𝚝. 𝙸𝚝 𝚕𝚎𝚝𝚜 𝚢𝚘𝚞 𝚜𝚎𝚗𝚍 𝚊 𝚝𝚊𝚜𝚔 𝚝𝚘 𝚢𝚘𝚞𝚛 𝚍𝚎𝚜𝚔𝚝𝚘𝚙 𝙰𝙸 𝚏𝚛𝚘𝚖 𝚢𝚘𝚞𝚛 𝚙𝚑𝚘𝚗𝚎 𝚊𝚗𝚍 𝚌𝚘𝚖𝚎 𝚋𝚊𝚌𝚔 𝚝𝚘 𝚝𝚑𝚎 𝚏𝚒𝚗𝚒𝚜𝚑𝚎𝚍 𝚠𝚘𝚛𝚔. 𝙲𝚕𝚊𝚞𝚍𝚎 𝚛𝚞𝚗𝚜 𝚘𝚗 𝚢𝚘𝚞𝚛 𝚌𝚘𝚖𝚙𝚞𝚝𝚎𝚛, 𝚛𝚎𝚊𝚍𝚜 𝚢𝚘𝚞𝚛 𝚊𝚌𝚝𝚞𝚊𝚕 𝚏𝚒𝚕𝚎𝚜, 𝚞𝚜𝚎𝚜 𝚢𝚘𝚞𝚛 𝚌𝚘𝚗𝚗𝚎𝚌𝚝𝚎𝚍 𝚝𝚘𝚘𝚕𝚜, 𝚊𝚗𝚍 𝚍𝚎𝚕𝚒𝚟𝚎𝚛𝚜 𝚝𝚑𝚎 𝚘𝚞𝚝𝚙𝚞𝚝 𝚋𝚊𝚌𝚔 𝚝𝚘 𝚝𝚑𝚎 𝚜𝚊𝚖𝚎 𝚌𝚘𝚗𝚟𝚎𝚛𝚜𝚊𝚝𝚒𝚘𝚗. 𝚈𝚘𝚞 𝚍𝚘𝚗’𝚝 𝚠𝚊𝚝𝚌𝚑 𝚒𝚝 𝚠𝚘𝚛𝚔. 𝚈𝚘𝚞 𝚓𝚞𝚜𝚝 𝚛𝚎𝚝𝚞𝚛𝚗 𝚝𝚘 𝚛𝚎𝚜𝚞𝚕𝚝𝚜.

𝙷𝚘𝚠 𝚝𝚘 𝚊𝚌𝚌𝚎𝚜𝚜: 𝙳𝚘𝚠𝚗𝚕𝚘𝚊𝚍 𝙲𝚕𝚊𝚞𝚍𝚎 𝙳𝚎𝚜𝚔𝚝𝚘𝚙 + 𝙲𝚕𝚊𝚞𝚍𝚎 𝚖𝚘𝚋𝚒𝚕𝚎 𝚊𝚙𝚙. 𝙴𝚗𝚊𝚋𝚕𝚎 𝙳𝚒𝚜𝚙𝚊𝚝𝚌𝚑 𝚒𝚗𝚜𝚒𝚍𝚎 𝚝𝚑𝚎 𝙲𝚘𝚠𝚘𝚛𝚔 𝚝𝚊𝚋.

𝙲𝚕𝚊𝚞𝚍𝚎 𝙳𝚒𝚜𝚙𝚊𝚝𝚌𝚑 𝚑𝚎𝚕𝚙𝚜 𝚢𝚘𝚞 𝚠𝚒𝚝𝚑:

• 𝚂𝚒𝚕𝚎𝚗𝚝 𝙲𝚘𝚖𝚙𝚒𝚕𝚎𝚛: 𝙲𝚘𝚖𝚙𝚒𝚕𝚎 𝚖𝚞𝚕𝚝𝚒-𝚏𝚒𝚕𝚎 𝚛𝚎𝚙𝚘𝚛𝚝𝚜 𝚠𝚑𝚒𝚕𝚎 𝚢𝚘𝚞’𝚛𝚎 𝚒𝚗 𝚝𝚛𝚊𝚗𝚜𝚒𝚝 𝚘𝚛 𝚋𝚎𝚝𝚠𝚎𝚎𝚗 𝚖𝚎𝚎𝚝𝚒𝚗𝚐𝚜.

• 𝙸𝚗𝚋𝚘𝚡 𝚂𝚎𝚗𝚝𝚒𝚗𝚎𝚕: 𝚃𝚛𝚒𝚊𝚐𝚎 𝚢𝚘𝚞𝚛 𝚒𝚗𝚋𝚘𝚡 𝚊𝚗𝚍 𝚏𝚕𝚊𝚐 𝚞𝚛𝚐𝚎𝚗𝚝 𝚎𝚖𝚊𝚒𝚕𝚜 𝚋𝚎𝚏𝚘𝚛𝚎 𝚢𝚘𝚞𝚛 𝚍𝚊𝚢 𝚋𝚎𝚐𝚒𝚗𝚜.

• 𝚁𝚎𝚌𝚞𝚛𝚛𝚒𝚗𝚐 𝙿𝚒𝚕𝚘𝚝: 𝚁𝚞𝚗 𝚛𝚎𝚌𝚞𝚛𝚛𝚒𝚗𝚐 𝚝𝚊𝚜𝚔𝚜 (𝚕𝚒𝚔𝚎 𝚍𝚊𝚒𝚕𝚢 𝚜𝚝𝚊𝚝𝚞𝚜 𝚛𝚎𝚙𝚘𝚛𝚝𝚜) 𝚘𝚗 𝚊 𝚜𝚌𝚑𝚎𝚍𝚞𝚕𝚎, 𝚖𝚘𝚗𝚒𝚝𝚘𝚛𝚎𝚍 𝚏𝚛𝚘𝚖 𝚢𝚘𝚞𝚛 𝚙𝚑𝚘𝚗𝚎.

𝙴𝚡𝚊𝚖𝚙𝚕𝚎: 𝙰 𝚌𝚘𝚗𝚜𝚞𝚕𝚝𝚊𝚗𝚝 𝚑𝚎𝚊𝚍𝚒𝚗𝚐 𝚒𝚗𝚝𝚘 𝚊 𝚌𝚕𝚒𝚎𝚗𝚝 𝚙𝚒𝚝𝚌𝚑 𝚛𝚎𝚊𝚕𝚒𝚣𝚎𝚜 𝚜𝚑𝚎 𝚑𝚊𝚜𝚗’𝚝 𝚏𝚞𝚕𝚕𝚢 𝚙𝚛𝚎𝚙𝚙𝚎𝚍 𝚘𝚗 𝚊 𝚌𝚘𝚖𝚙𝚎𝚝𝚒𝚝𝚘𝚛. 𝙵𝚛𝚘𝚖 𝚑𝚎𝚛 𝚙𝚑𝚘𝚗𝚎:

“𝙲𝚑𝚎𝚌𝚔 𝚖𝚢 /𝚁𝚎𝚜𝚎𝚊𝚛𝚌𝚑 𝚏𝚘𝚕𝚍𝚎𝚛 𝚊𝚗𝚍 𝚎𝚖𝚊𝚒𝚕𝚜 𝚏𝚛𝚘𝚖 𝚘𝚞𝚛 𝚜𝚎𝚌𝚝𝚘𝚛 𝚕𝚎𝚊𝚍. 𝙲𝚘𝚖𝚙𝚒𝚕𝚎 𝚊 𝟷-𝚙𝚊𝚐𝚎 𝚜𝚗𝚊𝚙𝚜𝚑𝚘𝚝 𝚊𝚗𝚍 𝚜𝚊𝚟𝚎 𝚝𝚘 /𝙿𝚒𝚝𝚌𝚑-𝙿𝚛𝚎𝚙.”

𝙱𝚢 𝚝𝚑𝚎 𝚝𝚒𝚖𝚎 𝚜𝚑𝚎 𝚠𝚊𝚕𝚔𝚜 𝚒𝚗𝚝𝚘 𝚝𝚑𝚎 𝚛𝚘𝚘𝚖, 𝚒𝚝’𝚜 𝚍𝚘𝚗𝚎.

𝙰𝚗 𝙷𝚁 𝚖𝚊𝚗𝚊𝚐𝚎𝚛 𝚠𝚛𝚊𝚙𝚙𝚒𝚗𝚐 𝚞𝚙 𝚋𝚊𝚌𝚔-𝚝𝚘-𝚋𝚊𝚌𝚔 𝚒𝚗𝚝𝚎𝚛𝚟𝚒𝚎𝚠𝚜 𝚗𝚎𝚎𝚍𝚜 𝚝𝚑𝚛𝚎𝚎 𝚌𝚊𝚗𝚍𝚒𝚍𝚊𝚝𝚎 𝚋𝚛𝚒𝚎𝚏𝚒𝚗𝚐 𝚙𝚊𝚌𝚔𝚜 𝚛𝚎𝚊𝚍𝚢 𝚏𝚘𝚛 𝚝𝚘𝚖𝚘𝚛𝚛𝚘𝚠’𝚜 𝚏𝚒𝚗𝚊𝚕 𝚛𝚘𝚞𝚗𝚍. 𝙵𝚛𝚘𝚖 𝚑𝚎𝚛 𝚙𝚑𝚘𝚗𝚎:

“𝙿𝚞𝚕𝚕 𝚝𝚑𝚎 𝚜𝚑𝚘𝚛𝚝𝚕𝚒𝚜𝚝𝚎𝚍 𝙲𝚅𝚜, 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎 𝚒𝚗𝚝𝚎𝚛𝚟𝚒𝚎𝚠 𝚚𝚞𝚎𝚜𝚝𝚒𝚘𝚗𝚜 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚝𝚑𝚎 𝙹𝙳, 𝚊𝚝𝚝𝚊𝚌𝚑 𝚝𝚑𝚎 𝚒𝚗𝚝𝚎𝚛𝚟𝚒𝚎𝚠 𝚗𝚘𝚝𝚎𝚜, 𝚊𝚗𝚍 𝚜𝚊𝚟𝚎 𝚘𝚗𝚎 𝚋𝚛𝚒𝚎𝚏𝚒𝚗𝚐 𝚍𝚘𝚌 𝚙𝚎𝚛 𝚌𝚊𝚗𝚍𝚒𝚍𝚊𝚝𝚎 𝚝𝚘 /𝙱𝚛𝚒𝚎𝚏𝚒𝚗𝚐-𝙿𝚊𝚌𝚔𝚜”

𝚂𝚑𝚎’𝚜 𝚑𝚘𝚖𝚎 𝚋𝚢 𝚝𝚑𝚎 𝚝𝚒𝚖𝚎 𝚝𝚑𝚎𝚢’𝚛𝚎 𝚛𝚎𝚊𝚍𝚢 𝚝𝚘 𝚜𝚑𝚊𝚛𝚎．

𝚆𝚑𝚊𝚝 𝚖𝚊𝚔𝚎𝚜 𝙲𝚕𝚊𝚞𝚍𝚎 𝙳𝚒𝚜𝚙𝚊𝚝𝚌𝚑 𝚜𝚙𝚎𝚌𝚒𝚊𝚕?

• 𝚃𝚛𝚞𝚕𝚢 𝚊𝚜𝚢𝚗𝚌: 𝚈𝚘𝚞 𝚊𝚜𝚜𝚒𝚐𝚗 𝚝𝚑𝚎 𝚝𝚊𝚜𝚔 𝚊𝚗𝚍 𝚕𝚎𝚊𝚟𝚎. 𝙲𝚕𝚊𝚞𝚍𝚎 𝚍𝚘𝚎𝚜𝚗’𝚝 𝚗𝚎𝚎𝚍 𝚢𝚘𝚞 𝚊𝚛𝚘𝚞𝚗𝚍.

• 𝚈𝚘𝚞𝚛 𝚏𝚒𝚕𝚎𝚜, 𝚢𝚘𝚞𝚛 𝚖𝚊𝚌𝚑𝚒𝚗𝚎: 𝙻𝚘𝚌𝚊𝚕 𝚙𝚛𝚘𝚌𝚎𝚜𝚜𝚒𝚗𝚐 𝚖𝚎𝚊𝚗𝚜 𝚜𝚎𝚗𝚜𝚒𝚝𝚒𝚟𝚎 𝚍𝚊𝚝𝚊 𝚜𝚝𝚊𝚢𝚜 𝚠𝚒𝚝𝚑 𝚢𝚘𝚞.

• 𝙾𝚗𝚎 𝚙𝚎𝚛𝚜𝚒𝚜𝚝𝚎𝚗𝚝 𝚝𝚑𝚛𝚎𝚊𝚍: 𝚈𝚘𝚞𝚛 𝚙𝚑𝚘𝚗𝚎 𝚊𝚗𝚍 𝚍𝚎𝚜𝚔𝚝𝚘𝚙 𝚜𝚝𝚊𝚢 𝚒𝚗 𝚜𝚢𝚗𝚌—𝚗𝚘 𝚜𝚠𝚒𝚝𝚌𝚑𝚒𝚗𝚐 𝚊𝚙𝚙𝚜, 𝚗𝚘 𝚌𝚘𝚙𝚢-𝚙𝚊𝚜𝚝𝚒𝚗𝚐 𝚋𝚎𝚝𝚠𝚎𝚎𝚗 𝚍𝚎𝚟𝚒𝚌𝚎𝚜.

𝙽𝚘𝚝𝚎: 𝚃𝚑𝚎 𝚝𝚘𝚘𝚕𝚜 𝚊𝚗𝚍 𝚊𝚗𝚊𝚕𝚢𝚜𝚒𝚜 𝚏𝚎𝚊𝚝𝚞𝚛𝚎𝚍 𝚒𝚗 𝚝𝚑𝚒𝚜 𝚜𝚎𝚌𝚝𝚒𝚘𝚗 𝚍𝚎𝚖𝚘𝚗𝚜𝚝𝚛𝚊𝚝𝚎𝚍 𝚌𝚕𝚎𝚊𝚛 𝚟𝚊𝚕𝚞𝚎 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚘𝚞𝚛 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚝𝚎𝚜𝚝𝚒𝚗𝚐. 𝙾𝚞𝚛 𝚛𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗𝚜 𝚊𝚛𝚎 𝚎𝚗𝚝𝚒𝚛𝚎𝚕𝚢 𝚒𝚗𝚍𝚎𝚙𝚎𝚗𝚍𝚎𝚗𝚝 𝚊𝚗𝚍 𝚗𝚘𝚝 𝚒𝚗𝚏𝚕𝚞𝚎𝚗𝚌𝚎𝚍 𝚋𝚢 𝚝𝚑𝚎 𝚝𝚘𝚘𝚕 𝚌𝚛𝚎𝚊𝚝𝚘𝚛𝚜.

AI BITS & BYTES

Elon Musk doubles down on Grok Imagine after Sora shutdown

SpaceX CEO Elon Musk has reacted to the shutdown of OpenAI Sora, claiming that that video-generation tools will be essential for achieving AGI.

Musk was responding to a post by a user who questioned why the billionaire continued to invest in video-generation tools, given that rivals like OpenAI had shut down the Sora service after reportedly losing $1 million per day. The billionaire noted that xAI remains heavily involved in video generation through Grok Imagine, because he believes it is a path to achieving AGI. Musk hypothesised that visual data carries much more information than text or other inputs. Read more.

What’s the mystery around Claude Mythos?

Anthropic’s latest AI model has been making waves on social media even before its official launch. The new model, codenamed “Claude Mythos”, was accidentally revealed after descriptions of the model were stored in a publicly accessible data cache that was first reported by Fortune.

After the news of the AI model leak came to light, an Anthropic spokesperson confirmed its existence and noted that the model represented “a step change” in AI performance and was “the most capable we’ve built to date”. Anthropic believes its upcoming model poses unprecedented cybersecurity risks. Read more.

Googlers are turning to Agent Smith to automate tasks

Google employees have begun using a new internal AI tool called “Agent Smith” that can automate tasks like coding, according to a report by Business Insider.

The AI agent has become so popular inside the company that the tech giant had to restrict access to handle the sudden influx of employees using it. Agent Smith is likely a reference to the popular character from the movie The Matrix, where Smith is a software agent inside the simulated world designed to monitor, control, and maintain order, and is capable of acting independently and intelligently. Read more.

Meta readies two prescription-supported Ray-Ban AI glasses

Meta Platforms Inc. plans to launch two new Ray-Ban smart glasses next week, especially aimed at prescription wearers. According to a report by Bloomberg’s Mark Gurman, the glasses will be sold alongside EssilorLuxottica SA and will come with prescription support.

This will be the first time that Meta will be introducing prescription support for its Ray-Ban series. The glasses, codenamed Scriber and Blazer, were earlier spotted in the US Federal Communications Commission filings by The Verge. Read more.

Oracle layoffs hit 12,000 employees in India

Oracle is laying off thousands of employees to cut costs, CNBC reported. Reports suggest the number can be as high as 30,000, with ~12,000 employees affected in India. Oracle had a headcount of 162,000 people as of May 2025, according to its most recent 10-K filing.

The Oracle layoffs come amid steps to increase spending on AI infrastructure in an effort to better compete with cloud rivals, such as Google Cloud and Amazon Web Services. You may read the full story here. Still, a detailed study by Yale University’s Budget Lab suggests the recent tech layoffs is not necessarily attributable to AI.

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

Context first. Fridman asked the Nvidia Corp. CEO if Agentic AI can build and run a billion-dollar business. The reference point was OpenClaw, an open-source agent framework that was recently acquired by OpenAI. Huang, who describes OpenClaw as “the next ChatGPT”, has already unveiled a software development tool kit, NemoClaw, aimed at making OpenClaw agents enterprise-ready.

Discussing OpenClaw, Huang acknowledged that some users could deploy it to build apps that gain rapid traction. But he added that even 100,000 such agents would not be enough to build an Nvidia. He also pointed out a paradox emerging in sectors such as radiology: Despite AI’s growing accuracy, fears of job loss have contributed to a shortage of radiologists. That, he argued, is counterproductive — AI can help clinicians scan faster and improve diagnosis.

Huang is right about accounting for the alarm and hype around AGI when discussing the benefits of advanced AI models and Agentic AI, especially given that there is no single definition today.

The reason is that systems like Anthropic’s Claude Code and Claude Cowork can already interact with files, browsers, and developer tools. Meanwhile, Meta’s Mark Zuckerberg is reportedly building a “CEO Agent” to compress decision-making layers. And OpenAI is working toward an autonomous “AI research intern”.

The trajectory is clear: AI is becoming more proactive, more embedded, and more capable of executing multi-step tasks, but none of this necessarily amounts to AGI, which typically is a point when machines become as good as humans at performing tasks—if not better.

Agentic AI — while action-oriented and not essentially prompt-driven — is scaffolded by predefined tools, guardrails, and training regimes. Remove the structure, and their limitations quickly show up as hallucinations, brittle reasoning, and poor long-horizon judgement.

In this week’s edition of Mint Tech Talk:

• Agentic AI’s IQ cannot match Human Intelligence

• The noble gas that can quietly choke the AI boom

• AI Tool of the Week: Google Stitch’s new Vibe Design

PREVIOUSLY, ON MINT TECH TALK

Machine IQ or illusion?

Meanwhile, Intelligence Quotient (IQ) has entered the race to define AI progress. Claims that models like Claude Opus 4.6 score 133, GPT-5.2 Thinking hits 141, and Gemini 3 Pro reaches 142 on the Mensa Norway test are hard to ignore. They suggest machines are not just improving but rivalling—even surpassing—human intelligence.

The reality is less compelling. IQ is a human construct, shaped by the Flynn effect. It measures how individuals perform relative to other people, against population norms built over decades. A score of 130 places a person in the top percentile. Machines, however, are not part of that population, making direct comparisons questionable.

The Mensa Norway exam itself leans heavily on pattern recognition (visual matrices and abstract sequences), an area where modern AI excels. Trained on vast datasets rich in similar structures, these systems are specialists masquerading as generalists. High scores here reflect training alignment, not broad reasoning ability. Further, even if models haven’t seen the exact questions, they have likely encountered multiple variations. It’s akin to testing a student after weeks of drilling near-identical problems.

To reiterate, while AI IQ leaps invite comparisons with human cognitive gains, we must realise that AI improves because engineers redesign architectures, scale compute, and refine data as opposed to the Flynn effect that accounts for humans improving gradually through better nutrition, education, and environments.

Effectively, while today’s systems can ace pattern puzzles, they continue to stumble on basic reasoning, long-horizon planning, or real-world judgment. Artificial General Intelligence, or AGI, implies generality with the ability to adapt across domains with consistent, reliable competence, much like a human. Today’s agents, for all their sophistication, are still context-bound optimisers. They can execute workflows impressively but struggle with ambiguity, causality, and real-world grounding.

It’s important to recognise that “AGI” is increasingly being used to describe systems that are economically useful or operationally autonomous, rather than cognitively general. By that looser standard, today’s systems may qualify. But equating autonomy with general intelligence is premature. It’s tempting to believe that AGI has been “achieved” but the reality is more about semantics than human intelligence.

The noble gas that could quietly choke the AI boom

While an LPG shortage in India continues to make headlines, there’s another gas which rarely features in conversations. And while helium has nothing to do with the Iran war, like the LPG crisis does, it has much to do with AI.

This inert gas, better known for balloons, sits deep inside the global semiconductor supply chain, and disruptions linked to the Strait of Hormuz are bringing its importance into sharper focus.

Helium is essential for chip manufacturing. It is used to cool wafers, stabilise ultra-clean environments, and support critical processes like etching and deposition. There are few viable substitutes at scale, making supply reliability crucial for semiconductor fabs.

A significant share of global helium comes from Qatar, where it is extracted as a by-product of liquefied natural gas. Much of this supply must transit through the Strait of Hormuz, which explains why any prolonged disruption here risks hitting both production and transportation simultaneously.

What heightens concern is helium’s physical nature. Unlike oil or gas, it is not easy to store. It requires specialised infrastructure and tends to escape over time due to its low density. As a result, stockpiling is limited, and supply shocks can translate into shortages faster than in other commodities.

For chipmakers, this creates a lagged but tangible risk. Most fabs maintain only weeks or months of helium inventory. A short disruption may raise costs, but a prolonged one could force production adjustments or slowdowns. In an industry already stretched by demand for AI chips, that matters.

For now, helium remains a background risk but in a fragile, globally interconnected supply chain, even an invisible gas can become a critical fault line.

AI TOOL OF THE WEEK

By AI&Beyond, with Jaspreet Bindra and Anuj Magazine

𝚃𝚑𝚎 𝙰𝙸 𝚑𝚊𝚌𝚔 𝚠𝚎 𝚞𝚗𝚕𝚘𝚌𝚔 𝚝𝚘𝚍𝚊𝚢 𝚒𝚜 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚊 𝚝𝚘𝚘𝚕: 𝙶𝚘𝚘𝚐𝚕𝚎 𝚂𝚝𝚒𝚝𝚌𝚑, 𝚊𝚗𝚍 𝚝𝚑𝚎 𝚛𝚒𝚜𝚎 𝚘𝚏 𝚅𝚒𝚋𝚎 𝙳𝚎𝚜𝚒𝚐𝚗.

𝚆𝚑𝚊𝚝 𝚙𝚛𝚘𝚋𝚕𝚎𝚖𝚜 𝚍𝚘𝚎𝚜 𝙶𝚘𝚘𝚐𝚕𝚎 𝚂𝚝𝚒𝚝𝚌𝚑’𝚜 𝚗𝚎𝚠 𝚅𝚒𝚋𝚎 𝙳𝚎𝚜𝚒𝚐𝚗 𝚞𝚙𝚍𝚊𝚝𝚎 𝚜𝚘𝚕𝚟𝚎? 𝙸𝚖𝚊𝚐𝚒𝚗𝚎 𝚝𝚑𝚒𝚜: 𝙰 𝚙𝚛𝚘𝚍𝚞𝚌𝚝 𝚖𝚊𝚗𝚊𝚐𝚎𝚛 𝚑𝚊𝚜 𝚊 𝚜𝚙𝚛𝚒𝚗𝚝 𝚙𝚕𝚊𝚗𝚗𝚒𝚗𝚐 𝚖𝚎𝚎𝚝𝚒𝚗𝚐 𝚝𝚘𝚖𝚘𝚛𝚛𝚘𝚠. 𝚃𝚑𝚎 𝚍𝚎𝚜𝚒𝚐𝚗𝚎𝚛 𝚒𝚜 𝚘𝚗 𝚕𝚎𝚊𝚟𝚎. 𝚃𝚑𝚎 𝚍𝚎𝚟𝚎𝚕𝚘𝚙𝚎𝚛 𝚠𝚘𝚗’𝚝 𝚜𝚝𝚊𝚛𝚝 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚊 𝚟𝚒𝚜𝚞𝚊𝚕 𝚜𝚙𝚎𝚌. 𝚃𝚑𝚎 𝚒𝚍𝚎𝚊 𝚒𝚜 𝚌𝚕𝚎𝚊𝚛 𝚒𝚗 𝚝𝚑𝚎 𝙿𝙼’𝚜 𝚑𝚎𝚊𝚍 𝚋𝚞𝚝 𝚜𝚝𝚞𝚌𝚔 𝚝𝚑𝚎𝚛𝚎.

𝚃𝚑𝚒𝚜 𝚌𝚢𝚌𝚕𝚎: 𝚒𝚍𝚎𝚊 → 𝚖𝚒𝚜𝚌𝚘𝚖𝚖𝚞𝚗𝚒𝚌𝚊𝚝𝚒𝚘𝚗 → 𝚛𝚎𝚠𝚘𝚛𝚔 — 𝚚𝚞𝚒𝚎𝚝𝚕𝚢 𝚌𝚘𝚜𝚝𝚜 𝚝𝚎𝚊𝚖𝚜 𝚠𝚎𝚎𝚔𝚜 𝚎𝚟𝚎𝚛𝚢 𝚚𝚞𝚊𝚛𝚝𝚎𝚛. 𝙸𝚝’𝚜 𝚗𝚘𝚝 𝚊 𝚍𝚎𝚜𝚒𝚐𝚗 𝚙𝚛𝚘𝚋𝚕𝚎𝚖. 𝙸𝚝’𝚜 𝚊 𝚝𝚛𝚊𝚗𝚜𝚕𝚊𝚝𝚒𝚘𝚗 𝚙𝚛𝚘𝚋𝚕𝚎𝚖. 𝙰𝚗𝚍 𝚒𝚝 𝚑𝚊𝚙𝚙𝚎𝚗𝚜 𝚒𝚗 𝚎𝚟𝚎𝚛𝚢 𝚒𝚗𝚍𝚞𝚜𝚝𝚛𝚢, 𝚎𝚟𝚎𝚛𝚢 𝚝𝚒𝚖𝚎 𝚊 𝚋𝚞𝚜𝚒𝚗𝚎𝚜𝚜 𝚕𝚎𝚊𝚍𝚎𝚛 𝚑𝚊𝚜 𝚊𝚗 𝚒𝚍𝚎𝚊 𝚝𝚛𝚊𝚙𝚙𝚎𝚍 𝚒𝚗𝚜𝚒𝚍𝚎 𝚊 𝙿𝚘𝚠𝚎𝚛𝙿𝚘𝚒𝚗𝚝 𝚜𝚕𝚒𝚍𝚎.

𝚃𝚑𝚎 𝚘𝚕𝚍 𝙶𝚘𝚘𝚐𝚕𝚎 𝚂𝚝𝚒𝚝𝚌𝚑 𝚑𝚎𝚕𝚙𝚎𝚍 𝚋𝚞𝚝 𝚒𝚝 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎𝚍 𝚘𝚗𝚎 𝚜𝚌𝚛𝚎𝚎𝚗 𝚊𝚝 𝚊 𝚝𝚒𝚖𝚎, 𝚑𝚊𝚍 𝚗𝚘 𝚍𝚎𝚜𝚒𝚐𝚗 𝚖𝚎𝚖𝚘𝚛𝚢 𝚊𝚌𝚛𝚘𝚜𝚜 𝚙𝚛𝚘𝚓𝚎𝚌𝚝𝚜, 𝚊𝚗𝚍 𝚔𝚎𝚙𝚝 𝚢𝚘𝚞 𝚒𝚗𝚜𝚒𝚍𝚎 𝚊 𝚝𝚎𝚡𝚝 𝚋𝚘𝚡. 𝚃𝚑𝚎 𝙼𝚊𝚛𝚌𝚑 𝟸𝟶𝟸𝟼 𝚅𝚒𝚋𝚎 𝙳𝚎𝚜𝚒𝚐𝚗 𝚞𝚙𝚍𝚊𝚝𝚎 𝚌𝚑𝚊𝚗𝚐𝚎𝚜 𝚝𝚑𝚎 𝚐𝚊𝚖𝚎 𝚎𝚗𝚝𝚒𝚛𝚎𝚕𝚢. 𝚈𝚘𝚞 𝚌𝚊𝚗 𝚗𝚘𝚠 𝚜𝚙𝚎𝚊𝚔 𝚢𝚘𝚞𝚛 𝚋𝚛𝚒𝚎𝚏 𝚘𝚞𝚝 𝚕𝚘𝚞𝚍, 𝚛𝚞𝚗 𝚝𝚑𝚛𝚎𝚎 𝚍𝚎𝚜𝚒𝚐𝚗 𝚍𝚒𝚛𝚎𝚌𝚝𝚒𝚘𝚗𝚜 𝚜𝚒𝚖𝚞𝚕𝚝𝚊𝚗𝚎𝚘𝚞𝚜𝚕𝚢, 𝚊𝚗𝚍 𝚠𝚊𝚕𝚔 𝚝𝚑𝚛𝚘𝚞𝚐𝚑 𝚊 𝚌𝚕𝚒𝚌𝚔𝚊𝚋𝚕𝚎 𝚙𝚛𝚘𝚝𝚘𝚝𝚢𝚙𝚎 — 𝚊𝚕𝚕 𝚒𝚗𝚜𝚒𝚍𝚎 𝚊 𝚜𝚒𝚗𝚐𝚕𝚎 𝙰𝙸-𝚗𝚊𝚝𝚒𝚟𝚎 𝚌𝚊𝚗𝚟𝚊𝚜. 𝙳𝚎𝚜𝚒𝚐𝚗 𝚒𝚜 𝚗𝚘 𝚕𝚘𝚗𝚐𝚎𝚛 𝚊 𝚜𝚙𝚎𝚌𝚒𝚊𝚕𝚒𝚜𝚝 𝚜𝚔𝚒𝚕𝚕. 𝙸𝚝’𝚜 𝚊 𝚝𝚑𝚒𝚗𝚔𝚒𝚗𝚐 𝚜𝚔𝚒𝚕𝚕.

𝙷𝚘𝚠 𝚝𝚘 𝚊𝚌𝚌𝚎𝚜𝚜: 𝚑𝚝𝚝𝚙𝚜://𝚜𝚝𝚒𝚝𝚌𝚑.𝚠𝚒𝚝𝚑𝚐𝚘𝚘𝚐𝚕𝚎.𝚌𝚘𝚖 (𝚏𝚛𝚎𝚎 𝚟𝚒𝚊 𝙶𝚘𝚘𝚐𝚕𝚎 𝙻𝚊𝚋𝚜)

𝙶𝚘𝚘𝚐𝚕𝚎 𝚂𝚝𝚒𝚝𝚌𝚑 𝚌𝚊𝚗 𝚑𝚎𝚕𝚙 𝚢𝚘𝚞:

• 𝚅𝚒𝚋𝚎 𝚍𝚎𝚜𝚒𝚐𝚗 𝚏𝚛𝚘𝚖 𝚊 𝚍𝚎𝚜𝚌𝚛𝚒𝚙𝚝𝚒𝚘𝚗: 𝙿𝚕𝚊𝚒𝚗-𝙴𝚗𝚐𝚕𝚒𝚜𝚑 𝚋𝚛𝚒𝚎𝚏 → 𝚙𝚘𝚕𝚒𝚜𝚑𝚎𝚍 𝚖𝚞𝚕𝚝𝚒-𝚜𝚌𝚛𝚎𝚎𝚗 𝚄𝙸 𝚒𝚗 𝚜𝚎𝚌𝚘𝚗𝚍𝚜

• 𝙿𝚛𝚘𝚝𝚘𝚝𝚢𝚙𝚎 𝚒𝚗𝚜𝚝𝚊𝚗𝚝𝚕𝚢: 𝙲𝚘𝚗𝚗𝚎𝚌𝚝 𝚜𝚌𝚛𝚎𝚎𝚗𝚜 𝚊𝚗𝚍 𝚑𝚒𝚝 “𝙿𝚕𝚊𝚢” 𝚝𝚘 𝚠𝚊𝚕𝚔 𝚝𝚑𝚛𝚘𝚞𝚐𝚑 𝚝𝚑𝚎 𝚏𝚞𝚕𝚕 𝚞𝚜𝚎𝚛 𝚓𝚘𝚞𝚛𝚗𝚎𝚢

• 𝙴𝚡𝚙𝚘𝚛𝚝 𝚊𝚗𝚍 𝚋𝚞𝚒𝚕𝚍: 𝙲𝚕𝚎𝚊𝚗 𝙷𝚃𝙼𝙻, 𝙲𝚂𝚂, 𝚘𝚛 𝚁𝚎𝚊𝚌𝚝 𝚌𝚘𝚍𝚎-𝚛𝚎𝚊𝚍𝚢 𝚏𝚘𝚛 𝚢𝚘𝚞𝚛 𝚍𝚎𝚟𝚎𝚕𝚘𝚙𝚎𝚛

𝙴𝚡𝚊𝚖𝚙𝚕𝚎:

𝙰 p𝚛𝚘𝚍𝚞𝚌𝚝 m𝚊𝚗𝚊𝚐𝚎𝚛 𝚗𝚎𝚎𝚍𝚜 𝚝𝚑𝚛𝚎𝚎 𝚕𝚊𝚢𝚘𝚞𝚝 𝚘𝚙𝚝𝚒𝚘𝚗𝚜 𝚏𝚘𝚛 𝚊 𝚗𝚎𝚠 𝚏𝚎𝚊𝚝𝚞𝚛𝚎 𝚋𝚎𝚏𝚘𝚛𝚎 𝚝𝚘𝚖𝚘𝚛𝚛𝚘𝚠’𝚜 𝚜𝚝𝚊𝚗𝚍𝚞𝚙. 𝙷𝚎𝚛𝚎’𝚜 𝚝𝚑𝚎 𝚞𝚙𝚍𝚊𝚝𝚎𝚍 𝚂𝚝𝚒𝚝𝚌𝚑 𝚠𝚘𝚛𝚔𝚏𝚕𝚘𝚠:

• 𝙴𝚡𝚝𝚛𝚊𝚌𝚝 𝚢𝚘𝚞𝚛 𝚋𝚛𝚊𝚗𝚍: 𝙿𝚊𝚜𝚝𝚎 𝚢𝚘𝚞𝚛 𝚌𝚘𝚖𝚙𝚊𝚗𝚢 𝚄𝚁𝙻 - 𝚂𝚝𝚒𝚝𝚌𝚑 𝚙𝚞𝚕𝚕𝚜 𝚢𝚘𝚞𝚛 𝚌𝚘𝚕𝚘𝚞𝚛𝚜, 𝚏𝚘𝚗𝚝𝚜, 𝚊𝚗𝚍 𝚌𝚘𝚖𝚙𝚘𝚗𝚎𝚗𝚝𝚜 𝚒𝚗𝚝𝚘 𝚊 𝙳𝙴𝚂𝙸𝙶𝙽.𝚖𝚍 𝚏𝚒𝚕𝚎. 𝙴𝚟𝚎𝚛𝚢 𝚜𝚌𝚛𝚎𝚎𝚗 𝚒𝚝 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎𝚜 𝚏𝚛𝚘𝚖 𝚑𝚎𝚛𝚎 𝚏𝚘𝚕𝚕𝚘𝚠𝚜 𝚢𝚘𝚞𝚛 𝚋𝚛𝚊𝚗𝚍 𝚊𝚞𝚝𝚘𝚖𝚊𝚝𝚒𝚌𝚊𝚕𝚕𝚢.

• 𝚁𝚞𝚗 𝚙𝚊𝚛𝚊𝚕𝚕𝚎𝚕 𝚍𝚒𝚛𝚎𝚌𝚝𝚒𝚘𝚗𝚜: 𝚄𝚜𝚎 𝚝𝚑𝚎 𝚗𝚎𝚠 𝙰𝚐𝚎𝚗𝚝 𝙼𝚊𝚗𝚊𝚐𝚎𝚛 𝚝𝚘 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎 𝟹 𝚍𝚎𝚜𝚒𝚐𝚗 𝚟𝚊𝚛𝚒𝚊𝚗𝚝𝚜 𝚜𝚒𝚖𝚞𝚕𝚝𝚊𝚗𝚎𝚘𝚞𝚜𝚕𝚢 — 𝚍𝚒𝚟𝚎𝚛𝚐𝚎 𝚒𝚗 𝚖𝚒𝚗𝚞𝚝𝚎𝚜, 𝚗𝚘𝚝 𝚍𝚊𝚢𝚜.

• 𝙲𝚕𝚒𝚌𝚔 𝙿𝚕𝚊𝚢: 𝚃𝚑𝚎 𝚗𝚎𝚠 𝚒𝚗𝚝𝚎𝚛𝚊𝚌𝚝𝚒𝚟𝚎 𝚙𝚛𝚘𝚝𝚘𝚝𝚢𝚙𝚎 𝚖𝚘𝚍𝚎 𝚌𝚘𝚗𝚗𝚎𝚌𝚝𝚜 𝚢𝚘𝚞𝚛 𝚜𝚌𝚛𝚎𝚎𝚗𝚜 𝚒𝚗𝚜𝚝𝚊𝚗𝚝𝚕𝚢 𝚠𝚊𝚕𝚔 𝚜𝚝𝚊𝚔𝚎𝚑𝚘𝚕𝚍𝚎𝚛𝚜 𝚝𝚑𝚛𝚘𝚞𝚐𝚑 𝚝𝚑𝚎 𝚏𝚞𝚕𝚕 𝚞𝚜𝚎𝚛 𝚓𝚘𝚞𝚛𝚗𝚎𝚢 𝚋𝚎𝚏𝚘𝚛𝚎 𝚊 𝚜𝚒𝚗𝚐𝚕𝚎 𝚕𝚒𝚗𝚎 𝚘𝚏 𝚌𝚘𝚍𝚎 𝚒𝚜 𝚠𝚛𝚒𝚝𝚝𝚎𝚗.

• 𝚂𝚙𝚎𝚊𝚔 𝚢𝚘𝚞𝚛 𝚎𝚍𝚒𝚝𝚜: 𝚂𝚠𝚒𝚝𝚌𝚑 𝚝𝚘 𝚅𝚘𝚒𝚌𝚎 𝙲𝚊𝚗𝚟𝚊𝚜 𝚊𝚗𝚍 𝚜𝚊𝚢 “𝚖𝚊𝚔𝚎 𝚝𝚑𝚎 𝚍𝚊𝚜𝚑𝚋𝚘𝚊𝚛𝚍 𝚍𝚊𝚛𝚔𝚎𝚛 𝚊𝚗𝚍 𝚜𝚑𝚘𝚠 𝚖𝚎 𝚝𝚠𝚘 𝚖𝚎𝚗𝚞 𝚘𝚙𝚝𝚒𝚘𝚗𝚜” — 𝚝𝚑𝚎 𝚊𝚐𝚎𝚗𝚝 𝚞𝚙𝚍𝚊𝚝𝚎𝚜 𝚒𝚗 𝚛𝚎𝚊𝚕 𝚝𝚒𝚖𝚎.

𝚆𝚑𝚊𝚝 𝚞𝚜𝚎𝚍 𝚝𝚘 𝚝𝚊𝚔𝚎 𝚊 𝚍𝚎𝚜𝚒𝚐𝚗𝚎𝚛 𝚝𝚠𝚘 𝚍𝚊𝚢𝚜 𝚗𝚘𝚠 𝚝𝚊𝚔𝚎𝚜 𝚝𝚠𝚎𝚗𝚝𝚢 𝚖𝚒𝚗𝚞𝚝𝚎𝚜.

𝚆𝚑𝚊𝚝 𝚖𝚊𝚔𝚎𝚜 𝙶𝚘𝚘𝚐𝚕𝚎 𝚂𝚝𝚒𝚝𝚌𝚑’𝚜 𝚅𝚒𝚋𝚎 𝙳𝚎𝚜𝚒𝚐𝚗 𝚞𝚙𝚍𝚊𝚝𝚎 𝚜𝚝𝚊𝚗𝚍 𝚘𝚞𝚝?

• 𝙳𝙴𝚂𝙸𝙶𝙽.𝚖𝚍: 𝚈𝚘𝚞𝚛 𝚋𝚛𝚊𝚗𝚍 𝚜𝚢𝚜𝚝𝚎𝚖, 𝚙𝚘𝚛𝚝𝚊𝚋𝚕𝚎 𝚊𝚌𝚛𝚘𝚜𝚜 𝚎𝚟𝚎𝚛𝚢 𝚙𝚛𝚘𝚓𝚎𝚌𝚝 — 𝚗𝚘 𝚖𝚘𝚛𝚎 𝚟𝚒𝚜𝚞𝚊𝚕 𝚍𝚛𝚒𝚏𝚝

• 𝙰𝚐𝚎𝚗𝚝 𝙼𝚊𝚗𝚊𝚐𝚎𝚛: 𝙴𝚡𝚙𝚕𝚘𝚛𝚎 𝚖𝚞𝚕𝚝𝚒𝚙𝚕𝚎 𝚍𝚎𝚜𝚒𝚐𝚗 𝚍𝚒𝚛𝚎𝚌𝚝𝚒𝚘𝚗𝚜 𝚒𝚗 𝚙𝚊𝚛𝚊𝚕𝚕𝚎𝚕, 𝚗𝚘𝚝 𝚜𝚎𝚚𝚞𝚎𝚗𝚝𝚒𝚊𝚕𝚕𝚢

• 𝙲𝚘𝚖𝚙𝚕𝚎𝚝𝚎𝚕𝚢 𝚏𝚛𝚎𝚎: 𝙽𝚘 𝚙𝚊𝚒𝚍 𝚙𝚕𝚊𝚗 𝚢𝚎𝚝 — 𝚊𝚟𝚊𝚒𝚕𝚊𝚋𝚕𝚎 𝚗𝚘𝚠 𝚟𝚒𝚊 𝙶𝚘𝚘𝚐𝚕𝚎 𝙻𝚊𝚋𝚜

𝙽𝚘𝚝𝚎: 𝚃𝚑𝚎 𝚝𝚘𝚘𝚕𝚜 𝚊𝚗𝚍 𝚊𝚗𝚊𝚕𝚢𝚜𝚒𝚜 𝚏𝚎𝚊𝚝𝚞𝚛𝚎𝚍 𝚒𝚗 𝚝𝚑𝚒𝚜 𝚜𝚎𝚌𝚝𝚒𝚘𝚗 𝚍𝚎𝚖𝚘𝚗𝚜𝚝𝚛𝚊𝚝𝚎𝚍 𝚌𝚕𝚎𝚊𝚛 𝚟𝚊𝚕𝚞𝚎 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚘𝚞𝚛 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚝𝚎𝚜𝚝𝚒𝚗𝚐. 𝙾𝚞𝚛 𝚛𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗𝚜 𝚊𝚛𝚎 𝚎𝚗𝚝𝚒𝚛𝚎𝚕𝚢 𝚒𝚗𝚍𝚎𝚙𝚎𝚗𝚍𝚎𝚗𝚝 𝚊𝚗𝚍 𝚗𝚘𝚝 𝚒𝚗𝚏𝚕𝚞𝚎𝚗𝚌𝚎𝚍 𝚋𝚢 𝚝𝚑𝚎 𝚝𝚘𝚘𝚕 𝚌𝚛𝚎𝚊𝚝𝚘𝚛𝚜.

AI BITS & BYTES

OpenAI’s unified desktop ‘super app’

OpenAI is combining its AI-powered web browser, native ChatGPT and Codex apps into a desktop super app, according to media reports. The company’s chief executive of applications, Fidji Simo, confirmed the reports in a post: “…when new bets start to work, like we’re seeing now with Codex, it’s very important to double down on them and avoid distractions…”. Read more.

Elon Musk launches Terafab, unveils AI chip factory

Elon Musk has embarked on the “most epic chip-building exercise in history” with the launch of “Terafab“. The new initiative will be a joint venture of Tesla and SpaceX and has begun with an advanced manufacturing facility in Texas. Musk claims the new facility is the only one in the world to have the entire chip-making lifecycle under one roof, including memory, packaging, testing, and manufacturing of lithography masks. Read more.

Why OpenAI has bid goodbye to Sora

OpenAI’s Sora was made available to the public in late 2024, but it wasn’t until the launch of Sora 2 and its standalone app in September 2025 that the video generation platform became a viral sensation. On Wednesday, the maker of ChatGPT announced it would shut down Sora.

More AI Reading

• How Nvidia keeps its iron grip on the AI boom

• Nvidia CEO blames recent tech layoffs on lack of ‘imagination’, not AI

• Here's where AI will take away jobs—and where it will create them

• America’s chief financial officers say AI is coming for admin jobs

• Can IBM reboot its AI play with $11 billion Confluent buyout?

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Madhumita Sen Choudhury. Produced by Tushar Deep Singh.

Taking the stage in his trademark black leather jacket, co-founder and CEO Jensen Huang laid out an audacious vision: Nvidia could generate as much as $1 trillion in annual revenue by 2027 from AI chip sales alone, doubling his earlier $500 billion forecast for 2026.

Huang’s confidence rests on a staggering premise that demand for compute has surged a million-fold over the past two years. That explosion, he argues, will drive cumulative purchase orders for Nvidia’s current Blackwell chips and its next-gen Vera Rubin architecture to at least $1 trillion.

If realised, the milestone would be historic: no company has ever generated $1 trillion in annual revenue. For context, Nvidia reported $215.9 billion in revenue in 2025, up from $130.5 billion a year-ago. Huang is signalling that this is just the beginning of a far steeper curve.

In this week’s edition of Tech Talk:

• Can AI cure cancer in humans?

• Can AI predict the next market crash?

• AI Tool of the Week: Google Pomelli

PREVIOUSLY, ON MINT TECH TALK

Addressing the inference challenge

For more than a decade, Nvidia has been the undisputed kingmaker of the AI boom. Its graphics processing units are the backbone of modern AI, but as the industry pivots from training models to running them at scale—also known as inference—the competitive landscape could widen, creating openings for a new set of challengers.

In the training stage, giant models ingest vast datasets and learn complex patterns. During inference, the trained models actually answer queries, recommend products, generate images, or power AI agents. Every user interaction requires inference, and as AI applications proliferate across industries, the volume of these tasks will grow exponentially.

Inference workloads also prioritise different characteristics than training. Instead of raw computational muscle, they emphasise latency, power efficiency, and cost per query. GPUs excel at highly parallel training tasks, but inference opens the door to specialised chips designed for narrower, more efficient workloads.

Suddenly, you don’t need Nvidia GPUs anymore.

• AMD has been steadily building its AI portfolio with accelerators such as the MI300 series, aimed directly at Nvidia’s data centre market.

• Intel Corp. is pushing its Gaudi accelerator line as a lower-cost alternative for AI workloads.

• Qualcomm Inc. is attempting to leverage its expertise in power-efficient chip design to produce inference-optimised data-centre processors.

Simultaneously, Nvidia’s own customers—from Alphabet Inc. and Microsoft Corp. to Meta Platforms Inc., are investing heavily in custom AI silicon. Alphabet has long relied on its proprietary Tensor chips, Microsoft is building its Maia accelerator, and Meta is designing in-house inference hardware for its giga-scale data centres.

That, even as new-age startups such as Cerebras Systems, are betting on specialised architecture to outperform general-purpose GPUs in inference tasks.

“Cerebras chose to use SRAM (GPUs use DRAM or HBM) so that we could move data from memory to compute faster (also known as inference). Not a little bit faster but more than 2,600 times faster than NVIDIA Blackwell GPUs. As a result, we can generate tokens faster 15 times faster.” — Andrew Feldman, CEO and founder of Cerebras, on X.

While their market share remains small, Cerebras’ technology highlights how the AI hardware landscape could fragment as workloads diversify even as Nvidia’s software ecosystem, particularly CUDA, remains deeply embedded across AI development.

Besides, Nvidia has begun its own inference pivot with Huang unveiling the Nvidia Groq 3 Language Processing Unit, or LPU. Likely to ship in the third quarter, this LPU has been integrated into the Vera Rubin platform, working alongside GPUs and aimed at speeding up the inference process. This is a space that we’ll keenly be watching.

AI TOOL OF THE WEEK

By AI&Beyond, with Jaspreet Bindra and Anuj Magazine

𝚃𝚑𝚎 𝙰𝙸 𝚝𝚘𝚘𝚕 𝚠𝚎 𝚞𝚗𝚕𝚘𝚌𝚔𝚎𝚍 𝚝𝚘𝚍𝚊𝚢 𝚒𝚜: 𝙶𝚘𝚘𝚐𝚕𝚎 𝙿𝚘𝚖𝚎𝚕𝚕𝚒

𝚆𝚑𝚊𝚝 𝚙𝚛𝚘𝚋𝚕𝚎𝚖 𝚍𝚘𝚎𝚜 𝚒𝚝 𝚜𝚘𝚕𝚟𝚎? 𝚈𝚘𝚞 𝚛𝚞𝚗 𝚊 𝚐𝚛𝚘𝚠𝚒𝚗𝚐 𝙳𝟸𝙲 𝚋𝚛𝚊𝚗𝚍 — 𝚜𝚔𝚒𝚗𝚌𝚊𝚛𝚎, 𝚏𝚘𝚘𝚍, 𝚏𝚊𝚜𝚑𝚒𝚘𝚗. 𝙿𝚛𝚘𝚍𝚞𝚌𝚝 𝚕𝚊𝚞𝚗𝚌𝚑 𝚒𝚗 𝚝𝚠𝚘 𝚠𝚎𝚎𝚔𝚜. 𝚈𝚘𝚞 𝚗𝚎𝚎𝚍 𝙸𝚗𝚜𝚝𝚊𝚐𝚛𝚊𝚖 𝚙𝚘𝚜𝚝𝚜, 𝙵𝚊𝚌𝚎𝚋𝚘𝚘𝚔 𝚊𝚍𝚜, 𝚙𝚛𝚘𝚍𝚞𝚌𝚝 𝚙𝚑𝚘𝚝𝚘𝚜, 𝚊 𝚁𝚎𝚎𝚕. 𝙰 𝚙𝚛𝚘𝚏𝚎𝚜𝚜𝚒𝚘𝚗𝚊𝚕 𝚜𝚑𝚘𝚘𝚝 𝚌𝚘𝚜𝚝𝚜 ₹𝟻𝟶,𝟶𝟶𝟶 𝚊𝚗𝚍 𝟷𝟶 𝚍𝚊𝚢𝚜 𝚢𝚘𝚞 𝚍𝚘𝚗’𝚝 𝚑𝚊𝚟𝚎. 𝚂𝚘 𝚢𝚘𝚞 𝚍𝚎𝚕𝚊𝚢. 𝙾𝚛 𝚙𝚘𝚜𝚝 𝚜𝚘𝚖𝚎𝚝𝚑𝚒𝚗𝚐 𝚘𝚏𝚏-𝚋𝚛𝚊𝚗𝚍. 𝙾𝚛 𝚜𝚙𝚎𝚗𝚍 𝚝𝚑𝚛𝚎𝚎 𝚎𝚟𝚎𝚗𝚒𝚗𝚐𝚜 𝚒𝚗 𝙲𝚊𝚗𝚟𝚊, 𝚊𝚗𝚍 𝚒𝚝 𝚜𝚝𝚒𝚕𝚕 𝚍𝚘𝚎𝚜𝚗’𝚝 𝚕𝚘𝚘𝚔 𝚛𝚒𝚐𝚑𝚝．

𝚃𝚘 𝚋𝚎 𝚜𝚞𝚛𝚎, 𝚒𝚗𝚌𝚘𝚗𝚜𝚒𝚜𝚝𝚎𝚗𝚝, 𝚕𝚘𝚠-𝚚𝚞𝚊𝚕𝚒𝚝𝚢 𝚖𝚊𝚛𝚔𝚎𝚝𝚒𝚗𝚐 𝚌𝚘𝚗𝚝𝚎𝚗𝚝 𝚌𝚘𝚜𝚝𝚜 𝚢𝚘𝚞 𝚜𝚊𝚕𝚎𝚜, 𝚝𝚛𝚞𝚜𝚝, 𝚊𝚗𝚍 𝚟𝚒𝚜𝚒𝚋𝚒𝚕𝚒𝚝𝚢. 𝙵𝚘𝚛 𝚜𝚖𝚊𝚕𝚕 𝚋𝚞𝚜𝚒𝚗𝚎𝚜𝚜𝚎𝚜, 𝚝𝚑𝚎 𝚐𝚊𝚙 𝚋𝚎𝚝𝚠𝚎𝚎𝚗 𝚠𝚑𝚊𝚝 𝚢𝚘𝚞 𝚠𝚊𝚗𝚝 𝚝𝚘 𝚜𝚊𝚢 𝚊𝚗𝚍 𝚠𝚑𝚊𝚝 𝚢𝚘𝚞 𝚌𝚊𝚗 𝚊𝚏𝚏𝚘𝚛𝚍 𝚝𝚘 𝚙𝚛𝚘𝚍𝚞𝚌𝚎 𝚒𝚜 𝚘𝚗𝚎 𝚘𝚏 𝚝𝚑𝚎 𝚖𝚘𝚜𝚝 𝚙𝚎𝚛𝚜𝚒𝚜𝚝𝚎𝚗𝚝, 𝚎𝚡𝚙𝚎𝚗𝚜𝚒𝚟𝚎 𝚙𝚛𝚘𝚋𝚕𝚎𝚖𝚜 𝚒𝚗 𝚝𝚑𝚎 𝚋𝚞𝚒𝚕𝚍𝚒𝚗𝚐．

𝙶𝚘𝚘𝚐𝚕𝚎 𝙿𝚘𝚖𝚎𝚕𝚕𝚒 𝚌𝚕𝚘𝚜𝚎𝚜 𝚝𝚑𝚊𝚝 𝚐𝚊𝚙. 𝙴𝚗𝚝𝚎𝚛 𝚢𝚘𝚞𝚛 𝚠𝚎𝚋𝚜𝚒𝚝𝚎 𝚄𝚁𝙻 — 𝚒𝚝 𝚛𝚎𝚊𝚍𝚜 𝚢𝚘𝚞𝚛 𝚋𝚛𝚊𝚗𝚍, 𝚎𝚡𝚝𝚛𝚊𝚌𝚝𝚜 𝚢𝚘𝚞𝚛 𝚌𝚘𝚕𝚘𝚞𝚛𝚜, 𝚏𝚘𝚗𝚝𝚜, 𝚊𝚗𝚍 𝚝𝚘𝚗𝚎, 𝚊𝚗𝚍 𝚋𝚞𝚒𝚕𝚍𝚜 𝚢𝚘𝚞𝚛 “𝙱𝚞𝚜𝚒𝚗𝚎𝚜𝚜 𝙳𝙽𝙰”. 𝙴𝚟𝚎𝚛𝚢𝚝𝚑𝚒𝚗𝚐 𝚒𝚝 𝚌𝚛𝚎𝚊𝚝𝚎𝚜 𝚏𝚛𝚘𝚖 𝚝𝚑𝚊𝚝 𝚙𝚘𝚒𝚗𝚝 𝚒𝚜 𝚊𝚗𝚌𝚑𝚘𝚛𝚎𝚍 𝚝𝚘 𝚠𝚑𝚘 𝚢𝚘𝚞 𝚊𝚌𝚝𝚞𝚊𝚕𝚕𝚢 𝚊𝚛𝚎．

𝙷𝚘𝚠 𝚝𝚘 𝚊𝚌𝚌𝚎𝚜𝚜: 𝚑𝚝𝚝𝚙𝚜://𝚕𝚊𝚋𝚜.𝚐𝚘𝚘𝚐𝚕𝚎.𝚌𝚘𝚖/𝚙𝚘𝚖𝚎𝚕𝚕𝚒

𝙶𝚘𝚘𝚐𝚕𝚎 𝙿𝚘𝚖𝚎𝚕𝚕𝚒 𝚌𝚊𝚗 𝚑𝚎𝚕𝚙 𝚢𝚘𝚞:

• 𝙶𝚎𝚗𝚎𝚛𝚊𝚝𝚎 𝚘𝚗-𝚋𝚛𝚊𝚗𝚍 𝚌𝚊𝚖𝚙𝚊𝚒𝚐𝚗𝚜: 𝙳𝚎𝚜𝚌𝚛𝚒𝚋𝚎 𝚢𝚘𝚞𝚛 𝚙𝚛𝚘𝚖𝚘𝚝𝚒𝚘𝚗 𝚒𝚗 𝚙𝚕𝚊𝚒𝚗 𝙴𝚗𝚐𝚕𝚒𝚜𝚑 𝚊𝚗𝚍 𝚐𝚎𝚝 𝚙𝚘𝚜𝚝𝚜, 𝚊𝚍𝚜, 𝚎𝚖𝚊𝚒𝚕 𝚋𝚊𝚗𝚗𝚎𝚛𝚜, 𝚊𝚗𝚍 𝚈𝚘𝚞𝚃𝚞𝚋𝚎 𝚝𝚑𝚞𝚖𝚋𝚗𝚊𝚒𝚕𝚜 𝚊𝚕𝚕 𝚒𝚗 𝚢𝚘𝚞𝚛 𝚋𝚛𝚊𝚗𝚍’𝚜 𝚟𝚘𝚒𝚌𝚎 𝚊𝚗𝚍 𝚙𝚊𝚕𝚎𝚝𝚝𝚎.

• 𝙲𝚛𝚎𝚊𝚝𝚎 𝚜𝚝𝚞𝚍𝚒𝚘-𝚚𝚞𝚊𝚕𝚒𝚝𝚢 𝚙𝚛𝚘𝚍𝚞𝚌𝚝 𝚙𝚑𝚘𝚝𝚘𝚜: 𝚄𝚙𝚕𝚘𝚊𝚍 𝚊 𝚜𝚖𝚊𝚛𝚝𝚙𝚑𝚘𝚗𝚎 𝚙𝚑𝚘𝚝𝚘 𝚘𝚏 𝚢𝚘𝚞𝚛 𝚙𝚛𝚘𝚍𝚞𝚌𝚝 𝚊𝚗𝚍 𝙿𝚘𝚖𝚎𝚕𝚕𝚒 𝚝𝚛𝚊𝚗𝚜𝚏𝚘𝚛𝚖𝚜 𝚒𝚝 𝚒𝚗𝚝𝚘 𝚙𝚛𝚘𝚏𝚎𝚜𝚜𝚒𝚘𝚗𝚊𝚕 𝚜𝚝𝚞𝚍𝚒𝚘, 𝚕𝚒𝚏𝚎𝚜𝚝𝚢𝚕𝚎, 𝚊𝚗𝚍 𝚒𝚗-𝚞𝚜𝚎 𝚜𝚑𝚘𝚝𝚜.

• 𝙿𝚛𝚘𝚍𝚞𝚌𝚎 𝚜𝚑𝚘𝚛𝚝-𝚏𝚘𝚛𝚖 𝚟𝚒𝚍𝚎𝚘: 𝚃𝚞𝚛𝚗 𝚊𝚗𝚢 𝚜𝚝𝚊𝚝𝚒𝚌 𝚊𝚜𝚜𝚎𝚝 𝚒𝚗𝚝𝚘 𝚊 𝚋𝚛𝚊𝚗𝚍𝚎𝚍 𝚁𝚎𝚎𝚕 𝚘𝚛 𝚃𝚒𝚔𝚃𝚘𝚔-𝚛𝚎𝚊𝚍𝚢 𝚊𝚗𝚒𝚖𝚊𝚝𝚒𝚘𝚗 𝚠𝚒𝚝𝚑 𝚘𝚗𝚎 𝚌𝚕𝚒𝚌𝚔．

𝙴𝚡𝚊𝚖𝚙𝚕𝚎:

𝙰 𝚏𝚘𝚘𝚍 𝚋𝚛𝚊𝚗𝚍 𝚒𝚜 𝚕𝚊𝚞𝚗𝚌𝚑𝚒𝚗𝚐 𝚊 𝚗𝚎𝚠 𝚟𝚊𝚛𝚒𝚊𝚗𝚝 𝚊𝚑𝚎𝚊𝚍 𝚘𝚏 𝚝𝚑𝚎 𝚏𝚎𝚜𝚝𝚒𝚟𝚎 𝚜𝚎𝚊𝚜𝚘𝚗. 𝙷𝚎𝚛𝚎’𝚜 𝚠𝚑𝚊𝚝 𝚝𝚑𝚎𝚢 𝚍𝚘 𝚠𝚒𝚝𝚑 𝙿𝚘𝚖𝚎𝚕𝚕𝚒:

• 𝙴𝚗𝚝𝚎𝚛 𝚝𝚑𝚎 𝚠𝚎𝚋𝚜𝚒𝚝𝚎: 𝙿𝚘𝚖𝚎𝚕𝚕𝚒 𝚜𝚌𝚊𝚗𝚜 𝚝𝚑𝚎 𝚜𝚒𝚝𝚎 𝚊𝚗𝚍 𝚋𝚞𝚒𝚕𝚍𝚜 𝚝𝚑𝚎 𝚋𝚛𝚊𝚗𝚍 𝚙𝚛𝚘𝚏𝚒𝚕𝚎．

• 𝙳𝚎𝚜𝚌𝚛𝚒𝚋𝚎 𝚝𝚑𝚎 𝚌𝚊𝚖𝚙𝚊𝚒𝚐𝚗: “𝙻𝚊𝚞𝚗𝚌𝚑 𝚘𝚞𝚛 𝚗𝚎𝚠 𝚖𝚊𝚗𝚐𝚘 𝚟𝚊𝚛𝚒𝚊𝚗𝚝 𝚝𝚊𝚛𝚐𝚎𝚝 𝚏𝚊𝚖𝚒𝚕𝚒𝚎𝚜 𝚌𝚎𝚕𝚎𝚋𝚛𝚊𝚝𝚒𝚗𝚐 𝚝𝚘𝚐𝚎𝚝𝚑𝚎𝚛 𝚝𝚑𝚒𝚜 𝚏𝚎𝚜𝚝𝚒𝚟𝚎 𝚜𝚎𝚊𝚜𝚘𝚗.”

• 𝙶𝚎𝚝 𝚊 𝚏𝚞𝚕𝚕 𝚌𝚊𝚖𝚙𝚊𝚒𝚐𝚗: 𝚃𝚑𝚛𝚎𝚎 𝚌𝚘𝚗𝚌𝚎𝚙𝚝 𝚍𝚒𝚛𝚎𝚌𝚝𝚒𝚘𝚗𝚜 𝚊𝚙𝚙𝚎𝚊𝚛. 𝚃𝚑𝚎𝚢 𝚙𝚒𝚌𝚔 𝚘𝚗𝚎. 𝙸𝚗 𝟼𝟶 𝚜𝚎𝚌𝚘𝚗𝚍𝚜, 𝚝𝚑𝚎𝚢 𝚑𝚊𝚟𝚎 𝙸𝚗𝚜𝚝𝚊𝚐𝚛𝚊𝚖 𝚙𝚘𝚜𝚝𝚜, 𝚊 𝙵𝚊𝚌𝚎𝚋𝚘𝚘𝚔 𝚊𝚍, 𝚊𝚗𝚍 𝚊𝚗 𝚎𝚖𝚊𝚒𝚕 𝚋𝚊𝚗𝚗𝚎𝚛 — 𝚊𝚕𝚕 𝚒𝚗 𝚝𝚑𝚎𝚒𝚛 𝚎𝚡𝚊𝚌𝚝 𝚋𝚛𝚊𝚗𝚍 𝚌𝚘𝚕𝚘𝚞𝚛𝚜 𝚊𝚗𝚍 𝚟𝚘𝚒𝚌𝚎.

• 𝙲𝚛𝚎𝚊𝚝𝚎 𝚝𝚑𝚎 𝚙𝚛𝚘𝚍𝚞𝚌𝚝 𝚙𝚑𝚘𝚝𝚘: 𝚃𝚑𝚎𝚢 𝚞𝚙𝚕𝚘𝚊𝚍 𝚊 𝚙𝚑𝚘𝚗𝚎 𝚜𝚑𝚘𝚝 𝚘𝚏 𝚝𝚑𝚎 𝚖𝚊𝚗𝚐𝚘 𝚓𝚊𝚛 𝚘𝚗 𝚊 𝚔𝚒𝚝𝚌𝚑𝚎𝚗 𝚌𝚘𝚞𝚗𝚝𝚎𝚛. 𝙿𝚘𝚖𝚎𝚕𝚕𝚒 𝚛𝚎𝚝𝚞𝚛𝚗𝚜 𝚏𝚘𝚞𝚛 𝚒𝚖𝚊𝚐𝚎𝚜 — 𝚊 𝚌𝚕𝚎𝚊𝚗 𝚜𝚝𝚞𝚍𝚒𝚘 𝚜𝚑𝚘𝚝 𝚏𝚘𝚛 𝚎-𝚌𝚘𝚖𝚖𝚎𝚛𝚌𝚎, 𝚊 𝚕𝚒𝚏𝚎𝚜𝚝𝚢𝚕𝚎 𝚜𝚑𝚘𝚝 𝚠𝚒𝚝𝚑 𝚊 𝚏𝚎𝚜𝚝𝚒𝚟𝚎 𝚝𝚊𝚋𝚕𝚎 𝚜𝚎𝚝𝚝𝚒𝚗𝚐, 𝚊𝚗𝚍 𝚊𝚗 𝚒𝚗-𝚞𝚜𝚎 𝚜𝚑𝚘𝚝. 𝚉𝚎𝚛𝚘 𝚙𝚑𝚘𝚝𝚘𝚐𝚛𝚊𝚙𝚑𝚢 𝚋𝚞𝚍𝚐𝚎𝚝.

𝚆𝚑𝚊𝚝 𝚖𝚊𝚔𝚎𝚜 𝙶𝚘𝚘𝚐𝚕𝚎 𝙿𝚘𝚖𝚎𝚕𝚕𝚒 𝚜𝚙𝚎𝚌𝚒𝚊𝚕?

• 𝚈𝚘𝚞𝚛 𝚋𝚛𝚊𝚗𝚍, 𝚗𝚘𝚝 𝚊 𝚝𝚎𝚖𝚙𝚕𝚊𝚝𝚎: 𝚄𝚗𝚕𝚒𝚔𝚎 𝙲𝚊𝚗𝚟𝚊 𝚘𝚛 𝙰𝚍𝚘𝚋𝚎 𝙴𝚡𝚙𝚛𝚎𝚜𝚜, 𝚠𝚑𝚎𝚛𝚎 𝚢𝚘𝚞 𝚜𝚝𝚊𝚛𝚝 𝚠𝚒𝚝𝚑 𝚜𝚘𝚖𝚎𝚘𝚗𝚎 𝚎𝚕𝚜𝚎’𝚜 𝚍𝚎𝚜𝚒𝚐𝚗, 𝙿𝚘𝚖𝚎𝚕𝚕𝚒 𝚜𝚝𝚊𝚛𝚝𝚜 𝚠𝚒𝚝𝚑 𝚢𝚘𝚞𝚛𝚜．

• 𝚃𝚎𝚡𝚝 + 𝚒𝚖𝚊𝚐𝚎 + 𝚟𝚒𝚍𝚎𝚘 𝚒𝚗 𝚘𝚗𝚎 𝚙𝚕𝚊𝚌𝚎: 𝙽𝚘 𝚘𝚝𝚑𝚎𝚛 𝚏𝚛𝚎𝚎 𝚝𝚘𝚘𝚕 𝚝𝚘𝚍𝚊𝚢 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎𝚜 𝚌𝚘𝚙𝚢, 𝚟𝚒𝚜𝚞𝚊𝚕𝚜, 𝚙𝚛𝚘𝚍𝚞𝚌𝚝 𝚙𝚑𝚘𝚝𝚘𝚐𝚛𝚊𝚙𝚑𝚢, 𝚊𝚗𝚍 𝚊𝚗𝚒𝚖𝚊𝚝𝚎𝚍 𝚟𝚒𝚍𝚎𝚘 𝚏𝚛𝚘𝚖 𝚊 𝚜𝚒𝚗𝚐𝚕𝚎 𝚋𝚛𝚊𝚗𝚍 𝚙𝚛𝚘𝚏𝚒𝚕𝚎, 𝚊𝚕𝚕 𝚒𝚗 𝚘𝚗𝚎 𝚠𝚘𝚛𝚔𝚏𝚕𝚘𝚠．

• 𝙵𝚛𝚎𝚎 𝚊𝚗𝚍 𝚗𝚘𝚠 𝚊𝚟𝚊𝚒𝚕𝚊𝚋𝚕𝚎 𝚒𝚗 𝙸𝚗𝚍𝚒𝚊: 𝙿𝚘𝚖𝚎𝚕𝚕𝚒 𝚎𝚡𝚙𝚊𝚗𝚍𝚎𝚍 𝚐𝚕𝚘𝚋𝚊𝚕𝚕𝚢 𝚒𝚗 𝙼𝚊𝚛𝚌𝚑 𝟸𝟶𝟸𝟼, 𝚒𝚗𝚌𝚕𝚞𝚍𝚒𝚗𝚐 𝙸𝚗𝚍𝚒𝚊. 𝙽𝚘 𝚌𝚛𝚎𝚍𝚒𝚝 𝚌𝚊𝚛𝚍, 𝚗𝚘 𝚠𝚊𝚒𝚝𝚕𝚒𝚜𝚝 — 𝚓𝚞𝚜𝚝 𝚢𝚘𝚞𝚛 𝚠𝚎𝚋𝚜𝚒𝚝𝚎 𝚄𝚁𝙻．

𝙽𝚘𝚝𝚎: 𝚃𝚑𝚎 𝚝𝚘𝚘𝚕𝚜 𝚊𝚗𝚍 𝚊𝚗𝚊𝚕𝚢𝚜𝚒𝚜 𝚏𝚎𝚊𝚝𝚞𝚛𝚎𝚍 𝚒𝚗 𝚝𝚑𝚒𝚜 𝚜𝚎𝚌𝚝𝚒𝚘𝚗 𝚍𝚎𝚖𝚘𝚗𝚜𝚝𝚛𝚊𝚝𝚎𝚍 𝚌𝚕𝚎𝚊𝚛 𝚟𝚊𝚕𝚞𝚎 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚘𝚞𝚛 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚝𝚎𝚜𝚝𝚒𝚗𝚐. 𝙾𝚞𝚛 𝚛𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗𝚜 𝚊𝚛𝚎 𝚎𝚗𝚝𝚒𝚛𝚎𝚕𝚢 𝚒𝚗𝚍𝚎𝚙𝚎𝚗𝚍𝚎𝚗𝚝 𝚊𝚗𝚍 𝚗𝚘𝚝 𝚒𝚗𝚏𝚕𝚞𝚎𝚗𝚌𝚎𝚍 𝚋𝚢 𝚝𝚑𝚎 𝚝𝚘𝚘𝚕 𝚌𝚛𝚎𝚊𝚝𝚘𝚛𝚜．

Paradox of AI pay: Fewer jobs, more tokens

Jensen Huang wants us to believe that tokens will become the currency of the AI economy. He posits that in the near future, engineers won’t just earn salaries but also receive annual “token budgets”—a pool of AI compute they can spend to amplify their productivity. The basic assumption is that the need for tokens will keep growing.

But what is a token? Simply put, a token is a small chunk of text, words or parts of words that an AI model reads or generates. Every prompt and response is broken into these units, and companies are billed based on how many tokens are processed.

But as systems get more efficient, models will require fewer tokens per task, thanks to better architectures, caching, and specialised systems. If that trend continues, tokens become cheaper, more abundant, and ultimately less meaningful as a unit of value.

Besides, as pointed above, more money is now flowing into AI inference than into training new LLMs from scratch. That complicates the token narrative. Inference is highly competitive, margins are thinner, and efficiency gains are relentless. If companies optimise for fewer tokens per task, the very metric Huang wants to monetise starts shrinking.

Further, token budgets are not like salaries; they are tied to specific models, vendors, and pricing structures. If the underlying AI stack shifts—say from proprietary systems to open-source alternatives, or from large general models to smaller task-specific ones—the “value” of those tokens can erode quickly.

Huang’s idea also assumes that tokens will remain the primary interface to AI. That may already be changing. The industry is moving from “generate text” to “complete tasks”. Agentic AI systems do not just consume tokens, they call APIs, execute the workflows, and deliver outcomes. In such a world, tokens used may not matter. What matters is whether the job gets done or not.

AI BITS & BYTES

Can AI cure cancer in humans?

A techie made headlines when he said that he used ChatGPT and Google DeepMind’s AlphaFold to make a vaccine for his dog that had just months to live. Elon Musk later reposted that “Paul Conyngham, a Sydney-based technology entrepreneur and AI consultant, used Grok to finalise an mRNA vaccine construct for his dog Rosie’s mast cell cancer”.

That said, Emilia Javorsky, MD, MPH and Director of the Futures Program at Future of Life Institute, cautions that “...current AI capabilities—thoughtfully applied by people who understand biomedical problems deeply—can help...but realising this potential at scale requires abandoning the illusion of intelligence and embracing the complex, unglamorous work of fixing systems that actually stand between discovery and cure”.

She emphasises that today’s “...AI capabilities already deliver real medical value, not through Big Tech’s pursuit of god-like machines, but through targeted solutions for specific problems. Even DeepMind’s AlphaFold succeeded by focusing on one well-defined challenge: protein folding...”

Can AI predict the next market crash?

MiroFish, touted as a “next-generation AI prediction engine” by its creators, extracts real-world information as ‘seed’ data (like breaking news, policy drafts and financial signals) and converts it into a knowledge graph.

The tool could be used to predict how traders and retail investors might react to specific news, forecast public reactions and opinions prior to policy announcements, test the effectiveness of marketing strategies by gauging user feedback in a virtual social environment, or build character worlds in order to deduce plot developments or write lost story endings.

More AI Reading Material

• Elon Musk sues OpenAI for $134 million with promise to donate his winnings

• AI backlash grows as people start movements promoting human-only products

• Elon Musk vows xAI will catch up with Gemini to Claude and ChatGPT

• Why is Meta delaying the rollout of its new AI model called Avocado?

• Will AI change how computer science is taught? Perplexity CEO thinks so

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Madhumita Sen Choudhury. Produced by Tushar Deep Singh.

The company told the court that hundreds of millions of dollars in revenue could be at risk. Its CFO said revenue from defence contractors and others reliant on the DoD could fall 50-100% if the designation stands.

Meanwhile, more than 30 employees from OpenAI and Google DeepMind have filed an amicus brief supporting Anthropic’s lawsuit. Among those named are Jeff Dean and researchers including Zhengdong Wang, Alexander Matt Turner and Noah Siegel, who argue that penalising a leading US AI firm could harm the country’s scientific and industrial competitiveness.

The dispute underscores the growing friction between national security priorities and the rapidly expanding AI industry. As governments increasingly treat Advanced AI as strategic infrastructure, decisions on procurement, security vetting and supply-chain risk could shape which companies gain access to lucrative defence contracts, and which are sidelined.

For AI firms, the outcome could set an important precedent on how far governments can go in restricting commercial players from sensitive national security ecosystems. In this context, you may also read:

• AI & War: Principle, Profit and Agentic AI combat

• Anthropic’s standoff with the Pentagon shakes up AI talent race

In this week’s edition of Tech Talk:

• Why GPT-5.4 signals the next leap in Autonomous AI

• AI Tool Of The Week: Claude Legal plugin

• Are you better off as a cook or plumber in this AI age?

PREVIOUSLY, ON TECH TALK

GPT-5.4: The next leap in Autonomous AI?

GPT-5.4 is described by OpenAI Inc. as its most token-efficient reasoning model, delivering faster responses and lower costs. It comes in two versions: GPT-5.4 and the higher-end GPT-5.4 Pro.

The large language model can interpret screenshots, browse the web, and execute keyboard and mouse commands across apps, enabling multistep workflows that previously required human input.

GPT-5.4 Thinking can outline its reasoning plan upfront, letting users adjust direction mid-response. It also strengthens deep web research for highly specific queries and maintains context better during longer reasoning tasks, producing faster, more relevant answers.

Alongside improvements in reasoning, coding and professional knowledge work, GPT-5.4 enables more reliable agents, faster developer workflows and higher-quality outputs across ChatGPT, the API and OpenAI Codex.

In Codex and the API, GPT-5.4 is the first general-purpose model that allows agents to operate computers and run complex workflows across applications. It supports up to one million tokens of context, enabling planning and execution across long tasks.

GPT-5.4 also improves how models interact with large tool ecosystems through tool search, helping agents find and use the right tools more efficiently.

According to OpenAI, GPT-5.4 is its most factual model yet. On the MMMU-Pro benchmark for visual reasoning, GPT-5.4 scores 81.2%, up from GPT-5.2’s 79.5%. On Toolathlon, which measures how well AI agents use real-world tools and APIs, GPT-5.4 achieves higher accuracy than GPT-5.2 in fewer turns.

Even so, OpenAI faces resistance from rivals such as Google’s Gemini and Anthropic’s Claude. The chart below tells you the story.

AI TOOL OF THE WEEK

By AI&Beyond, with Jaspreet Bindra and Anuj Magazine

𝚃𝚑𝚎 𝙰𝙸 𝚑𝚊𝚌𝚔 𝚠𝚎 𝚞𝚗𝚕𝚘𝚌𝚔𝚎𝚍 𝚝𝚘𝚍𝚊𝚢 𝚒𝚜 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝙲𝚕𝚊𝚞𝚍𝚎 𝙻𝚎𝚐𝚊𝚕 𝚙𝚕𝚞𝚐𝚒𝚗．

𝚆𝚑𝚊𝚝 𝚙𝚛𝚘𝚋𝚕𝚎𝚖 𝚍𝚘𝚎𝚜 𝚒𝚝 𝚜𝚘𝚕𝚟𝚎? 𝙻𝚎𝚐𝚊𝚕 𝚝𝚎𝚊𝚖𝚜 𝚏𝚊𝚌𝚎 𝚊 𝚚𝚞𝚒𝚎𝚝 𝚌𝚛𝚒𝚜𝚒𝚜: 𝚖𝚘𝚞𝚗𝚝𝚊𝚒𝚗𝚜 𝚘𝚏 𝚛𝚘𝚞𝚝𝚒𝚗𝚎 𝚌𝚘𝚗𝚝𝚛𝚊𝚌𝚝𝚜, 𝙽𝙳𝙰𝚜 𝚊𝚗𝚍 𝚌𝚘𝚖𝚙𝚕𝚒𝚊𝚗𝚌𝚎 𝚍𝚘𝚌𝚞𝚖𝚎𝚗𝚝𝚜 𝚝𝚑𝚊𝚝 𝚍𝚎𝚖𝚊𝚗𝚍 𝚌𝚊𝚛𝚎𝚏𝚞𝚕 𝚛𝚎𝚟𝚒𝚎𝚠 𝚋𝚞𝚝 𝚌𝚘𝚗𝚜𝚞𝚖𝚎 𝚑𝚞𝚐𝚎 𝚝𝚒𝚖𝚎 𝚊𝚗𝚍 𝚖𝚘𝚗𝚎𝚢. 𝙰 𝚜𝚒𝚗𝚐𝚕𝚎 𝚟𝚎𝚗𝚍𝚘𝚛 𝚌𝚘𝚗𝚝𝚛𝚊𝚌𝚝 𝚌𝚊𝚗 𝚝𝚊𝚔𝚎 𝚝𝚠𝚘 𝚑𝚘𝚞𝚛𝚜; 𝟻𝟶 𝙽𝙳𝙰𝚜 𝚌𝚊𝚗 𝚖𝚎𝚊𝚗 𝚍𝚊𝚢𝚜 𝚘𝚏 𝚙𝚊𝚛𝚊𝚕𝚎𝚐𝚊𝚕 𝚠𝚘𝚛𝚔. 𝚈𝚎𝚝 𝚖𝚒𝚜𝚜𝚒𝚗𝚐 𝚊 𝚜𝚒𝚗𝚐𝚕𝚎 𝙶𝙳𝙿𝚁 𝚌𝚕𝚊𝚞𝚜𝚎, 𝚘𝚟𝚎𝚛𝚕𝚘𝚘𝚔𝚒𝚗𝚐 𝚊𝚗 𝚊𝚞𝚝𝚘-𝚛𝚎𝚗𝚎𝚠𝚊𝚕 𝚝𝚛𝚊𝚙 𝚘𝚛 𝚒𝚐𝚗𝚘𝚛𝚒𝚗𝚐 𝚊𝚗 𝚎𝚡𝚙𝚒𝚛𝚎𝚍 𝚌𝚎𝚛𝚝𝚒𝚏𝚒𝚌𝚊𝚝𝚒𝚘𝚗 𝚌𝚊𝚗 𝚙𝚛𝚘𝚟𝚎 𝚌𝚘𝚜𝚝𝚕𝚢.

𝚃𝚑𝚒𝚜 𝚒𝚜 𝚗𝚘𝚝 𝚓𝚞𝚜𝚝 𝚊 𝙱𝚒𝚐𝙻𝚊𝚠 𝚙𝚛𝚘𝚋𝚕𝚎𝚖. 𝙸𝚗-𝚑𝚘𝚞𝚜𝚎 𝚝𝚎𝚊𝚖𝚜 𝚊𝚝 𝚖𝚒𝚍-𝚖𝚊𝚛𝚔𝚎𝚝 𝚏𝚒𝚛𝚖𝚜, 𝚌𝚘𝚖𝚙𝚕𝚒𝚊𝚗𝚌𝚎 𝚘𝚏𝚏𝚒𝚌𝚎𝚛𝚜 𝚊𝚗𝚍 𝙷𝚁 𝚍𝚎𝚙𝚊𝚛𝚝𝚖𝚎𝚗𝚝𝚜 𝚏𝚊𝚌𝚎 𝚝𝚑𝚎 𝚜𝚊𝚖𝚎 𝚋𝚘𝚝𝚝𝚕𝚎𝚗𝚎𝚌𝚔 𝚍𝚊𝚒𝚕𝚢. 𝚃𝚑𝚎 𝚠𝚘𝚛𝚔 𝚒𝚜 𝚛𝚎𝚙𝚎𝚝𝚒𝚝𝚒𝚟𝚎, 𝚑𝚒𝚐𝚑-𝚜𝚝𝚊𝚔𝚎𝚜 𝚊𝚗𝚍 𝚎𝚡𝚙𝚎𝚗𝚜𝚒𝚟𝚎 𝚝𝚘 𝚘𝚞𝚝𝚜𝚘𝚞𝚛𝚌𝚎. 𝚃𝚑𝚎 s𝚙𝚎𝚌𝚒𝚊𝚕𝚒𝚜𝚎𝚍 𝚕𝚎𝚐𝚊𝚕 𝙰𝙸 𝚝𝚘𝚘𝚕𝚜 𝚛𝚎𝚖𝚊𝚒𝚗 𝚞𝚗𝚊𝚏𝚏𝚘𝚛𝚍𝚊𝚋𝚕𝚎 𝚏𝚘𝚛 𝚖𝚊𝚗𝚢. 𝚃𝚑𝚎 𝙲𝚕𝚊𝚞𝚍𝚎 𝙻𝚎𝚐𝚊𝚕 𝚙𝚕𝚞𝚐𝚒𝚗 𝚊𝚒𝚖𝚜 𝚝𝚘 𝚌𝚑𝚊𝚗𝚐𝚎 𝚝𝚑𝚊𝚝 𝚎𝚚𝚞𝚊𝚝𝚒𝚘𝚗.

𝙷𝚘𝚠 𝚝𝚘 𝚊𝚌𝚌𝚎𝚜𝚜: 𝚌𝚕𝚊𝚞𝚍𝚎.𝚊𝚒 𝚘𝚛 𝙲𝚕𝚊𝚞𝚍𝚎 𝙳𝚎𝚜𝚔𝚝𝚘𝚙 𝚊𝚙𝚙 𝚏𝚘𝚛 𝚊𝚐𝚎𝚗𝚝𝚒𝚌 𝙲𝚘𝚠𝚘𝚛𝚔 𝚖𝚘𝚍𝚎．

𝙲𝚕𝚊𝚞𝚍𝚎 𝙻𝚎𝚐𝚊𝚕 𝙰𝙸 𝚌𝚊𝚗 𝚑𝚎𝚕𝚙 𝚢𝚘𝚞:

• 𝚁𝚎𝚟𝚒𝚎𝚠 𝚌𝚘𝚗𝚝𝚛𝚊𝚌𝚝𝚜 𝚊𝚞𝚝𝚘𝚖𝚊𝚝𝚒𝚌𝚊𝚕𝚕𝚢, 𝚞𝚜𝚒𝚗𝚐 /𝚛𝚎𝚟𝚒𝚎𝚠-𝚌𝚘𝚗𝚝𝚛𝚊𝚌𝚝: U𝚙𝚕𝚘𝚊𝚍 𝚊 𝚟𝚎𝚗𝚍𝚘𝚛 𝚊𝚐𝚛𝚎𝚎𝚖𝚎𝚗𝚝 𝚊𝚗𝚍 𝚛𝚎𝚌𝚎𝚒𝚟𝚎 𝚊 𝙶𝚁𝙴𝙴𝙽/𝚈𝙴𝙻𝙻𝙾𝚆/𝚁𝙴𝙳 𝚛𝚒𝚜𝚔 𝚖𝚊𝚝𝚛𝚒𝚡 𝚠𝚒𝚝𝚑 𝚜𝚙𝚎𝚌𝚒𝚏𝚒𝚌 𝚛𝚎𝚍𝚕𝚒𝚗𝚎 𝚜𝚞𝚐𝚐𝚎𝚜𝚝𝚒𝚘𝚗𝚜, 𝚒𝚗 𝚖𝚒𝚗𝚞𝚝𝚎𝚜．

• 𝚃𝚛𝚒𝚊𝚐𝚎 𝙽𝙳𝙰𝚜 𝚒𝚗 𝚋𝚞𝚕𝚔 𝚞𝚜𝚒𝚗𝚐 /𝚝𝚛𝚒𝚊𝚐𝚎-𝚗𝚍𝚊: P𝚛𝚘𝚌𝚎𝚜𝚜 𝚍𝚘𝚣𝚎𝚗𝚜 𝚘𝚏 𝙽𝙳𝙰𝚜 𝚊𝚝 𝚘𝚗𝚌𝚎, 𝚊𝚞𝚝𝚘-𝚌𝚊𝚝𝚎𝚐𝚘𝚛𝚒𝚜𝚎 𝚝𝚑𝚎𝚖, 𝚊𝚗𝚍 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎 𝚊 𝚜𝚞𝚖𝚖𝚊𝚛𝚢 𝚛𝚎𝚙𝚘𝚛𝚝 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚝𝚘𝚞𝚌𝚑𝚒𝚗𝚐 𝚎𝚊𝚌𝚑 𝚏𝚒𝚕𝚎 𝚒𝚗𝚍𝚒𝚟𝚒𝚍𝚞𝚊𝚕𝚕𝚢．

• 𝙶𝚎𝚗𝚎𝚛𝚊𝚝𝚎 𝚛𝚎𝚐𝚞𝚕𝚊𝚝𝚘𝚛𝚢 𝚋𝚛𝚒𝚎𝚏𝚒𝚗𝚐𝚜 𝚞𝚜𝚒𝚗𝚐 /𝚋𝚛𝚒𝚎𝚏: G𝚎𝚝 𝚜𝚝𝚛𝚞𝚌𝚝𝚞𝚛𝚎𝚍 𝚍𝚊𝚒𝚕𝚢 𝚞𝚙𝚍𝚊𝚝𝚎𝚜 𝚘𝚗 𝚝𝚘𝚙𝚒𝚌𝚜 𝚕𝚒𝚔𝚎 𝙴𝚄 𝙰𝙸 𝙰𝚌𝚝 𝚍𝚎𝚊𝚍𝚕𝚒𝚗𝚎𝚜 𝚘𝚛 𝙶𝙳𝙿𝚁 𝚌𝚑𝚊𝚗𝚐𝚎𝚜, 𝚏𝚘𝚛𝚖𝚊𝚝𝚝𝚎𝚍 𝚏𝚘𝚛 𝚎𝚡𝚎𝚌𝚞𝚝𝚒𝚟𝚎 𝚍𝚒𝚜𝚝𝚛𝚒𝚋𝚞𝚝𝚒𝚘𝚗．

𝚆𝚑𝚊𝚝 𝚖𝚊𝚔𝚎𝚜 𝙲𝚕𝚊𝚞𝚍𝚎 𝙻𝚎𝚐𝚊𝚕 𝙰𝙸 𝚜𝚝𝚊𝚗𝚍 𝚘𝚞𝚝?

•⁠ 𝚈𝚘𝚞𝚛 𝚙𝚕𝚊𝚢𝚋𝚘𝚘𝚔, 𝚗𝚘𝚝 𝚊 𝚐𝚎𝚗𝚎𝚛𝚒𝚌 𝚝𝚎𝚖𝚙𝚕𝚊𝚝𝚎: 𝙴𝚟𝚎𝚛𝚢 𝚛𝚎𝚟𝚒𝚎𝚠 𝚒𝚜 𝚖𝚎𝚊𝚜𝚞𝚛𝚎𝚍 𝚊𝚐𝚊𝚒𝚗𝚜𝚝 𝚝𝚑𝚎 𝚜𝚝𝚊𝚗𝚍𝚊𝚛𝚍𝚜 𝚢𝚘𝚞 𝚍𝚎𝚏𝚒𝚗𝚎, 𝚖𝚊𝚔𝚒𝚗𝚐 𝚘𝚞𝚝𝚙𝚞𝚝𝚜 𝚌𝚘𝚗𝚜𝚒𝚜𝚝𝚎𝚗𝚝 𝚠𝚒𝚝𝚑 𝚢𝚘𝚞𝚛 𝚘𝚛𝚐𝚊𝚗𝚒𝚜𝚊𝚝𝚒𝚘𝚗’𝚜 𝚊𝚌𝚝𝚞𝚊𝚕 𝚛𝚒𝚜𝚔 𝚝𝚘𝚕𝚎𝚛𝚊𝚗𝚌𝚎, 𝚗𝚘𝚝 𝚊 𝚘𝚗𝚎-𝚜𝚒𝚣𝚎-𝚏𝚒𝚝𝚜-𝚊𝚕𝚕 𝚏𝚛𝚊𝚖𝚎𝚠𝚘𝚛𝚔．

•⁠ 𝙰𝚐𝚎𝚗𝚝𝚒𝚌 𝚋𝚊𝚝𝚌𝚑 𝚙𝚛𝚘𝚌𝚎𝚜𝚜𝚒𝚗𝚐: 𝙸𝚗 𝙲𝚘𝚠𝚘𝚛𝚔 𝚖𝚘𝚍𝚎, 𝚢𝚘𝚞 𝚑𝚊𝚗𝚍 𝚘𝚏𝚏 𝚊 𝚏𝚘𝚕𝚍𝚎𝚛 𝚘𝚏 𝚍𝚘𝚌𝚞𝚖𝚎𝚗𝚝𝚜 𝚊𝚗𝚍 𝚛𝚎𝚝𝚞𝚛𝚗 𝚝𝚘 𝚏𝚒𝚗𝚒𝚜𝚑𝚎𝚍, 𝚌𝚊𝚝𝚎𝚐𝚘𝚛𝚒𝚜𝚎𝚍 𝚠𝚘𝚛𝚔 — 𝚗𝚘𝚝 𝚊 𝚝𝚞𝚛𝚗-𝚋𝚢-𝚝𝚞𝚛𝚗 𝚌𝚘𝚗𝚟𝚎𝚛𝚜𝚊𝚝𝚒𝚘𝚗．

•⁠ 𝙳𝚛𝚊𝚖𝚊𝚝𝚒𝚌𝚊𝚕𝚕𝚢 𝚖𝚘𝚛𝚎 𝚊𝚌𝚌𝚎𝚜𝚜𝚒𝚋𝚕𝚎: 𝙸𝚝 𝚞𝚗𝚍𝚎𝚛𝚌𝚞𝚝𝚜 𝚜𝚙𝚎𝚌𝚒𝚊𝚕𝚒𝚜𝚎𝚍 𝚕𝚎𝚐𝚊𝚕 𝙰𝙸 𝚝𝚘𝚘𝚕𝚜 𝚜𝚒𝚐𝚗𝚒𝚏𝚒𝚌𝚊𝚗𝚝𝚕𝚢, 𝚋𝚛𝚒𝚗𝚐𝚒𝚗𝚐 𝚌𝚊𝚙𝚊𝚋𝚒𝚕𝚒𝚝𝚒𝚎𝚜 𝚙𝚛𝚎𝚟𝚒𝚘𝚞𝚜𝚕𝚢 𝚊𝚟𝚊𝚒𝚕𝚊𝚋𝚕𝚎 𝚘𝚗𝚕𝚢 𝚝𝚘 𝚕𝚊𝚛𝚐𝚎 𝚕𝚊𝚠 𝚏𝚒𝚛𝚖𝚜 𝚠𝚒𝚝𝚑𝚒𝚗 𝚛𝚎𝚊𝚌𝚑 𝚘𝚏 𝚒𝚗-𝚑𝚘𝚞𝚜𝚎 𝚝𝚎𝚊𝚖𝚜 𝚊𝚗𝚍 𝚖𝚒𝚍-𝚖𝚊𝚛𝚔𝚎𝚝 𝚘𝚛𝚐𝚊𝚗𝚒𝚜𝚊𝚝𝚒𝚘𝚗𝚜.

𝙽𝚘𝚝𝚎: 𝚃𝚑𝚎 𝚝𝚘𝚘𝚕𝚜 𝚊𝚗𝚍 𝚊𝚗𝚊𝚕𝚢𝚜𝚒𝚜 𝚏𝚎𝚊𝚝𝚞𝚛𝚎𝚍 𝚒𝚗 𝚝𝚑𝚒𝚜 𝚜𝚎𝚌𝚝𝚒𝚘𝚗 𝚍𝚎𝚖𝚘𝚗𝚜𝚝𝚛𝚊𝚝𝚎𝚍 𝚌𝚕𝚎𝚊𝚛 𝚟𝚊𝚕𝚞𝚎 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚘𝚞𝚛 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚝𝚎𝚜𝚝𝚒𝚗𝚐. 𝙾𝚞𝚛 𝚛𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗𝚜 𝚊𝚛𝚎 𝚎𝚗𝚝𝚒𝚛𝚎𝚕𝚢 𝚒𝚗𝚍𝚎𝚙𝚎𝚗𝚍𝚎𝚗𝚝 𝚊𝚗𝚍 𝚗𝚘𝚝 𝚒𝚗𝚏𝚕𝚞𝚎𝚗𝚌𝚎𝚍 𝚋𝚢 𝚝𝚑𝚎 𝚝𝚘𝚘𝚕 𝚌𝚛𝚎𝚊𝚝𝚘𝚛𝚜．

AI BITS & BYTES

Are you better off as a cook or plumber in this AI age?

While earlier reports from the World Economic Forum and Microsoft Corp. suggest that blue-collar workers may weather the AI tsunami better than many white-collar roles, a new study by Anthropic reinforces the trend. It finds that cooks, motorcycle mechanics, lifeguards, bartenders and dishwashers face the least risk from AI, while computer programmers, customer service agents and data-entry operators are among the most exposed.

The study combines data from the O*NET occupational database — Anthropic’s own economic index — and research by Tyna Eloundou on whether large language models can significantly accelerate workplace tasks.

You can find the full Anthropic report, here.

AI creates 46 new billionaires, but at a cost

AI is continuing to create new billionaires, with 114 of the world’s 4,020 billionaires now linked to AI companies. However, this wealth creation coincides with a parallel wealth destruction, according to the Hurun Global Rich List 2026. The number of AI billionaires currently stands at 114, with a little less than half (46) being new entrants to the list, suggesting a breakneck pace of wealth creation.

AI models can reveal your real identity, warn researchers

A new study has warned that Generative AI has made it vastly easier for malicious hackers to uncover the real identities of anonymous social media users.

According to a report by The Guardian citing the study, LLMs can now successfully match anonymous online profiles with actual identities by synthesising seemingly harmless information posted across different platforms.

The report notes that AI researchers Simon Lermen and Daniel Paleka found that the technology behind platforms like ChatGPT makes sophisticated privacy attacks highly cost-effective. The researchers ran an experiment where they fed anonymous accounts into an AI and instructed it to scrape all available information it could.

OpenAI’s robotics head quits after Pentagon deal

Caitlin Kalinowski, head of OpenAI’s robotics team has resigned from the position. In a post on X (formerly Twitter), she said the decision was directly tied to the company’s new deal with the Department of War. “I resigned from OpenAI. I care deeply about the robotics team and the work we built together. This wasn’t an easy call,” she wrote.

Microsoft integrates Anthropic AI into Copilot

Microsoft is integrating Anthropic’s AI tools into its Copilot platform to meet rising interest in autonomous agents, shortly after the startup’s latest innovations triggered a decline in software equities.

‘Copilot Cowork’, a utility modelled after Anthropic’s popular Claude Cowork system, has intrigued Silicon Valley through its capacity to manage intricate operations like app development, spreadsheet construction and large-scale data management with minimal human intervention, reported Reuters.

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Madhumita Sen Choudhury. Produced by Tushar Deep Singh.

Anthropic said it cannot “in good conscience accede to their (Department of War’s) request”, but as we pointed out last week, the confrontation effectively jeopardises Anthropic’s long-standing defence contracts with the government, and may make its investors jittery till things become clearer.

In this week’s edition of Tech Talk:

• Citrini vs Citadel and AI’s days of future-past

• AI Tool of the Week: Lyria 3 in Google Gemini

• Did Grok predict the Iran-Israel-US war?

A Matter of Principle vs Profit

An Axios article highlighted that Anthropic has raised more than $60 billion from over 200 “venture capital” investors, half of which came in just the last two months. These new investors will expectedly be nervous as US defence contractors—including the likes of Lockheed Martin—are expected to remove Anthropic’s AI tools from their supply chains days after President Donald Trump ordered all federal agencies to immediately stop using them.

Meanwhile, Sam Altman has been making many mixed comments. On the one hand, he admitted OpenAI’s rushed deal with the DoW amid its battle with Anthropic over safety and ethical concerns looked “opportunistic and sloppy”. He said his company wants to work “through democratic processes” and urged DoW to reconsider its decision about Anthropic.

On the other hand, Altman’s latest post says that the ChatGPT maker has made some “additions in our agreement to make our principles very clear…”.

“[We]…shouldn’t have rushed to get this out (referring to the deal with DoW) on Friday. The issues are super complex, and demand clear communication. We were genuinely trying to de-escalate things and avoid a much worse outcome, but I think it just looked opportunistic and sloppy... In my conversations over the weekend, I reiterated that Anthropic should not be designated as a SCR, and that we hope the DoW offers them the same terms we’ve agreed to.”

Expectedly, Amodei has accused Altman of “gaslighting”, saying that Altman is falsely “presenting himself as a peacemaker and dealmaker”. Netizens, too, are not amused. They applauded Anthropic’s “principled” stand, pushing Claude to No.1 on the iOS free app rankings in late January, even as many users uninstalled ChatGPT as a mark of protest. Users on social media platforms like Reddit and X claim to be switching to Claude due to the OpenAI’s “opportunistic” stance.

While these moves are unlikely to change the DoW’s perception of Anthropic, or even trigger a change of heart, it is sending a strong signal to Big Tech and governments that principles and guardrails remain a valuable and treasured commodity.

Beginning of Agentic AI wars

The continuously changing geopolitical scenario with abrupt wars is queering the pitch. In the context of AI, it has brought to fore the question of governments unleashing semi- and fully-autonomous weapons on their enemies.

And It’s with good reason. Unlike traditional AI or even Generative AI, agentic systems act as autonomous officers. Given an objective, they pursue it, adapt when things go wrong, and coordinate with other agents. Agentic AI can thus close the OODA (Observe-Orient-Decide-Act) loop on its own, planning missions, fusing multi-sensor data, adapting mid-course, and executing tasks with minimal human input.

Three conflicts illustrate the spectrum:

• Israel-Iran saw AI as a Command-and-Control (C2) integration layer synchronising multi-domain strikes, with humans retaining lethal authorisation.

• Ukraine’s “Operation Spider’s Web” deployed autonomous targeting that activated when human control was jammed with the Gogol-M mothership-subagent architecture approaching true autonomy.

• The Mexico cartel operation was AI-assisted ISR feeding a human raid but alarmingly, Jalisco New Generation Cartel operatives embedded in Ukrainian drone units are actively reverse-engineering battlefield AI for criminal use.

All these developments signal a decisive shift in how wars may be fought even as regulation hasn’t kept pace because of ambiguity in defining these terms, dual-use concerns, strategic competition, and verification issues.

Moreover, as conflicts in Ukraine and West Asia intensify, questions are mounting about whether multi-agent AI architectures, systems where multiple autonomous agents coordinate and adapt in real time, are already shaping the battlefield.

In the Ukraine war, for instance, both sides have deployed AI-enabled drones, including loitering munitions such as the Shahed 136. These systems can navigate autonomously and strike pre-programmed targets without live piloting. But they rely on narrow AI, computer vision, GPS, and fixed targeting logic, not decentralised multi-agent coordination.

Similarly, Turkey’s Kargu-2 drew global attention after a UN panel suggested it may have operated in an autonomous engagement mode in Libya. Even here, autonomy referred to onboard target recognition and strike capability, not to multiple AI agents dynamically coordinating strategy.

The most advanced demonstrations of multi-agent architectures remain experimental. DARPA’s Offensive Swarm-enabled Tactics (OFFSET) Program tested large drone swarms capable of decentralised coordination and real-time adaptation. Likewise, Pentagon’s Perdix microdrones showcased distributed “swarm intelligence” without a central controller. But these were controlled demonstrations, not confirmed battlefield deployments.

In recent strikes involving Iran, reporting suggests AI models were used for analysis and planning, not autonomous swarm combat. Multi-agent AI implies decentralised decision-making, role allocation, and emergent behaviour under battlefield stress. Public evidence shows militaries are advancing toward that capability but have not yet operationalised fully self-directed, multi-agent lethal systems.

For now, autonomy in war is real. True multi-agent AI warfare remains largely in the lab. In the absence of sufficient guardrails, we must be grateful that it’s so. At least, for now.

AI TOOL OF THE WEEK

by AI&Beyond, with Jaspreet Bindra and Anuj Magazine

𝚃𝚑𝚎 𝙰𝙸 𝚑𝚊𝚌𝚔 𝚠𝚎 𝚞𝚗𝚕𝚘𝚌𝚔𝚎𝚍 𝚝𝚘𝚍𝚊𝚢 𝚒𝚜 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚊 𝚝𝚘𝚘𝚕: 𝙻𝚢𝚛𝚒𝚊 𝟹 𝚒𝚗 𝙶𝚘𝚘𝚐𝚕𝚎 𝙶𝚎𝚖𝚒𝚗𝚒．

𝚆𝚑𝚊𝚝 𝚙𝚛𝚘𝚋𝚕𝚎𝚖 𝚍𝚘𝚎𝚜 𝚒𝚝 𝚜𝚘𝚕𝚟𝚎? 𝙴𝚟𝚎𝚛𝚢 𝚠𝚎𝚎𝚔, 𝚝𝚎𝚊𝚖𝚜 𝚊𝚌𝚛𝚘𝚜𝚜 𝚖𝚊𝚛𝚔𝚎𝚝𝚒𝚗𝚐, 𝙻&𝙳, 𝚊𝚗𝚍 𝚌𝚘𝚖𝚖𝚞𝚗𝚒𝚌𝚊𝚝𝚒𝚘𝚗𝚜 𝚑𝚒𝚝 𝚝𝚑𝚎 𝚜𝚒𝚖𝚒𝚕𝚊𝚛 𝚒𝚗𝚟𝚒𝚜𝚒𝚋𝚕𝚎 𝚠𝚊𝚕𝚕: 𝚝𝚑𝚎𝚢 𝚑𝚊𝚟𝚎 𝚐𝚛𝚎𝚊𝚝 𝚌𝚘𝚗𝚝𝚎𝚗𝚝, 𝚖𝚊𝚢𝚋𝚎 𝚊 𝚝𝚛𝚊𝚒𝚗𝚒𝚗𝚐 𝚟𝚒𝚍𝚎𝚘, 𝚊 𝚙𝚛𝚘𝚍𝚞𝚌𝚝 𝚕𝚊𝚞𝚗𝚌𝚑 𝚛𝚎𝚎𝚕, 𝚊𝚗 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚝𝚘𝚠𝚗𝚑𝚊𝚕𝚕 𝚛𝚎𝚌𝚘𝚛𝚍𝚒𝚗𝚐 𝚋𝚞𝚝 𝚗𝚘 𝚋𝚞𝚍𝚐𝚎𝚝, 𝚗𝚘 𝚝𝚒𝚖𝚎, 𝚊𝚗𝚍 𝚗𝚘 𝚎𝚡𝚙𝚎𝚛𝚝𝚒𝚜𝚎 𝚝𝚘 𝚐𝚒𝚟𝚎 𝚒𝚝 𝚊 𝚜𝚘𝚞𝚗𝚍𝚝𝚛𝚊𝚌𝚔 𝚝𝚑𝚊𝚝 𝚊𝚌𝚝𝚞𝚊𝚕𝚕𝚢 𝚏𝚒𝚝𝚜. 𝚂𝚘, 𝚝𝚑𝚎𝚢 𝚎𝚒𝚝𝚑𝚎𝚛 𝚙𝚊𝚢 𝚏𝚘𝚛 𝚜𝚝𝚘𝚌𝚔 𝚖𝚞𝚜𝚒𝚌 𝚝𝚑𝚊𝚝 𝚜𝚘𝚞𝚗𝚍𝚜 𝚐𝚎𝚗𝚎𝚛𝚒𝚌, 𝚜𝚙𝚎𝚗𝚍 𝚠𝚎𝚎𝚔𝚜 𝚌𝚘𝚘𝚛𝚍𝚒𝚗𝚊𝚝𝚒𝚗𝚐 𝚠𝚒𝚝𝚑 𝚊 𝚌𝚘𝚖𝚙𝚘𝚜𝚎𝚛, 𝚘𝚛 𝚜𝚒𝚖𝚙𝚕𝚢 𝚜𝚑𝚒𝚙 𝚝𝚑𝚎 𝚌𝚘𝚗𝚝𝚎𝚗𝚝 𝚠𝚒𝚝𝚑 𝚗𝚘 𝚊𝚞𝚍𝚒𝚘.

𝙽𝚘𝚗𝚎 𝚘𝚏 𝚝𝚑𝚎𝚜𝚎 𝚘𝚙𝚝𝚒𝚘𝚗𝚜 𝚊𝚛𝚎 𝚐𝚘𝚘𝚍 𝚎𝚗𝚘𝚞𝚐𝚑. 𝙰 𝚖𝚒𝚍-𝚜𝚒𝚣𝚎𝚍 𝚌𝚘𝚖𝚙𝚊𝚗𝚢 𝚙𝚛𝚘𝚍𝚞𝚌𝚎𝚜 𝚍𝚘𝚣𝚎𝚗𝚜 𝚘𝚏 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚟𝚒𝚍𝚎𝚘𝚜, 𝚌𝚕𝚒𝚎𝚗𝚝-𝚏𝚊𝚌𝚒𝚗𝚐 𝚙𝚛𝚎𝚜𝚎𝚗𝚝𝚊𝚝𝚒𝚘𝚗𝚜, 𝚊𝚗𝚍 𝚜𝚘𝚌𝚒𝚊𝚕 𝚖𝚎𝚍𝚒𝚊 𝚊𝚜𝚜𝚎𝚝𝚜 𝚎𝚟𝚎𝚛𝚢 𝚖𝚘𝚗𝚝𝚑. 𝙴𝚊𝚌𝚑 𝚘𝚗𝚎 𝚒𝚜 𝚊𝚗 𝚘𝚙𝚙𝚘𝚛𝚝𝚞𝚗𝚒𝚝𝚢 𝚝𝚘 𝚌𝚘𝚖𝚖𝚞𝚗𝚒𝚌𝚊𝚝𝚎 𝚠𝚒𝚝𝚑 𝚎𝚖𝚘𝚝𝚒𝚘𝚗 𝚊𝚗𝚍 𝚖𝚘𝚜𝚝 𝚜𝚚𝚞𝚊𝚗𝚍𝚎𝚛 𝚒𝚝 𝚋𝚎𝚌𝚊𝚞𝚜𝚎 𝚘𝚛𝚒𝚐𝚒𝚗𝚊𝚕 𝚖𝚞𝚜𝚒𝚌 𝚏𝚎𝚎𝚕𝚜 𝚘𝚞𝚝 𝚘𝚏 𝚛𝚎𝚊𝚌𝚑．

𝙶𝚘𝚘𝚐𝚕𝚎 𝙶𝚎𝚖𝚒𝚗𝚒 𝙻𝚢𝚛𝚒𝚊 𝟹 𝚜𝚘𝚕𝚟𝚎𝚜 𝚝𝚑𝚎𝚜𝚎 𝚙𝚛𝚘𝚋𝚕𝚎𝚖𝚜. 𝚃𝚢𝚙𝚎 𝚊 𝚜𝚎𝚗𝚝𝚎𝚗𝚌𝚎 𝚍𝚎𝚜𝚌𝚛𝚒𝚋𝚒𝚗𝚐 𝚢𝚘𝚞𝚛 𝚋𝚛𝚊𝚗𝚍, 𝚢𝚘𝚞𝚛 𝚖𝚎𝚜𝚜𝚊𝚐𝚎, 𝚘𝚛 𝚢𝚘𝚞𝚛 𝚖𝚘𝚘𝚍. 𝙸𝚗 𝚞𝚗𝚍𝚎𝚛 𝟹𝟶 𝚜𝚎𝚌𝚘𝚗𝚍𝚜, 𝚢𝚘𝚞 𝚑𝚊𝚟𝚎 𝚊𝚗 𝚘𝚛𝚒𝚐𝚒𝚗𝚊𝚕, 𝟹𝟶-𝚜𝚎𝚌𝚘𝚗𝚍, 𝟺𝟾 𝚔𝙷𝚣 𝚜𝚝𝚎𝚛𝚎𝚘 𝚝𝚛𝚊𝚌𝚔 𝚠𝚒𝚝𝚑 𝚟𝚘𝚌𝚊𝚕𝚜, 𝚊𝚞𝚝𝚘-𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎𝚍 𝚕𝚢𝚛𝚒𝚌𝚜, 𝚊𝚗𝚍 𝚌𝚞𝚜𝚝𝚘𝚖 𝚌𝚘𝚟𝚎𝚛 𝚊𝚛𝚝. 𝙽𝚘 𝚌𝚘𝚖𝚙𝚘𝚜𝚎𝚛. 𝙽𝚘 𝚕𝚒𝚌𝚎𝚗𝚜𝚒𝚗𝚐. 𝙽𝚘 𝚙𝚛𝚘𝚖𝚙𝚝 𝚎𝚗𝚐𝚒𝚗𝚎𝚎𝚛𝚒𝚗𝚐 𝚍𝚎𝚐𝚛𝚎𝚎 𝚛𝚎𝚚𝚞𝚒𝚛𝚎𝚍.

𝙷𝚘𝚠 𝚝𝚘 𝚊𝚌𝚌𝚎𝚜𝚜: 𝚐𝚎𝚖𝚒𝚗𝚒.𝚐𝚘𝚘𝚐𝚕𝚎.𝚌𝚘𝚖

𝙻𝚢𝚛𝚒𝚊 𝟹 𝚌𝚊𝚗 𝚑𝚎𝚕𝚙 𝚢𝚘𝚞:

• 𝙶𝚎𝚗𝚎𝚛𝚊𝚝𝚎 𝚘𝚛𝚒𝚐𝚒𝚗𝚊𝚕 𝚜𝚘𝚞𝚗𝚍𝚝𝚛𝚊𝚌𝚔𝚜 𝚏𝚘𝚛 𝚝𝚛𝚊𝚒𝚗𝚒𝚗𝚐 𝚟𝚒𝚍𝚎𝚘𝚜 𝚊𝚗𝚍 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚌𝚘𝚖𝚖𝚞𝚗𝚒𝚌𝚊𝚝𝚒𝚘𝚗𝚜 — 𝚗𝚘 𝚕𝚒𝚌𝚎𝚗𝚜𝚒𝚗𝚐 𝚏𝚎𝚎𝚜.

• 𝙲𝚛𝚎𝚊𝚝𝚎 𝚌𝚞𝚜𝚝𝚘𝚖 𝚓𝚒𝚗𝚐𝚕𝚎𝚜 𝚏𝚘𝚛 𝚙𝚛𝚘𝚍𝚞𝚌𝚝 𝚕𝚊𝚞𝚗𝚌𝚑𝚎𝚜, 𝚎𝚟𝚎𝚗𝚝𝚜, 𝚊𝚗𝚍 𝚜𝚘𝚌𝚒𝚊𝚕 𝚌𝚘𝚗𝚝𝚎𝚗𝚝 𝚒𝚗 𝚜𝚎𝚌𝚘𝚗𝚍𝚜.

• 𝙿𝚑𝚘𝚝𝚘 𝚝𝚘 𝚖𝚞𝚜𝚒𝚌: 𝚃𝚞𝚛𝚗 𝚊 𝚙𝚑𝚘𝚝𝚘 𝚘𝚛 𝚌𝚊𝚖𝚙𝚊𝚒𝚐𝚗 𝚒𝚖𝚊𝚐𝚎 𝚍𝚒𝚛𝚎𝚌𝚝𝚕𝚢 𝚒𝚗𝚝𝚘 𝚊 𝚖𝚘𝚘𝚍-𝚖𝚊𝚝𝚌𝚑𝚎𝚍 𝚊𝚞𝚍𝚒𝚘 𝚝𝚛𝚊𝚌𝚔.

𝙴𝚡𝚊𝚖𝚙𝚕𝚎: 𝙷𝚎𝚛𝚎’𝚜 𝚑𝚘𝚠 𝚝𝚘 𝚞𝚜𝚎 𝙻𝚢𝚛𝚒𝚊 𝟹 𝚊𝚌𝚛𝚘𝚜𝚜 𝚝𝚑𝚛𝚎𝚎 𝚛𝚎𝚊𝚕 𝚖𝚘𝚖𝚎𝚗𝚝𝚜:

1. 𝚃𝚑𝚎 𝚝𝚛𝚊𝚒𝚗𝚒𝚗𝚐 𝚟𝚒𝚍𝚎𝚘 𝚝𝚑𝚊𝚝 𝚏𝚒𝚗𝚊𝚕𝚕𝚢 𝚜𝚘𝚞𝚗𝚍𝚜 𝚒𝚗𝚝𝚎𝚗𝚝𝚒𝚘𝚗𝚊𝚕: 𝚃𝚑𝚎 𝚝𝚎𝚊𝚖 𝚘𝚙𝚎𝚗𝚜 𝙶𝚎𝚖𝚒𝚗𝚒, 𝚜𝚎𝚕𝚎𝚌𝚝𝚜 m𝚞𝚜𝚒𝚌, 𝚊𝚗𝚍 𝚝𝚢𝚙𝚎𝚜: “𝙿𝚛𝚘𝚏𝚎𝚜𝚜𝚒𝚘𝚗𝚊𝚕, 𝚏𝚘𝚌𝚞𝚜𝚎𝚍 𝚋𝚊𝚌𝚔𝚐𝚛𝚘𝚞𝚗𝚍 𝚖𝚞𝚜𝚒𝚌 𝚏𝚘𝚛 𝚊 𝚌𝚘𝚖𝚙𝚕𝚒𝚊𝚗𝚌𝚎 𝚝𝚛𝚊𝚒𝚗𝚒𝚗𝚐 𝚟𝚒𝚍𝚎𝚘. 𝙽𝚎𝚞𝚝𝚛𝚊𝚕, 𝚖𝚘𝚍𝚎𝚛𝚗, 𝚜𝚕𝚒𝚐𝚑𝚝𝚕𝚢 𝚞𝚙𝚋𝚎𝚊𝚝. 𝙸𝚗𝚜𝚝𝚛𝚞𝚖𝚎𝚗𝚝𝚊𝚕 𝚘𝚗𝚕𝚢, 𝚗𝚘 𝚟𝚘𝚌𝚊𝚕𝚜.” 𝙸𝚗 𝟹𝟶 𝚜𝚎𝚌𝚘𝚗𝚍𝚜 𝚝𝚑𝚎𝚢 𝚑𝚊𝚟𝚎 𝚊 𝚞𝚗𝚒𝚚𝚞𝚎 𝚝𝚛𝚊𝚌𝚔.

𝚃𝚑𝚎𝚗 𝚏𝚘𝚕𝚕𝚘𝚠 𝚞𝚙: “𝙼𝚊𝚔𝚎 𝚒𝚝 𝚜𝚕𝚒𝚐𝚑𝚝𝚕𝚢 𝚠𝚊𝚛𝚖𝚎𝚛.” 𝙾𝚗𝚎 𝚖𝚘𝚛𝚎 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚒𝚘𝚗. 𝙳𝚘𝚗𝚎．

𝟸. 𝚃𝚑𝚎 𝚝𝚘𝚠𝚗 𝚑𝚊𝚕𝚕 𝚝𝚑𝚊𝚝 𝚗𝚎𝚎𝚍𝚎𝚍 𝚊𝚗 𝚎𝚗𝚝𝚛𝚊𝚗𝚌𝚎: 𝚃𝚑𝚎 𝙲𝙴𝙾’𝚜 𝚚𝚞𝚊𝚛𝚝𝚎𝚛𝚕𝚢 𝚊𝚕𝚕-𝚑𝚊𝚗𝚍𝚜 𝚑𝚊𝚜 𝚊 𝚜𝚕𝚒𝚍𝚎 𝚍𝚎𝚌𝚔, 𝚊 𝚜𝚝𝚛𝚘𝚗𝚐 𝚊𝚐𝚎𝚗𝚍𝚊, 𝚊𝚗𝚍 𝚣𝚎𝚛𝚘 𝚊𝚞𝚍𝚒𝚘 𝚒𝚍𝚎𝚗𝚝𝚒𝚝𝚢. 𝙿𝚛𝚘𝚖𝚙𝚝: “𝙰𝚗 𝚎𝚗𝚎𝚛𝚐𝚒𝚜𝚒𝚗𝚐, 𝚌𝚒𝚗𝚎𝚖𝚊𝚝𝚒𝚌 𝟹𝟶-𝚜𝚎𝚌𝚘𝚗𝚍 𝚘𝚙𝚎𝚗𝚎𝚛 𝚏𝚘𝚛 𝚊 𝚌𝚘𝚖𝚙𝚊𝚗𝚢 𝚝𝚘𝚠𝚗 𝚑𝚊𝚕𝚕. 𝙼𝚘𝚍𝚎𝚛𝚗, 𝚌𝚘𝚗𝚏𝚒𝚍𝚎𝚗𝚝, 𝚋𝚞𝚒𝚕𝚍𝚜 𝚝𝚘 𝚊 𝚌𝚛𝚎𝚜𝚌𝚎𝚗𝚍𝚘. 𝙽𝚘 𝚕𝚢𝚛𝚒𝚌𝚜”. 𝚃𝚑𝚎 𝚝𝚛𝚊𝚌𝚔 𝚙𝚕𝚊𝚢𝚜 𝚊𝚜 𝚎𝚖𝚙𝚕𝚘𝚢𝚎𝚎𝚜 𝚓𝚘𝚒𝚗 𝚝𝚑𝚎 𝚌𝚊𝚕𝚕．

𝟹. 𝚃𝚑𝚎 𝚙𝚛𝚘𝚍𝚞𝚌𝚝 𝚕𝚊𝚞𝚗𝚌𝚑 𝚛𝚎𝚎𝚕 𝚝𝚑𝚊𝚝 𝚌𝚊𝚖𝚎 𝚝𝚘𝚐𝚎𝚝𝚑𝚎𝚛 𝚕𝚊𝚜𝚝 𝚖𝚒𝚗𝚞𝚝𝚎: 𝙼𝚊𝚛𝚔𝚎𝚝𝚒𝚗𝚐 𝚑𝚊𝚜 𝟺𝟾 𝚑𝚘𝚞𝚛𝚜 𝚋𝚎𝚏𝚘𝚛𝚎 𝚊 𝚙𝚛𝚘𝚍𝚞𝚌𝚝 𝚐𝚘𝚎𝚜 𝚕𝚒𝚟𝚎. 𝚃𝚑𝚎 𝚌𝚊𝚖𝚙𝚊𝚒𝚐𝚗 𝚟𝚒𝚍𝚎𝚘 𝚒𝚜 𝚛𝚎𝚊𝚍𝚢 𝚋𝚞𝚝 𝚝𝚑𝚎 𝚕𝚒𝚌𝚎𝚗𝚜𝚎𝚍 𝚝𝚛𝚊𝚌𝚔 𝚝𝚑𝚎𝚢 𝚠𝚊𝚗𝚝𝚎𝚍 𝚌𝚘𝚜𝚝𝚜 $𝟾𝟶𝟶 𝚊𝚗𝚍 𝚝𝚊𝚔𝚎𝚜 𝚝𝚑𝚛𝚎𝚎 𝚍𝚊𝚢𝚜 𝚝𝚘 𝚌𝚕𝚎𝚊𝚛. 𝙸𝚗𝚜𝚝𝚎𝚊𝚍, 𝚝𝚑𝚎 𝚖𝚊𝚛𝚔𝚎𝚝𝚒𝚗𝚐 𝚖𝚊𝚗𝚊𝚐𝚎𝚛 𝚞𝚙𝚕𝚘𝚊𝚍𝚜 𝚝𝚑𝚎 𝚑𝚎𝚛𝚘 𝚌𝚊𝚖𝚙𝚊𝚒𝚐𝚗 𝚒𝚖𝚊𝚐𝚎 𝚝𝚘 𝙶𝚎𝚖𝚒𝚗𝚒 𝚊𝚗𝚍 𝚙𝚛𝚘𝚖𝚙𝚝𝚜: “𝙲𝚛𝚎𝚊𝚝𝚎 𝚊 𝚝𝚛𝚊𝚌𝚔 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚝𝚑𝚒𝚜 𝚒𝚖𝚊𝚐𝚎. 𝙱𝚘𝚕𝚍, 𝚊𝚜𝚙𝚒𝚛𝚊𝚝𝚒𝚘𝚗𝚊𝚕, 𝚖𝚒𝚕𝚕𝚎𝚗𝚗𝚒𝚊𝚕 𝚎𝚗𝚎𝚛𝚐𝚢, 𝚜𝚝𝚛𝚘𝚗𝚐 𝚏𝚎𝚖𝚊𝚕𝚎 𝚟𝚘𝚌𝚊𝚕𝚜.”

𝙻𝚢𝚛𝚒𝚊 𝟹 𝚛𝚎𝚊𝚍𝚜 𝚝𝚑𝚎 𝚟𝚒𝚜𝚞𝚊𝚕- 𝚋𝚛𝚒𝚐𝚑𝚝 𝚌𝚘𝚕𝚘𝚞𝚛𝚜, 𝚏𝚘𝚛𝚠𝚊𝚛𝚍 𝚖𝚘𝚝𝚒𝚘𝚗 𝚊𝚗𝚍 𝚍𝚎𝚕𝚒𝚟𝚎𝚛𝚜 𝚊 𝚝𝚛𝚊𝚌𝚔 𝚝𝚑𝚊𝚝 𝚏𝚎𝚎𝚕𝚜 𝚖𝚊𝚍𝚎 𝚏𝚘𝚛 𝚝𝚑𝚎 𝚟𝚒𝚍𝚎𝚘. 𝙱𝚎𝚌𝚊𝚞𝚜𝚎 𝚒𝚝 𝚠𝚊𝚜．

𝚆𝚑𝚊𝚝 𝚖𝚊𝚔𝚎𝚜 𝙻𝚢𝚛𝚒𝚊 𝟹 𝚜𝚙𝚎𝚌𝚒𝚊𝚕?

𝙸𝚝 𝚕𝚒𝚟𝚎𝚜 𝚒𝚗𝚜𝚒𝚍𝚎 𝚢𝚘𝚞𝚛 𝚎𝚡𝚒𝚜𝚝𝚒𝚗𝚐 𝚠𝚘𝚛𝚔𝚏𝚕𝚘𝚠, 𝚗𝚘𝚝 𝚘𝚞𝚝𝚜𝚒𝚍𝚎 𝚒𝚝: 𝚄𝚗𝚕𝚒𝚔𝚎 𝚂𝚞𝚗𝚘, 𝚄𝚍𝚒𝚘, 𝚘𝚛 𝚊𝚗𝚢 𝚘𝚝𝚑𝚎𝚛 𝚜𝚝𝚊𝚗𝚍𝚊𝚕𝚘𝚗𝚎 𝚖𝚞𝚜𝚒𝚌 𝚝𝚘𝚘𝚕, 𝙻𝚢𝚛𝚒𝚊 𝟹 𝚒𝚜 𝚋𝚞𝚒𝚕𝚝 𝚍𝚒𝚛𝚎𝚌𝚝𝚕𝚢 𝚒𝚗𝚝𝚘 𝙶𝚎𝚖𝚒𝚗𝚒, 𝚝𝚑𝚎 𝚜𝚊𝚖𝚎 𝚠𝚒𝚗𝚍𝚘𝚠 𝚠𝚑𝚎𝚛𝚎 𝚢𝚘𝚞’𝚛𝚎 𝚊𝚕𝚛𝚎𝚊𝚍𝚢 𝚠𝚛𝚒𝚝𝚒𝚗𝚐, 𝚛𝚎𝚜𝚎𝚊𝚛𝚌𝚑𝚒𝚗𝚐, 𝚊𝚗𝚍 𝚌𝚛𝚎𝚊𝚝𝚒𝚗𝚐．

• 𝙽𝚘 m𝚞𝚜𝚒𝚌 k𝚗𝚘𝚠𝚕𝚎𝚍𝚐𝚎 r𝚎𝚚𝚞𝚒𝚛𝚎𝚍: 𝙶𝚎𝚗𝚛𝚎, 𝚝𝚎𝚖𝚙𝚘, 𝚟𝚘𝚌𝚊𝚕 𝚜𝚝𝚢𝚕𝚎 — 𝚊𝚕𝚕 𝚍𝚒𝚛𝚎𝚌𝚝𝚎𝚍 𝚒𝚗 𝚙𝚕𝚊𝚒𝚗 𝙴𝚗𝚐𝚕𝚒𝚜𝚑, 𝚝𝚑𝚎 𝚠𝚊𝚢 𝚢𝚘𝚞’𝚍 𝚋𝚛𝚒𝚎𝚏 𝚊 𝚌𝚘𝚕𝚕𝚎𝚊𝚐𝚞𝚎.

• 𝙱𝚞𝚒𝚕𝚝-𝚒𝚗 𝙰𝙸 t𝚛𝚊𝚗𝚜𝚙𝚊𝚛𝚎𝚗𝚌𝚢: 𝙴𝚟𝚎𝚛𝚢 𝚝𝚛𝚊𝚌𝚔 𝚒𝚜 𝚊𝚞𝚝𝚘𝚖𝚊𝚝𝚒𝚌𝚊𝚕𝚕𝚢 𝚠𝚊𝚝𝚎𝚛𝚖𝚊𝚛𝚔𝚎𝚍 𝚠𝚒𝚝𝚑 𝚂𝚢𝚗𝚝𝚑𝙸𝙳- 𝙶𝚘𝚘𝚐𝚕𝚎’𝚜 𝚒𝚗𝚟𝚒𝚜𝚒𝚋𝚕𝚎 𝚊𝚞𝚍𝚒𝚘 𝚏𝚒𝚗𝚐𝚎𝚛𝚙𝚛𝚒𝚗𝚝, 𝚜𝚘 𝙰𝙸-𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎𝚍 𝚌𝚘𝚗𝚝𝚎𝚗𝚝 𝚒𝚜 𝚊𝚕𝚠𝚊𝚢𝚜 𝚒𝚍𝚎𝚗𝚝𝚒𝚏𝚒𝚊𝚋𝚕𝚎, 𝚎𝚟𝚎𝚗 𝚊𝚏𝚝𝚎𝚛 𝚌𝚘𝚖𝚙𝚛𝚎𝚜𝚜𝚒𝚘𝚗 𝚘𝚛 𝚛𝚎-𝚛𝚎𝚌𝚘𝚛𝚍𝚒𝚗𝚐．

𝙽𝚘𝚝𝚎: 𝚃𝚑𝚎 𝚝𝚘𝚘𝚕𝚜 𝚊𝚗𝚍 𝚊𝚗𝚊𝚕𝚢𝚜𝚒𝚜 𝚏𝚎𝚊𝚝𝚞𝚛𝚎𝚍 𝚒𝚗 𝚝𝚑𝚒𝚜 𝚜𝚎𝚌𝚝𝚒𝚘𝚗 𝚍𝚎𝚖𝚘𝚗𝚜𝚝𝚛𝚊𝚝𝚎𝚍 𝚌𝚕𝚎𝚊𝚛 𝚟𝚊𝚕𝚞𝚎 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚘𝚞𝚛 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚝𝚎𝚜𝚝𝚒𝚗𝚐. 𝙾𝚞𝚛 𝚛𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗𝚜 𝚊𝚛𝚎 𝚎𝚗𝚝𝚒𝚛𝚎𝚕𝚢 𝚒𝚗𝚍𝚎𝚙𝚎𝚗𝚍𝚎𝚗𝚝 𝚊𝚗𝚍 𝚗𝚘𝚝 𝚒𝚗𝚏𝚕𝚞𝚎𝚗𝚌𝚎𝚍 𝚋𝚢 𝚝𝚑𝚎 𝚝𝚘𝚘𝚕 𝚌𝚛𝚎𝚊𝚝𝚘𝚛𝚜.

AI BITS & BYTES

Why slower adoption lowers displacement risk

Countering last week’s Citrini Research post which rattled IT stocks specifically, a Citadel Securities note argues that technological diffusion has historically followed an S-curve. Early adoption is slow and expensive. Growth accelerates as costs fall, and complementary infrastructure develops. Eventually, saturation sets in, and marginal adopter is less productive or less profitable which causes growth to decelerate.

Despite this, markets often extrapolate the acceleration phase linearly but history implies pace of adoption plateaus as organisational integration is costly, regulation emerges and diminishing marginal returns exist in economic deployment. The risk of displacement declines with a slower pace of adoption.

Did Grok predict the US attack on Iran

Elon Musk’s chatbot uniquely predicted the date of the US strike on Iran as 28 February, even prior to the attack. The prediction was made by the AI chatbot in a methodological test on 25 February.

Iran war puts spotlight on Big Tech’s AI bets in West Asia

Countries in West Asia are investing heavily in AI, semiconductor partnerships and cloud computing as part of broader transformation plans designed to attract foreign capital and build domestic technology ecosystems. Here are some of the Big Tech AI plans in the region.

Elon Musk’s X cracks down on AI deepfakes on Iran war

X Product Head Nikita Bier said that the company is trying to maintain access to authentic information on the ground in times of war. This has led X to now warn creators that they will be suspended from the revenue-sharing for 90 days if they post AI-generated videos of an armed conflict without adding a disclosure.

India adopting pragmatic approach to regulate AI: IAIRO

India has adopted a balanced and pragmatic approach on AI, avoiding the extremes of lightly regulated US model and heavily compliance-driven framework of the EU, said an Indian AI Research Organisation official. IAIRO Founding Director Amit Sheth said India has adopted a balanced regulatory approach which enables innovation while safeguarding users.

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

Sample these developments. When Anthropic (Claude Opus 4.6), OpenAI (GPT-5.3-Codex), and Google (Gemini 3.1 Pro) unveiled more capable Agentic AI systems—tools that can autonomously analyse, refactor and execute complex enterprise work flows—investors reacted sharply. And it’s most likely that software and IT stocks with heavy exposure to billable-hour models will continue to see volatility.

The tremors reflect a broader fear that Agentic AI systems can independently plan and execute multi-step coding or operational functions may upend valuation models that assume labour-intensive services.

In this week’s edition of Tech Talk:

• How Anthropic’s AI agents continue to unsettle IT stocks

• Why Google Gemini may steal the AI show altogether

• AI Tool of the Week: Claude AI in Powerpoint

• AI Bits & Bytes: Skills, not titles to bring in the monies

PREVIOUSLY, ON TECH TALK

Nasscom, has consistently argued that AI is augmenting work rather than eliminating it—causing near-term pain, slower hiring, and margin pressure—but leaving the long-term arc of digital transformation intact.

The software lobby projects that India’s IT services industry will grow at 6.1% year-on-year to $315 billion in FY26. There will more jobs too—up 2.3% to 135,000 on a net basis—driven by AI-led IT services and the expanding footprint of India’s roughly 1,700 global capability centres (GCCs).

Are these growth numbers realistic? Big Tech’s aggressive push into Agentic AI, capable of automating complex, multi-step tasks, hangs like the Sword of Damocles over traditional IT services.

• Citrini Research’s latest Substack post rattled markets by highlighting how Generative AI and Agentic AI could pose a structural threat to the offshore labour-arbitrage model over the next two to three years.

• Venture capitalist Vinod Khosla has argued that large portions of traditional IT services could be automated away.

• Investor notes now refer to the disruption of software-as-a-service (SaaS) companies as a “SaaSpocalypse”, where AI-native platforms compress both software economics and the consulting layers that support them.

And, Anthropic co-founder and CEO Dario Amodei has repeatedly warned that Advanced AI could sharply reduce demand for entry-level coders. He has to be taken seriously. Just this month, Anthropic roiled markets: first with new automation plug-ins for handling legal, sales and analytics tasks; then with a code-security tool that identified hundreds of long-undetected vulnerabilities; and, most recently, with claims its models can streamline COBOL code running on legacy systems.

• Anthropic’s first blow cost the markets almost $285 billion.

• Next, cybersecurity stocks such as CrowdStrike and Okta slid sharply.

• On Monday, IBM suffered its steepest one-day fall in over a decade after reports suggested that AI could modernise COBOL-heavy environments—systems deeply embedded in banking and government. In India, Tata Consultancy Services Ltd. and Infosys Ltd. employ thousands tied to mainframe and legacy maintenance.

That said, the truth lies somewhere in the middle. Between apocalyptic forecasts and industry defensiveness, AI will redraw the economics of IT services. But it won’t do so overnight, and certainly not without friction.

MORE ON ANTHROPIC

• New Anthropic AI tool wipes billions off cybersecurity stocks

• Anthropic says China’s DeepSeek, others misused Claude AI

• Anthropic boss says AI can surpass humans at ‘almost everything’

Why Google Gemini may steal the AI show

Even as investors despair, the partnership between Microsoft Corp. and OpenAI is entering a more complicated phase. OpenAI is expanding enterprise relationships and signalling infrastructure diversification, while Microsoft has also struck a strategic partnership with Anthropic (OpenAI’s rival) even as it continues embedding GPT models deep into Azure and its productivity stack.

That turbulence is already opening up space for Google, which holds roughly a 14% stake in Anthropic, and whose valuation has exceeded $60 billion in recent rounds. It gives Alphabet Inc. financial exposure to the very company unsettling stock-market assumptions. Through DeepMind and its Gemini models, Google competes directly at the frontier. This makes its bet on Anthropic both a hedge and rival.

When we wrote this newsletter, we reported that Pentagon queered the pitch for Anthropic when it threatened to designate the Claude maker as a “supply chain risk” due to the latter’s reluctance to allow use of its models without guardrails.

Anthropic reportedly succumbed to Pentagon’s pressure by weakening its foundational safety principle. “We are releasing the third version of our Responsible Scaling Policy (RSP), the voluntary framework we use to mitigate catastrophic risks from AI systems,” the company said in a statement to Business Insider on its revised security policy. Anthropic says that it was “convinced” that “effective government engagement on AI safety is both necessary and achievable”.

That, of course, raised serious concerns about AI being misused for mass surveillance or to power autonomous weapons, regardless of government assurances. From a business perspective too, a hit to Anthropic’s valuation would affect the paper value of Google’s stake.

However, late Thursday, Anthropic categorically rejected the Pentagon’s ultimatum over military use of its flagship product, Claude AI, despite threats of blacklisting. “These threats do not change our position: we cannot in good conscience accede to their request,” Anthropic CEO Dario Amodei said in a statement.

“...We have never raised objections to particular military operations nor attempted to limit use of our technology in an ad hoc manner. However, in a narrow set of cases, we believe AI can undermine, rather than defend, democratic values. Some uses are also simply outside the bounds of what today’s technology can safely and reliably do. Two such use cases have never been included in our contracts with the Department of War, and we believe they should not be included now...”

The two cases are: “Mass domestic surveillance”, and “Fully autonomous weapons”.

Amodei concluded, “It is the Department’s prerogative to select contractors most aligned with their vision. But given the substantial value that Anthropic’s technology provides to our armed forces, we hope they reconsider....”

Implications The confrontation effectively jeopardises Anthropic’s long-standing defence contracts with the government. If the Pentagon designates Anthropic as a “supply chain risk”, the implications could also extend beyond defence contracts. Such a label could limit access to sensitive federal workloads, influence procurement decisions across agencies, and prompt enterprise CIOs to reassess risk exposure. Markets could likely respond swiftly, compressing Anthropic’s valuation and injecting fresh volatility into AI-linked equities.

For Google, the impact would be mixed. A hit to Anthropic’s valuation would affect the paper value of its minority stake. Yet Google Cloud, now a multi-billion-dollar business with improving margins, could benefit if federal or enterprise clients diversify toward multi-cloud strategies rather than deepen dependence on a single Microsoft-OpenAI stack. In an era when AI workloads require hyperscale compute, owning the infras layer may prove more defensible than owning a single model.

Microsoft, conversely, could see near-term gains in government AI contracts if a competitor faces restrictions. But broader national security scrutiny of frontier AI supply chains could also expand regulatory oversight across the sector, tightening compliance demands for all major players.

We will continue to watch this space closely.

AI TOOL OF THE WEEK

by AI&Beyond, with Jaspreet Bindra and Anuj Magazine

𝚃𝚑𝚎 𝙰𝙸 𝚌𝚊𝚙𝚊𝚋𝚒𝚕𝚒𝚝𝚢 𝚠𝚎 𝚏𝚎𝚊𝚝𝚞𝚛𝚎 𝚝𝚘𝚍𝚊𝚢 𝚒𝚜 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝙲𝚕𝚊𝚞𝚍𝚎 𝚒𝚗 𝙿𝚘𝚠𝚎𝚛𝚙𝚘𝚒𝚗𝚝．

𝚆𝚑𝚊𝚝 𝚙𝚛𝚘𝚋𝚕𝚎𝚖 𝚍𝚘𝚎𝚜 𝚒𝚝 𝚜𝚘𝚕𝚟𝚎？𝙰𝚜𝚔 𝚊𝚗𝚢 𝚌𝚘𝚗𝚜𝚞𝚕𝚝𝚊𝚗𝚝, 𝚊𝚗𝚊𝚕𝚢𝚜𝚝, 𝚘𝚛 𝚋𝚞𝚜𝚒𝚗𝚎𝚜𝚜 𝚕𝚎𝚊𝚍𝚎𝚛 𝚠𝚑𝚊𝚝 𝚎𝚊𝚝𝚜 𝚝𝚑𝚎𝚒𝚛 𝚝𝚒𝚖𝚎, 𝚊𝚗𝚍 𝚢𝚘𝚞 𝚠𝚒𝚕𝚕 𝚑𝚎𝚊𝚛 𝚝𝚑𝚎 𝚜𝚊𝚖𝚎 𝚊𝚗𝚜𝚠𝚎𝚛: 𝚋𝚞𝚒𝚕𝚍𝚒𝚗𝚐 𝚍𝚎𝚌𝚔𝚜. 𝙽𝚘𝚝 𝚝𝚑𝚒𝚗𝚔𝚒𝚗𝚐, 𝚋𝚞𝚒𝚕𝚍𝚒𝚗𝚐. 𝚂𝚕𝚒𝚍𝚎 𝚜𝚝𝚛𝚞𝚌𝚝𝚞𝚛𝚎, 𝚏𝚘𝚗𝚝 𝚌𝚘𝚖𝚙𝚕𝚒𝚊𝚗𝚌𝚎, 𝚌𝚘𝚗𝚟𝚎𝚛𝚝𝚒𝚗𝚐 𝚏𝚒𝚟𝚎-𝚋𝚞𝚕𝚕𝚎𝚝 𝚙𝚘𝚒𝚗𝚝𝚜 𝚒𝚗𝚝𝚘 𝚊 𝚟𝚒𝚜𝚞𝚊𝚕, 𝚖𝚊𝚝𝚌𝚑𝚒𝚗𝚐 𝚝𝚑𝚎 𝚌𝚘𝚛𝚙𝚘𝚛𝚊𝚝𝚎 𝚝𝚎𝚖𝚙𝚕𝚊𝚝𝚎 𝚊𝚝 𝟷𝟷 𝙿𝙼 𝚋𝚎𝚏𝚘𝚛𝚎 𝚊 𝟿 𝙰𝙼 𝚋𝚘𝚊𝚛𝚍 𝚖𝚎𝚎𝚝𝚒𝚗𝚐. 𝚃𝚑𝚎 𝚝𝚑𝚒𝚗𝚔𝚒𝚗𝚐 𝚝𝚊𝚔𝚎𝚜 𝟹𝟶 𝚖𝚒𝚗𝚞𝚝𝚎𝚜; 𝚝𝚑𝚎 𝚏𝚘𝚛𝚖𝚊𝚝𝚝𝚒𝚗𝚐 𝚝𝚊𝚔𝚎𝚜 𝟹 𝚑𝚘𝚞𝚛𝚜． 𝙰𝚗𝚍 𝚝𝚑𝚎 𝚠𝚘𝚛𝚜𝚝 𝚙𝚊𝚛𝚝? 𝙴𝚟𝚎𝚛𝚢 𝚗𝚎𝚠 𝚎𝚗𝚐𝚊𝚐𝚎𝚖𝚎𝚗𝚝 𝚜𝚝𝚊𝚛𝚝𝚜 𝚏𝚛𝚘𝚖 𝚊 𝚋𝚕𝚊𝚗𝚔 𝚜𝚕𝚒𝚍𝚎．

𝙲𝚕𝚊𝚞𝚍𝚎 𝚒𝚗 𝙿𝚘𝚠𝚎𝚛𝙿𝚘𝚒𝚗𝚝 𝚑𝚎𝚕𝚙𝚜 𝚢𝚘𝚞 𝚜𝚘𝚕𝚟𝚎 𝚝𝚑𝚒𝚜. 𝙸𝚝 𝚜𝚒𝚝𝚜 𝚒𝚗𝚜𝚒𝚍𝚎 𝚢𝚘𝚞𝚛 𝙿𝚘𝚠𝚎𝚛𝙿𝚘𝚒𝚗𝚝 𝚜𝚒𝚍𝚎𝚋𝚊𝚛 𝚊𝚗𝚍 𝚊𝚌𝚝𝚞𝚊𝚕𝚕𝚢 𝚛𝚎𝚊𝚍𝚜 𝚢𝚘𝚞𝚛 𝚜𝚕𝚒𝚍𝚎 𝚖𝚊𝚜𝚝𝚎𝚛—𝚏𝚘𝚗𝚝𝚜, 𝚕𝚊𝚢𝚘𝚞𝚝𝚜, 𝚌𝚘𝚕𝚘𝚞𝚛𝚜—𝚋𝚎𝚏𝚘𝚛𝚎 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚒𝚗𝚐 𝚊𝚗𝚢𝚝𝚑𝚒𝚗𝚐．

𝙷𝚘𝚠 𝚝𝚘 𝚊𝚌𝚌𝚎𝚜𝚜: 𝙸𝚗𝚜𝚝𝚊𝚕𝚕 “𝙲𝚕𝚊𝚞𝚍𝚎 𝚋𝚢 𝙰𝚗𝚝𝚑𝚛𝚘𝚙𝚒𝚌 𝚒𝚗 𝙿𝚘𝚠𝚎𝚛𝙿𝚘𝚒𝚗𝚝” 𝚏𝚛𝚘𝚖 𝙼𝚒𝚌𝚛𝚘𝚜𝚘𝚏𝚝 𝙰𝚙𝚙𝚂𝚘𝚞𝚛𝚌𝚎 𝙼𝚊𝚛𝚔𝚎𝚝𝚙𝚕𝚊𝚌𝚎. 𝙰𝚟𝚊𝚒𝚕𝚊𝚋𝚕𝚎 𝚘𝚗 𝙿𝚛𝚘, 𝙼𝚊𝚡, 𝚃𝚎𝚊𝚖, 𝚊𝚗𝚍 𝙴𝚗𝚝𝚎𝚛𝚙𝚛𝚒𝚜𝚎 𝚙𝚕𝚊𝚗𝚜．

𝙲𝚕𝚊𝚞𝚍𝚎 𝚒𝚗 𝙿𝚘𝚠𝚎𝚛𝙿𝚘𝚒𝚗𝚝 𝚌𝚊𝚗 𝚑𝚎𝚕𝚙 𝚢𝚘𝚞:

• 𝙶𝚎𝚗𝚎𝚛𝚊𝚝𝚎 𝚌𝚘𝚖𝚙𝚕𝚎𝚝𝚎 𝚜𝚕𝚒𝚍𝚎𝚜 𝚏𝚛𝚘𝚖 𝚊 𝚗𝚊𝚝𝚞𝚛𝚊𝚕 𝚕𝚊𝚗𝚐𝚞𝚊𝚐𝚎 𝚍𝚎𝚜𝚌𝚛𝚒𝚙𝚝𝚒𝚘𝚗, 𝚞𝚜𝚒𝚗𝚐 𝚢𝚘𝚞𝚛 𝚎𝚡𝚒𝚜𝚝𝚒𝚗𝚐 𝚝𝚎𝚖𝚙𝚕𝚊𝚝𝚎.

• 𝙲𝚘𝚗𝚟𝚎𝚛𝚝 𝚋𝚞𝚕𝚕𝚎𝚝-𝚑𝚎𝚊𝚟𝚢 𝚜𝚕𝚒𝚍𝚎𝚜 𝚒𝚗𝚝𝚘 𝚎𝚍𝚒𝚝𝚊𝚋𝚕𝚎 𝚍𝚒𝚊𝚐𝚛𝚊𝚖𝚜, 𝚙𝚛𝚘𝚌𝚎𝚜𝚜 𝚏𝚕𝚘𝚠𝚜, 𝚘𝚛 𝚌𝚑𝚊𝚛𝚝𝚜 - 𝚗𝚘𝚝 𝚜𝚝𝚊𝚝𝚒𝚌 𝚒𝚖𝚊𝚐𝚎𝚜.

• 𝚁𝚎𝚠𝚛𝚒𝚝𝚎 𝚊𝚗𝚍 𝚛𝚎𝚜𝚝𝚛𝚞𝚌𝚝𝚞𝚛𝚎 𝚊 𝚍𝚎𝚌𝚔’𝚜 𝚜𝚝𝚘𝚛𝚢𝚕𝚒𝚗𝚎 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚝𝚘𝚞𝚌𝚑𝚒𝚗𝚐 𝚝𝚑𝚎 𝚏𝚘𝚛𝚖𝚊𝚝𝚝𝚒𝚗𝚐.

𝙴𝚡𝚊𝚖𝚙𝚕𝚎: 𝙰 𝚜𝚝𝚛𝚊𝚝𝚎𝚐𝚢 𝚝𝚎𝚊𝚖 𝚗𝚎𝚎𝚍𝚜 𝚝𝚘 𝚙𝚛𝚎𝚜𝚎𝚗𝚝 𝚊 𝚖𝚊𝚛𝚔𝚎𝚝 𝚎𝚗𝚝𝚛𝚢 𝚙𝚛𝚘𝚙𝚘𝚜𝚊𝚕 𝚝𝚘 𝚝𝚑𝚎 𝚕𝚎𝚊𝚍𝚎𝚛𝚜𝚑𝚒𝚙 𝚋𝚢 𝚖𝚘𝚛𝚗𝚒𝚗𝚐. 𝙷𝚎𝚛𝚎’𝚜 𝚑𝚘𝚠 𝙲𝚕𝚊𝚞𝚍𝚎 𝚒𝚗 𝙿𝚘𝚠𝚎𝚛𝙿𝚘𝚒𝚗𝚝 𝚖𝚘𝚟𝚎𝚜 𝚝𝚑𝚎𝚖 𝚏𝚛𝚘𝚖 𝚋𝚕𝚊𝚗𝚔 𝚙𝚊𝚐𝚎 𝚝𝚘 𝚋𝚘𝚊𝚛𝚍𝚛𝚘𝚘𝚖-𝚛𝚎𝚊𝚍𝚢:

• 𝙱𝚕𝚊𝚗𝚔 𝚙𝚊𝚐𝚎 → 𝙵𝚒𝚛𝚜𝚝 𝚍𝚛𝚊𝚏𝚝: 𝚃𝚑𝚎𝚢 𝚕𝚘𝚊𝚍 𝚝𝚑𝚎𝚒𝚛 𝚌𝚘𝚛𝚙𝚘𝚛𝚊𝚝𝚎 𝚝𝚎𝚖𝚙𝚕𝚊𝚝𝚎, 𝚘𝚙𝚎𝚗 𝚝𝚑𝚎 𝙲𝚕𝚊𝚞𝚍𝚎 𝚜𝚒𝚍𝚎𝚋𝚊𝚛, 𝚊𝚗𝚍 𝚝𝚢𝚙𝚎: “𝙱𝚞𝚒𝚕𝚍 𝚊𝚗 𝟾-𝚜𝚕𝚒𝚍𝚎 𝚖𝚊𝚛𝚔𝚎𝚝 𝚎𝚗𝚝𝚛𝚢 𝚙𝚛𝚘𝚙𝚘𝚜𝚊𝚕 𝚏𝚘𝚛 𝚝𝚑𝚎 𝙸𝚗𝚍𝚒𝚊𝚗 𝚕𝚘𝚐𝚒𝚜𝚝𝚒𝚌𝚜 𝚜𝚎𝚌𝚝𝚘𝚛—𝚌𝚘𝚟𝚎𝚛𝚒𝚗𝚐 𝚖𝚊𝚛𝚔𝚎𝚝 𝚜𝚒𝚣𝚎, 𝚌𝚘𝚖𝚙𝚎𝚝𝚒𝚝𝚒𝚟𝚎 𝚕𝚊𝚗𝚍𝚜𝚌𝚊𝚙𝚎, 𝚎𝚗𝚝𝚛𝚢 𝚜𝚝𝚛𝚊𝚝𝚎𝚐𝚢, 𝚛𝚒𝚜𝚔𝚜, 𝚊𝚗𝚍 𝚛𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗.” 𝙸𝚗 𝚖𝚒𝚗𝚞𝚝𝚎𝚜, 𝙲𝚕𝚊𝚞𝚍𝚎 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎𝚜 𝚊𝚕𝚕 𝚎𝚒𝚐𝚑𝚝 𝚜𝚕𝚒𝚍𝚎𝚜 𝚒𝚗 𝚝𝚑𝚎 𝚌𝚘𝚛𝚛𝚎𝚌𝚝 𝚏𝚘𝚗𝚝𝚜, 𝚌𝚘𝚕𝚘𝚞𝚛𝚜, 𝚊𝚗𝚍 𝚕𝚊𝚢𝚘𝚞𝚝𝚜. 𝚃𝚑𝚎 𝚋𝚕𝚊𝚗𝚔 𝚙𝚊𝚐𝚎 𝚙𝚛𝚘𝚋𝚕𝚎𝚖 𝚒𝚜 𝚐𝚘𝚗𝚎.

• 𝙵𝚒𝚛𝚜𝚝 𝚍𝚛𝚊𝚏𝚝 → 𝙴𝚍𝚒𝚝𝚜: 𝚃𝚑𝚎 𝚝𝚎𝚊𝚖 𝚛𝚎𝚟𝚒𝚎𝚠𝚜 𝚊𝚗𝚍 𝚜𝚙𝚘𝚝𝚜 𝚝𝚑𝚊𝚝 𝚜𝚕𝚒𝚍𝚎 𝟻 𝚒𝚜 𝚝𝚘𝚘 𝚝𝚎𝚡𝚝-𝚑𝚎𝚊𝚟𝚢 𝚊𝚗𝚍 𝚜𝚕𝚒𝚍𝚎 𝟽 𝚗𝚎𝚎𝚍𝚜 𝚊 𝚟𝚒𝚜𝚞𝚊𝚕. 𝚃𝚑𝚎𝚢 𝚝𝚎𝚕𝚕 𝙲𝚕𝚊𝚞𝚍𝚎: “𝙲𝚘𝚗𝚟𝚎𝚛𝚝 𝚜𝚕𝚒𝚍𝚎 𝟻 𝚒𝚗𝚝𝚘 𝚊 𝟸𝚡𝟸 𝚘𝚙𝚙𝚘𝚛𝚝𝚞𝚗𝚒𝚝𝚢 𝚖𝚊𝚝𝚛𝚒𝚡” 𝚊𝚗𝚍 “𝙼𝚊𝚔𝚎 𝚜𝚕𝚒𝚍𝚎 𝟽 𝚊 𝚝𝚑𝚛𝚎𝚎-𝚜𝚝𝚎𝚙 𝚙𝚛𝚘𝚌𝚎𝚜𝚜 𝚏𝚕𝚘𝚠.” 𝙲𝚕𝚊𝚞𝚍𝚎 𝚛𝚎𝚋𝚞𝚒𝚕𝚍𝚜 𝚋𝚘𝚝𝚑 𝚜𝚕𝚒𝚍𝚎𝚜 𝚊𝚜 𝚎𝚍𝚒𝚝𝚊𝚋𝚕𝚎 𝚗𝚊𝚝𝚒𝚟𝚎 𝙿𝚘𝚠𝚎𝚛𝙿𝚘𝚒𝚗𝚝 𝚜𝚑𝚊𝚙𝚎𝚜—𝚗𝚘𝚝 𝚒𝚖𝚊𝚐𝚎𝚜—𝚠𝚑𝚒𝚕𝚎 𝚕𝚎𝚊𝚟𝚒𝚗𝚐 𝚝𝚑𝚎 𝚛𝚎𝚜𝚝 𝚘𝚏 𝚝𝚑𝚎 𝚍𝚎𝚌𝚔 𝚞𝚗𝚝𝚘𝚞𝚌𝚑𝚎𝚍．

• 𝙴𝚍𝚒𝚝𝚜 → 𝙵𝚒𝚗𝚊𝚕 𝚍𝚛𝚊𝚏𝚝: 𝚆𝚒𝚝𝚑 𝚝𝚑𝚎 𝚜𝚝𝚛𝚞𝚌𝚝𝚞𝚛𝚎 𝚜𝚘𝚕𝚒𝚍, 𝚝𝚑𝚎𝚢 𝚍𝚘 𝚊 𝚏𝚒𝚗𝚊𝚕 𝚙𝚊𝚜𝚜: “𝚃𝚒𝚐𝚑𝚝𝚎𝚗 𝚜𝚕𝚒𝚍𝚎 𝟹 𝚝𝚘 𝚘𝚗𝚎 𝚔𝚎𝚢 𝚖𝚎𝚜𝚜𝚊𝚐𝚎” 𝚊𝚗𝚍 “𝚁𝚎𝚠𝚛𝚒𝚝𝚎 𝚝𝚑𝚎 𝚛𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗 𝚜𝚕𝚒𝚍𝚎 𝚏𝚘𝚛 𝚊 𝙲𝙵𝙾 𝚊𝚞𝚍𝚒𝚎𝚗𝚌𝚎—𝚖𝚘𝚛𝚎 𝚍𝚒𝚛𝚎𝚌𝚝, 𝚗𝚞𝚖𝚋𝚎𝚛𝚜-𝚏𝚒𝚛𝚜𝚝.” 𝙲𝚕𝚊𝚞𝚍𝚎 𝚊𝚍𝚓𝚞𝚜𝚝𝚜 𝚕𝚊𝚗𝚐𝚞𝚊𝚐𝚎 𝚊𝚗𝚍 𝚕𝚊𝚢𝚘𝚞𝚝 𝚊𝚌𝚌𝚘𝚛𝚍𝚒𝚗𝚐𝚕𝚢.

𝚆𝚑𝚊𝚝 𝚖𝚊𝚔𝚎𝚜 𝙲𝚕𝚊𝚞𝚍𝚎 𝚒𝚗 𝙿𝚘𝚠𝚎𝚛𝙿𝚘𝚒𝚗𝚝 𝚜𝚙𝚎𝚌𝚒𝚊𝚕？

• 𝚃𝚎𝚖𝚙𝚕𝚊𝚝𝚎-𝚊𝚠𝚊𝚛𝚎 𝚋𝚢 𝚍𝚎𝚏𝚊𝚞𝚕𝚝: 𝙲𝚕𝚊𝚞𝚍𝚎 𝚛𝚎𝚊𝚍𝚜 𝚢𝚘𝚞𝚛 𝚜𝚕𝚒𝚍𝚎 𝚖𝚊𝚜𝚝𝚎𝚛 𝚋𝚎𝚏𝚘𝚛𝚎 𝚒𝚝 𝚠𝚛𝚒𝚝𝚎𝚜 𝚊 𝚜𝚒𝚗𝚐𝚕𝚎 𝚠𝚘𝚛𝚍 - 𝚜𝚘 𝚎𝚟𝚎𝚛𝚢 𝚐𝚎𝚗𝚎𝚛𝚊𝚝𝚎𝚍 𝚜𝚕𝚒𝚍𝚎 𝚊𝚕𝚛𝚎𝚊𝚍𝚢 𝚖𝚊𝚝𝚌𝚑𝚎𝚜 𝚢𝚘𝚞𝚛 𝚋𝚛𝚊𝚗𝚍．

• 𝚁𝚎𝚊𝚜𝚘𝚗𝚒𝚗𝚐, 𝚗𝚘𝚝 𝚓𝚞𝚜𝚝 𝚠𝚛𝚒𝚝𝚒𝚗𝚐: 𝙲𝚕𝚊𝚞𝚍𝚎 𝚜𝚝𝚛𝚞𝚌𝚝𝚞𝚛𝚎𝚜 𝚊𝚛𝚐𝚞𝚖𝚎𝚗𝚝𝚜 𝚕𝚘𝚐𝚒𝚌𝚊𝚕𝚕𝚢 (𝙿𝚛𝚘𝚋𝚕𝚎𝚖 → 𝙳𝚊𝚝𝚊 → 𝙸𝚗𝚜𝚒𝚐𝚑𝚝 → 𝚁𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗), 𝚜𝚘 𝚝𝚑𝚎 𝚍𝚎𝚌𝚔 𝚖𝚊𝚔𝚎𝚜 𝚜𝚎𝚗𝚜𝚎, 𝚗𝚘𝚝 𝚓𝚞𝚜𝚝 𝚕𝚘𝚘𝚔𝚜 𝚐𝚘𝚘𝚍．

• 𝙿𝚒𝚗𝚙𝚘𝚒𝚗𝚝 𝚎𝚍𝚒𝚝𝚒𝚗𝚐: 𝚃𝚊𝚛𝚐𝚎𝚝 𝚘𝚗𝚎 𝚜𝚕𝚒𝚍𝚎 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚍𝚒𝚜𝚝𝚞𝚛𝚋𝚒𝚗𝚐 𝚝𝚑𝚎 𝚛𝚎𝚜𝚝 — 𝚊 𝚕𝚎𝚟𝚎𝚕 𝚘𝚏 𝚌𝚘𝚗𝚝𝚛𝚘𝚕 𝚖𝚘𝚜𝚝 𝙰𝙸 𝚝𝚘𝚘𝚕𝚜 𝚍𝚘𝚗’𝚝 𝚘𝚏𝚏𝚎𝚛.

𝙽𝚘𝚝𝚎: 𝚃𝚑𝚎 𝚝𝚘𝚘𝚕𝚜 𝚊𝚗𝚍 𝚊𝚗𝚊𝚕𝚢𝚜𝚒𝚜 𝚏𝚎𝚊𝚝𝚞𝚛𝚎𝚍 𝚒𝚗 𝚝𝚑𝚒𝚜 𝚜𝚎𝚌𝚝𝚒𝚘𝚗 𝚍𝚎𝚖𝚘𝚗𝚜𝚝𝚛𝚊𝚝𝚎𝚍 𝚌𝚕𝚎𝚊𝚛 𝚟𝚊𝚕𝚞𝚎 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚘𝚞𝚛 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚝𝚎𝚜𝚝𝚒𝚗𝚐. 𝙾𝚞𝚛 𝚛𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗𝚜 𝚊𝚛𝚎 𝚎𝚗𝚝𝚒𝚛𝚎𝚕𝚢 𝚒𝚗𝚍𝚎𝚙𝚎𝚗𝚍𝚎𝚗𝚝 𝚊𝚗𝚍 𝚗𝚘𝚝 𝚒𝚗𝚏𝚕𝚞𝚎𝚗𝚌𝚎𝚍 𝚋𝚢 𝚝𝚑𝚎 𝚝𝚘𝚘𝚕 𝚌𝚛𝚎𝚊𝚝𝚘𝚛𝚜.

AI BITS & BYTES

Skills, not titles will drive India’s pay boom: EY

India’s compensation landscape is moving decisively from role-based to skill-based pay, according to the EY Future of Pay report. Nearly half of surveyed organisations are adopting skill-linked salary frameworks, with emerging tech talent in AI, Gen AI, machine learning, and engineering commanding premiums of up to 40%.

The report projects overall salary hikes of 9.1% in 2026, led by GCCs at 10.4%, reflecting sustained global demand for specialised digital expertise. Financial Services is expected to see increments of around 10%, followed by e-commerce at 9.9% and lifesciences and pharma at 9.7%.

As AI adoption accelerates, companies are recalibrating pay to reward productivity, applied skills and measurable business impact. Around 50-60% of large firms now use analytics in compensation planning, signalling a clear shift toward data-driven rewards strategies.

You may read the full article here.

ALSO READS

• AI will ‘automate whole economy’, Sam Altman says

• India ranks 6th in ‘AI Brain Race’ led by US, China

• Has OpenAI bought itself a security nightmare with OpenClaw

• Meta, AMD agree to AI chips deal worth over $100 billion

• AI agent invasion has people trying to pick winners

Twin AI data leaks expose over 1 billion KYC records

Cybersecurity researchers have uncovered two massive data leaks linked to two AI-related apps that have exposed the sensitive personal data and media files of millions of users globally.

The first leak has been attributed to an AI-powered Know Your Customer (KYC) tool used by digital identity verification provider IDMerit. The company is an AI-powered digital identity verification solutions provider that serves the fintech and financial services sectors by providing real-time verification tools.

The second leak is linked to an Android app named “Video AI Art Generator & Maker”, which has been downloaded more than 500,000 times on Google Play and rated 4.3 stars with over 11,000 reviews.

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

The expo started on Monday, 16 February, with ambition to position India as the AI voice of the Global South—championing inclusive policy built on open-source models and digital public infrastructure—and sought to marry geopolitical aspiration with technological leadership.

However, while large summits (over 20 heads of state, more than 250,000 visitors from 45 countries, and 600 startups spread across a 70,000 square-metre venue) promise visibility and access, they also demand precision planning and seamless execution.

In this week’s edition of Tech Talk:

• India’s MANAV vision for AI

• AGI likely in 5-8 years? Demis Hasabbis has cues

• Will AI replace you at work? CEOs are divided

• Who said what at India AI Summit

• AI Tool of the Week: How to use Claude in Excel

Pageantry & Chaos

Mint cautioned earlier this week — global gatherings often risk flirting with pageantry, and chaos. Day 1 was marked by chaos with registration systems crashing, cash-only payment counters at an AI summit (an irony in a country that touts UPI as a global payments model), serpentine queues, patchy networks, and food stalls shutting down without notice. The security tightened abruptly to accommodate PM Modi’s revised schedule, adding to the confusion. In the melee, a founder reported wearable products stolen from a stall.

For exhibitors who paid between ₹15 lakh and ₹50 lakh for the week, the opening day was close to a washout. Several privately questioned whether the investment would translate into meaningful leads, partnerships, or policy traction.

Then, confusion erupted over Bill Gates’ participation at the India AI Summit following conflicting statements from government sources and his foundation. Gates finally dropped out. Nvidia CEO Jensen Huang also pulled out.

These setbacks were accompanied by moments of embarrassment: a professor manning the Galgotias University booth claimed that a Chinese-made Unitree Go2 robotic dog, named ‘Orion’, had been built by the varsity’s centre of excellence.

Following a massive online backlash, the varsity issued a statement that the professor was “ill-informed“, but it was asked to leave the summit.

“The main problem is they claimed to be something they are not. They misled. The whole world is here,” Abhishek Singh, additional secretary at MeitY and India AI Mission CEO, said while acknowledging the controversy. Union IT Minister Ashwini Vaishnaw too apologised for the issues on the first day of the India AI Summit, assuring attendees that efforts are being made to enhance their experience.

That said, while these incidents cast a blot on the event, the buzz around AI will endure since the summit successfully brought together policymakers, industry leaders and technology experts to deliberate on AI innovation, governance and real-world applications. World leaders underscored the need for equitable, collaborative and responsible development of AI. India also secured a place in the Guinness World Records after nearly a quarter of a million people signed up for an AI responsibility campaign, within 24 hours.

Who said What at India AI Summit

We are entering an era where humans and intelligence system co create, co-work and co-evolve. — Narendra Modi, Prime Minister of India.

He announced India's MANAV vision for AI — M: Moral and Ethical System; A: Accountable of Governance; N: National Sovereignty (right to data); A: Accessible and inclusive; V: Valid and legitimate.

We are at a threshold moment where Artificial General Intelligence (AGI) is on the horizon, maybe in the next 5-8 years. Demis Hassabis, co-founder and CEO, Google DeepMind

A scientific method should be adopted to understand AGI's full capabilities, build good guardrails and monitoring systems—an approach Hassabis believes will help unlock the full potential of technology for the benefit of humans.

India will emerge as one of the greatest AI powers in the world in the 21st century. AI is the mantra that powers every yantra (machine) to perform better. — Mukesh Ambani, chairman, Reliance Industries

Reliance and Jio will plow as much as ₹10 trillion over seven years into AI-related infrastructure, joining the global rush into technology’s fastest-growing arena.

I remember Vizag being a quiet and modest coastal city brimming with potential. Now in that same city, Google is establishing a full-stack AI hub. — Sundar Pichai, CEO, Alphabet

Google is investing $15 billion in India to build a full-stack AI hub in Vizag, including a gigawatt-scale data centre and a new international subsea cable. That will bring jobs and cutting-edge AI to people and businesses across India.

AI, in my mind, is the next big infrastructure. It is the infrastructure of intelligence. It will have a very profound impact, exactly the same way…as electricity, or the internet. — N. Chandrasekaran, chairman, Tata Sons Pvt. Ltd.

OpenAI is partnering with Tata Group for an AI push into India, which includes up to gigawatt-scale data centres and embedding Agentic AI tools for clients handled by Tata Consultancy Services.

India is currently leading the world in AI adoption and is poised to become one of the largest markets for the technology. — Sam Altman, CEO, OpenAI

Acknowledging that AI advancement will definitely impact the job market, Altman said he remains confident in human adaptability.

AI TOOL OF THE WEEK

by AI&Beyond, with Jaspreet Bindra and Anuj Magazine

𝚃𝚑𝚎 𝙰𝙸 𝚌𝚊𝚙𝚊𝚋𝚒𝚕𝚒𝚝𝚢 𝚠𝚎 𝚞𝚗𝚕𝚘𝚌𝚔 𝚝𝚘𝚍𝚊𝚢 𝚒𝚜 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝙲𝚕𝚊𝚞𝚍𝚎 𝚒𝚗 𝙴𝚡𝚌𝚎𝚕．

𝚆𝚑𝚊𝚝 𝚙𝚛𝚘𝚋𝚕𝚎𝚖 𝚍𝚘𝚎𝚜 𝚒𝚝 𝚜𝚘𝚕𝚟𝚎? 𝙴𝚟𝚎𝚛𝚢 𝚏𝚒𝚗𝚊𝚗𝚌𝚎 𝚙𝚛𝚘𝚏𝚎𝚜𝚜𝚒𝚘𝚗𝚊𝚕 𝚔𝚗𝚘𝚠𝚜 𝚝𝚑𝚎 𝚏𝚎𝚎𝚕𝚒𝚗𝚐: 𝚒𝚝’𝚜 𝟷𝟷:𝟶𝟶 𝚊𝚖, 𝚋𝚎𝚏𝚘𝚛𝚎 𝚊 𝚋𝚘𝚊𝚛𝚍 𝚖𝚎𝚎𝚝𝚒𝚗𝚐, 𝚢𝚘𝚞’𝚛𝚎 𝚜𝚝𝚊𝚛𝚒𝚗𝚐 𝚊𝚝 𝚊 𝟺𝟶-𝚝𝚊𝚋 𝚏𝚒𝚗𝚊𝚗𝚌𝚒𝚊𝚕 𝚖𝚘𝚍𝚎𝚕 𝚜𝚘𝚖𝚎𝚘𝚗𝚎 𝚎𝚕𝚜𝚎 𝚋𝚞𝚒𝚕𝚝, 𝚊 𝚏𝚘𝚛𝚖𝚞𝚕𝚊 𝚒𝚜 𝚋𝚛𝚘𝚔𝚎𝚗, 𝚊𝚗𝚍 𝚢𝚘𝚞 𝚑𝚊𝚟𝚎 𝚗𝚘 𝚒𝚍𝚎𝚊 𝚠𝚑𝚎𝚛𝚎 𝚝𝚘 𝚎𝚟𝚎𝚗 𝚜𝚝𝚊𝚛𝚝 𝚕𝚘𝚘𝚔𝚒𝚗𝚐. 𝙰𝚗𝚍 𝚠𝚑𝚎𝚗 𝚕𝚎𝚊𝚍𝚎𝚛𝚜𝚑𝚒𝚙 𝚊𝚜𝚔𝚜 “𝚠𝚑𝚊𝚝 𝚑𝚊𝚙𝚙𝚎𝚗𝚜 𝚝𝚘 𝚘𝚞𝚛 𝙿𝙰𝚃 𝚒𝚏 𝚖𝚊𝚛𝚐𝚒𝚗𝚜 𝚍𝚛𝚘𝚙 𝟹%?”, 𝚛𝚎𝚋𝚞𝚒𝚕𝚍𝚒𝚗𝚐 𝚝𝚑𝚎 𝚜𝚌𝚎𝚗𝚊𝚛𝚒𝚘 𝚖𝚊𝚗𝚞𝚊𝚕𝚕𝚢 𝚝𝚊𝚔𝚎𝚜 𝚊𝚗𝚘𝚝𝚑𝚎𝚛 𝚑𝚘𝚞𝚛 𝚢𝚘𝚞 𝚍𝚘𝚗’𝚝 𝚑𝚊𝚟𝚎.

𝙲𝚕𝚊𝚞𝚍𝚎 𝚒𝚗 𝙴𝚡𝚌𝚎𝚕 𝚙𝚞𝚝𝚜 𝚊𝚗 𝙰𝙸 𝚊𝚗𝚊𝚕𝚢𝚜𝚝 𝚍𝚒𝚛𝚎𝚌𝚝𝚕𝚢 𝚒𝚗𝚜𝚒𝚍𝚎 𝚢𝚘𝚞𝚛 𝚜𝚙𝚛𝚎𝚊𝚍𝚜𝚑𝚎𝚎𝚝, 𝚍𝚎𝚜𝚒𝚐𝚗𝚎𝚍 𝚝𝚘 𝚛𝚎𝚊𝚍 𝚏𝚘𝚛𝚖𝚞𝚕𝚊𝚜, 𝚝𝚊𝚋𝚜 𝚊𝚗𝚍 𝚍𝚎𝚙𝚎𝚗𝚍𝚎𝚗𝚌𝚒𝚎𝚜 𝚊𝚌𝚛𝚘𝚜𝚜 𝚊 𝚠𝚘𝚛𝚔𝚋𝚘𝚘𝚔 𝚋𝚎𝚏𝚘𝚛𝚎 𝚛𝚎𝚜𝚙𝚘𝚗𝚍𝚒𝚗𝚐 𝚝𝚘 𝚚𝚞𝚎𝚛𝚒𝚎𝚜. 𝙸𝚝 𝚘𝚙𝚎𝚛𝚊𝚝𝚎𝚜 𝚠𝚒𝚝𝚑𝚒𝚗 𝚝𝚑𝚎 𝚕𝚒𝚟𝚎 𝚏𝚒𝚕𝚎, 𝚌𝚒𝚝𝚎𝚜 𝚝𝚑𝚎 𝚌𝚎𝚕𝚕𝚜 𝚋𝚎𝚑𝚒𝚗𝚍 𝚎𝚊𝚌𝚑 𝚊𝚗𝚜𝚠𝚎𝚛, 𝚊𝚗𝚍 𝚖𝚊𝚔𝚎𝚜 𝚌𝚑𝚊𝚗𝚐𝚎𝚜 𝚘𝚗𝚕𝚢 𝚠𝚒𝚝𝚑 𝚞𝚜𝚎𝚛 𝚊𝚙𝚙𝚛𝚘𝚟𝚊𝚕．

𝙷𝚘𝚠 𝚝𝚘 𝚊𝚌𝚌𝚎𝚜𝚜: 𝙸𝚗𝚜𝚝𝚊𝚕𝚕 𝚏𝚛𝚘𝚖 𝚝𝚑𝚎 𝙼𝚒𝚌𝚛𝚘𝚜𝚘𝚏𝚝 𝙼𝚊𝚛𝚔𝚎𝚝𝚙𝚕𝚊𝚌𝚎 → 𝚜𝚎𝚊𝚛𝚌𝚑 ‘𝙲𝚕𝚊𝚞𝚍𝚎 𝚋𝚢 𝙰𝚗𝚝𝚑𝚛𝚘𝚙𝚒𝚌 𝚏𝚘𝚛 𝙴𝚡𝚌𝚎𝚕’ → 𝚜𝚒𝚐𝚗 𝚒𝚗 𝚠𝚒𝚝𝚑 𝚢𝚘𝚞𝚛 𝙲𝚕𝚊𝚞𝚍𝚎 𝚊𝚌𝚌𝚘𝚞𝚗𝚝 (𝙿𝚛𝚘, 𝙼𝚊𝚡, 𝚃𝚎𝚊𝚖 𝚘𝚛 𝙴𝚗𝚝𝚎𝚛𝚙𝚛𝚒𝚜𝚎 𝚙𝚕𝚊𝚗 𝚛𝚎𝚚𝚞𝚒𝚛𝚎𝚍)

𝙲𝚕𝚊𝚞𝚍𝚎 𝚒𝚗 𝙴𝚡𝚌𝚎𝚕 𝚌𝚊𝚗 𝚑𝚎𝚕𝚙 𝚢𝚘𝚞:

• 𝙴𝚡𝚙𝚕𝚊𝚒𝚗 𝚊𝚗𝚢 𝚖𝚘𝚍𝚎𝚕 𝚒𝚗𝚜𝚝𝚊𝚗𝚝𝚕𝚢: 𝙰𝚜𝚔 𝚠𝚑𝚊𝚝’𝚜 𝚍𝚛𝚒𝚟𝚒𝚗𝚐 𝚊 𝚗𝚞𝚖𝚋𝚎𝚛 𝚊𝚗𝚍 𝚐𝚎𝚝 𝚊 𝚌𝚎𝚕𝚕-𝚌𝚒𝚝𝚎𝚍 𝚊𝚗𝚜𝚠𝚎𝚛 𝚒𝚗 𝚜𝚎𝚌𝚘𝚗𝚍𝚜.

• 𝚃𝚎𝚜𝚝 𝚜𝚌𝚎𝚗𝚊𝚛𝚒𝚘𝚜 𝚕𝚒𝚟𝚎: 𝙲𝚑𝚊𝚗𝚐𝚎 𝚊𝚗 𝚊𝚜𝚜𝚞𝚖𝚙𝚝𝚒𝚘𝚗 𝚊𝚗𝚍 𝚠𝚊𝚝𝚌𝚑 𝚝𝚑𝚎 𝚎𝚗𝚝𝚒𝚛𝚎 𝙿&𝙻 𝚛𝚎𝚌𝚊𝚕𝚌𝚞𝚕𝚊𝚝𝚎 𝚠𝚒𝚝𝚑 𝚎𝚟𝚎𝚛𝚢 𝚒𝚖𝚙𝚊𝚌𝚝 𝚑𝚒𝚐𝚑𝚕𝚒𝚐𝚑𝚝𝚎𝚍.

• 𝙳𝚎𝚋𝚞𝚐 𝚎𝚛𝚛𝚘𝚛𝚜 𝚒𝚗 𝚖𝚘𝚖𝚎𝚗𝚝𝚜: 𝚃𝚛𝚊𝚌𝚎 𝚊 #𝚁𝙴𝙵! 𝚘𝚛 #𝚅𝙰𝙻𝚄𝙴! 𝚎𝚛𝚛𝚘𝚛 𝚝𝚘 𝚒𝚝𝚜 𝚜𝚘𝚞𝚛𝚌𝚎 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚖𝚊𝚗𝚞𝚊𝚕𝚕𝚢 𝚌𝚑𝚎𝚌𝚔𝚒𝚗𝚐 𝚎𝚟𝚎𝚛𝚢 𝚏𝚘𝚛𝚖𝚞𝚕𝚊.

𝙴𝚡𝚊𝚖𝚙𝚕𝚎: 𝙰 𝙲𝙵𝙾’𝚜 𝚝𝚎𝚊𝚖 𝚒𝚜 𝚛𝚎𝚟𝚒𝚎𝚠𝚒𝚗𝚐 𝚚𝚞𝚊𝚛𝚝𝚎𝚛𝚕𝚢 𝚙𝚛𝚘𝚓𝚎𝚌𝚝𝚒𝚘𝚗𝚜 𝚋𝚎𝚏𝚘𝚛𝚎 𝚊 𝚕𝚎𝚊𝚍𝚎𝚛𝚜𝚑𝚒𝚙 𝚙𝚛𝚎𝚜𝚎𝚗𝚝𝚊𝚝𝚒𝚘𝚗. 𝚃𝚑𝚎 𝙴𝙱𝙸𝚃𝙳𝙰 𝚖𝚊𝚛𝚐𝚒𝚗 𝚑𝚊𝚜 𝚜𝚑𝚒𝚏𝚝𝚎𝚍 𝚞𝚗𝚎𝚡𝚙𝚎𝚌𝚝𝚎𝚍𝚕𝚢 𝚊𝚗𝚍 𝚗𝚘 𝚘𝚗𝚎 𝚌𝚊𝚗 𝚎𝚡𝚙𝚕𝚊𝚒𝚗 𝚠𝚑𝚢. 𝙷𝚎𝚛𝚎’𝚜 𝚑𝚘𝚠 𝙲𝚕𝚊𝚞𝚍𝚎 𝚒𝚗 𝙴𝚡𝚌𝚎𝚕 𝚑𝚎𝚕𝚙𝚜 𝚏𝚛𝚘𝚖 𝚍𝚒𝚊𝚐𝚗𝚘𝚜𝚒𝚜 𝚝𝚘 𝚋𝚘𝚊𝚛𝚍-𝚛𝚎𝚊𝚍𝚢 𝚘𝚞𝚝𝚙𝚞𝚝:

#𝟷: 𝙴𝚡𝚙𝚕𝚊𝚒𝚗 𝚠𝚑𝚊𝚝’𝚜 𝚍𝚛𝚒𝚟𝚒𝚗𝚐 𝚊 𝚗𝚞𝚖𝚋𝚎𝚛

𝚃𝚑𝚎 𝚝𝚎𝚊𝚖 𝚌𝚊𝚗’𝚝 𝚎𝚡𝚙𝚕𝚊𝚒𝚗 𝚠𝚑𝚢 𝚝𝚑𝚎 𝙴𝙱𝙸𝚃𝙳𝙰 𝚖𝚊𝚛𝚐𝚒𝚗 𝚍𝚛𝚘𝚙𝚙𝚎𝚍 𝚋𝚎𝚝𝚠𝚎𝚎𝚗 𝚝𝚠𝚘 𝚚𝚞𝚊𝚛𝚝𝚎𝚛𝚜． 𝙸𝚗𝚜𝚝𝚎𝚊𝚍 𝚘𝚏 𝚝𝚛𝚊𝚌𝚒𝚗𝚐 𝚏𝚘𝚛𝚖𝚞𝚕𝚊𝚜 𝚖𝚊𝚗𝚞𝚊𝚕𝚕𝚢, 𝚝𝚑𝚎𝚢 𝚊𝚜𝚔:

“𝚆𝚑𝚢 𝚒𝚜 𝚝𝚑𝚎 𝙴𝙱𝙸𝚃𝙳𝙰 𝚖𝚊𝚛𝚐𝚒𝚗 𝚕𝚘𝚠𝚎𝚛 𝚒𝚗 𝚀𝟹 𝚟𝚎𝚛𝚜𝚞𝚜 𝚀𝟸? 𝚃𝚛𝚊𝚌𝚎 𝚝𝚑𝚎 𝚌𝚊𝚞𝚜𝚎 𝚊𝚗𝚍 𝚌𝚒𝚝𝚎 𝚝𝚑𝚎 𝚜𝚙𝚎𝚌𝚒𝚏𝚒𝚌 𝚛𝚘𝚠𝚜 𝚍𝚛𝚒𝚟𝚒𝚗𝚐 𝚝𝚑𝚎 𝚌𝚑𝚊𝚗𝚐𝚎.”

𝙲𝚕𝚊𝚞𝚍𝚎 𝚛𝚎𝚊𝚍𝚜 𝚝𝚑𝚎 𝚎𝚗𝚝𝚒𝚛𝚎 𝚖𝚘𝚍𝚎𝚕 𝚊𝚗𝚍 𝚛𝚎𝚜𝚙𝚘𝚗𝚍𝚜 𝚒𝚗 𝚜𝚎𝚌𝚘𝚗𝚍𝚜, 𝚙𝚘𝚒𝚗𝚝𝚒𝚗𝚐 𝚝𝚘 𝚝𝚑𝚎 𝚎𝚡𝚊𝚌𝚝 𝚌𝚘𝚜𝚝 𝚕𝚒𝚗𝚎 𝚝𝚑𝚊𝚝 𝚎𝚡𝚙𝚊𝚗𝚍𝚎𝚍 𝚊𝚜 𝚊 𝚙𝚎𝚛𝚌𝚎𝚗𝚝𝚊𝚐𝚎 𝚘𝚏 𝚛𝚎𝚟𝚎𝚗𝚞𝚎, 𝚠𝚒𝚝𝚑 𝚌𝚎𝚕𝚕 𝚛𝚎𝚏𝚎𝚛𝚎𝚗𝚌𝚎𝚜 𝚒𝚗𝚌𝚕𝚞𝚍𝚎𝚍．

#𝟸: 𝙻𝚒𝚟𝚎 𝚜𝚌𝚎𝚗𝚊𝚛𝚒𝚘 𝚝𝚎𝚜𝚝𝚒𝚗𝚐

𝙻𝚎𝚊𝚍𝚎𝚛𝚜𝚑𝚒𝚙 𝚠𝚊𝚗𝚝𝚜 𝚝𝚘 𝚔𝚗𝚘𝚠 𝚝𝚑𝚎 𝚍𝚘𝚠𝚗𝚜𝚒𝚍𝚎 𝚒𝚖𝚙𝚊𝚌𝚝 𝚋𝚎𝚏𝚘𝚛𝚎 𝚝𝚑𝚎 𝚋𝚘𝚊𝚛𝚍 𝚖𝚎𝚎𝚝𝚒𝚗𝚐:

“𝙲𝚑𝚊𝚗𝚐𝚎 𝚝𝚑𝚎 𝙴𝙱𝙸𝚃𝙳𝙰 𝚖𝚊𝚛𝚐𝚒𝚗 𝚊𝚜𝚜𝚞𝚖𝚙𝚝𝚒𝚘𝚗 𝚝𝚘 𝟹𝟶% 𝚊𝚗𝚍 𝚜𝚑𝚘𝚠 𝚖𝚎 𝚝𝚑𝚎 𝚗𝚎𝚠 𝚏𝚞𝚕𝚕-𝚢𝚎𝚊𝚛 𝙿𝙰𝚃. 𝙷𝚒𝚐𝚑𝚕𝚒𝚐𝚑𝚝 𝚎𝚟𝚎𝚛𝚢 𝚌𝚎𝚕𝚕 𝚝𝚑𝚊𝚝 𝚌𝚑𝚊𝚗𝚐𝚎𝚜.”

𝙲𝚕𝚊𝚞𝚍𝚎 𝚞𝚙𝚍𝚊𝚝𝚎𝚜 𝚝𝚑𝚎 𝚊𝚜𝚜𝚞𝚖𝚙𝚝𝚒𝚘𝚗, 𝚛𝚎𝚌𝚊𝚕𝚌𝚞𝚕𝚊𝚝𝚎𝚜 𝚝𝚑𝚎 𝚎𝚗𝚝𝚒𝚛𝚎 𝙿&𝙻 𝚌𝚊𝚜𝚌𝚊𝚍𝚎, 𝚑𝚒𝚐𝚑𝚕𝚒𝚐𝚑𝚝𝚜 𝚒𝚖𝚙𝚊𝚌𝚝𝚎𝚍 𝚌𝚎𝚕𝚕𝚜 𝚒𝚗 𝚢𝚎𝚕𝚕𝚘𝚠, 𝚊𝚗𝚍 𝚜𝚞𝚖𝚖𝚊𝚛𝚒𝚜𝚎𝚜 𝚝𝚑𝚎 𝙿𝙰𝚃 𝚟𝚊𝚛𝚒𝚊𝚗𝚌𝚎 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚝𝚘𝚞𝚌𝚑𝚒𝚗𝚐 𝚊 𝚜𝚒𝚗𝚐𝚕𝚎 𝚏𝚘𝚛𝚖𝚞𝚕𝚊 𝚜𝚝𝚛𝚞𝚌𝚝𝚞𝚛𝚎.

#𝟹 𝙿𝚛𝚘𝚓𝚎𝚌𝚝 𝚛𝚒𝚜𝚔 𝚏𝚕𝚊𝚐𝚐𝚒𝚗𝚐

𝚃𝚑𝚎 𝙲𝙵𝙾 𝚠𝚊𝚗𝚝𝚜 𝚝𝚘 𝚔𝚗𝚘𝚠 𝚠𝚑𝚒𝚌𝚑 𝚙𝚛𝚘𝚓𝚎𝚌𝚝𝚜 𝚗𝚎𝚎𝚍 𝚞𝚛𝚐𝚎𝚗𝚝 𝚊𝚝𝚝𝚎𝚗𝚝𝚒𝚘𝚗 𝚋𝚎𝚏𝚘𝚛𝚎 𝚚𝚞𝚊𝚛𝚝𝚎𝚛-𝚎𝚗𝚍:

“𝚆𝚑𝚒𝚌𝚑 𝚙𝚛𝚘𝚓𝚎𝚌𝚝𝚜 𝚑𝚊𝚟𝚎 𝚌𝚘𝚕𝚕𝚎𝚌𝚝𝚎𝚍 𝚕𝚎𝚜𝚜 𝚝𝚑𝚊𝚗 𝟻𝟶% 𝚘𝚏 𝚝𝚑𝚎𝚒𝚛 𝚋𝚘𝚘𝚔𝚒𝚗𝚐𝚜 𝚟𝚊𝚕𝚞𝚎? 𝙽𝚊𝚖𝚎 𝚝𝚑𝚎𝚖, 𝚜𝚝𝚊𝚝𝚎 𝚝𝚑𝚎 𝚞𝚗𝚌𝚘𝚕𝚕𝚎𝚌𝚝𝚎𝚍 𝚊𝚖𝚘𝚞𝚗𝚝, 𝚊𝚗𝚍 𝚏𝚕𝚊𝚐 𝚠𝚑𝚒𝚌𝚑 𝚒𝚜 𝚖𝚘𝚜𝚝 𝚊𝚝 𝚛𝚒𝚜𝚔 𝚐𝚒𝚟𝚎𝚗 𝚒𝚝𝚜 𝚍𝚎𝚊𝚍𝚕𝚒𝚗𝚎.”

𝙲𝚕𝚊𝚞𝚍𝚎 𝚜𝚌𝚊𝚗𝚜 𝚝𝚑𝚎 𝚏𝚞𝚕𝚕 𝚙𝚘𝚛𝚝𝚏𝚘𝚕𝚒𝚘 𝚝𝚛𝚊𝚌𝚔𝚎𝚛 𝚊𝚗𝚍 𝚛𝚎𝚝𝚞𝚛𝚗𝚜 𝚊 𝚛𝚊𝚗𝚔𝚎𝚍 𝚛𝚒𝚜𝚔 𝚜𝚞𝚖𝚖𝚊𝚛𝚢—𝚗𝚘 𝚙𝚒𝚟𝚘𝚝 𝚝𝚊𝚋𝚕𝚎 𝚗𝚎𝚎𝚍𝚎𝚍.

#𝟺 𝙱𝚘𝚊𝚛𝚍 𝚌𝚘𝚖𝚖𝚎𝚗𝚝𝚊𝚛𝚢, 𝚠𝚛𝚒𝚝𝚝𝚎𝚗 𝚒𝚗𝚜𝚝𝚊𝚗𝚝𝚕𝚢

𝚆𝚒𝚝𝚑 𝚗𝚞𝚖𝚋𝚎𝚛𝚜 𝚏𝚒𝚗𝚊𝚕𝚒𝚜𝚎𝚍, 𝚝𝚑𝚎 𝚝𝚎𝚊𝚖 𝚗𝚎𝚎𝚍𝚜 𝚊 𝙲𝙵𝙾 𝚌𝚘𝚖𝚖𝚎𝚗𝚝𝚊𝚛𝚢:

“𝚄𝚜𝚒𝚗𝚐 𝚝𝚑𝚎 𝙵𝚈𝟸𝟻 𝚊𝚌𝚝𝚞𝚊𝚕𝚜 𝚊𝚗𝚍 𝙵𝚈𝟸𝟼 𝚙𝚛𝚘𝚓𝚎𝚌𝚝𝚒𝚘𝚗𝚜 𝚒𝚗 𝚝𝚑𝚒𝚜 𝚜𝚑𝚎𝚎𝚝, 𝚠𝚛𝚒𝚝𝚎 𝚊 𝟹-𝚜𝚎𝚗𝚝𝚎𝚗𝚌𝚎 𝚋𝚘𝚊𝚛𝚍 𝚌𝚘𝚖𝚖𝚎𝚗𝚝𝚊𝚛𝚢 𝚜𝚞𝚖𝚖𝚊𝚛𝚒𝚜𝚒𝚗𝚐 𝚙𝚎𝚛𝚏𝚘𝚛𝚖𝚊𝚗𝚌𝚎, 𝚘𝚞𝚝𝚕𝚘𝚘𝚔, 𝚊𝚗𝚍 𝚝𝚑𝚎 𝚔𝚎𝚢 𝚊𝚜𝚜𝚞𝚖𝚙𝚝𝚒𝚘𝚗 𝚝𝚘 𝚠𝚊𝚝𝚌𝚑.”

𝙲𝚕𝚊𝚞𝚍𝚎 𝚍𝚛𝚊𝚏𝚝𝚜 𝚒𝚝 𝚍𝚒𝚛𝚎𝚌𝚝𝚕𝚢 𝚏𝚛𝚘𝚖 𝚝𝚑𝚎 𝚕𝚒𝚟𝚎 𝚌𝚎𝚕𝚕 𝚟𝚊𝚕𝚞𝚎𝚜—𝚏𝚊𝚌𝚝𝚞𝚊𝚕, 𝚌𝚒𝚝𝚎𝚍, 𝚊𝚗𝚍 𝚛𝚎𝚊𝚍𝚢 𝚝𝚘 𝚙𝚛𝚎𝚜𝚎𝚗𝚝.

𝚆𝚑𝚊𝚝 𝚖𝚊𝚔𝚎𝚜 𝙲𝚕𝚊𝚞𝚍𝚎 𝚒𝚗 𝙴𝚡𝚌𝚎𝚕 𝚜𝚙𝚎𝚌𝚒𝚊𝚕？

𝙸𝚝 𝚛𝚎𝚊𝚍𝚜 𝚢𝚘𝚞𝚛 𝚠𝚑𝚘𝚕𝚎 𝚠𝚘𝚛𝚔𝚋𝚘𝚘𝚔: 𝙽𝚘𝚝 𝚓𝚞𝚜𝚝 𝚝𝚑𝚎 𝚜𝚑𝚎𝚎𝚝 𝚢𝚘𝚞’𝚛𝚎 𝚘𝚗 𝚎𝚟𝚎𝚛𝚢 𝚝𝚊𝚋, 𝚎𝚟𝚎𝚛𝚢 𝚏𝚘𝚛𝚖𝚞𝚕𝚊, 𝚎𝚟𝚎𝚛𝚢 𝚍𝚎𝚙𝚎𝚗𝚍𝚎𝚗𝚌𝚢—𝚊𝚕𝚕 𝚊𝚝 𝚘𝚗𝚌𝚎．

𝙲𝚎𝚕𝚕-𝚕𝚎𝚟𝚎𝚕 𝚌𝚒𝚝𝚊𝚝𝚒𝚘𝚗𝚜: 𝙴𝚟𝚎𝚛𝚢 𝚊𝚗𝚜𝚠𝚎𝚛 𝚛𝚎𝚏𝚎𝚛𝚎𝚗𝚌𝚎𝚜 𝚝𝚑𝚎 𝚎𝚡𝚊𝚌𝚝 𝚌𝚎𝚕𝚕𝚜 𝚋𝚎𝚑𝚒𝚗𝚍 𝚒𝚝, 𝚜𝚘 𝚢𝚘𝚞 𝚌𝚊𝚗 𝚟𝚎𝚛𝚒𝚏𝚢 𝚝𝚑𝚎 𝚕𝚘𝚐𝚒𝚌 𝚛𝚊𝚝𝚑𝚎𝚛 𝚝𝚑𝚊𝚗 𝚓𝚞𝚜𝚝 𝚝𝚛𝚞𝚜𝚝 𝚒𝚝．

𝙲𝚘𝚗𝚗𝚎𝚌𝚝𝚎𝚍 𝚝𝚘 𝚕𝚒𝚟𝚎 𝚏𝚒𝚗𝚊𝚗𝚌𝚒𝚊𝚕 𝚍𝚊𝚝𝚊: 𝚅𝚒𝚊 𝙼𝙲𝙿 𝚌𝚘𝚗𝚗𝚎𝚌𝚝𝚘𝚛𝚜, 𝚒𝚝 𝚌𝚊𝚗 𝚙𝚞𝚕𝚕 𝚒𝚗 𝙵𝚊𝚌𝚝𝚂𝚎𝚝, 𝚂&𝙿 𝙶𝚕𝚘𝚋𝚊𝚕, 𝚊𝚗𝚍 𝙱𝚕𝚘𝚘𝚖𝚋𝚎𝚛𝚐 𝚍𝚊𝚝𝚊 𝚍𝚒𝚛𝚎𝚌𝚝𝚕𝚢 𝚒𝚗𝚝𝚘 𝚢𝚘𝚞𝚛 𝚖𝚘𝚍𝚎𝚕．

𝙽𝚘𝚝𝚎: 𝚃𝚑𝚎 𝚝𝚘𝚘𝚕𝚜 𝚊𝚗𝚍 𝚊𝚗𝚊𝚕𝚢𝚜𝚒𝚜 𝚏𝚎𝚊𝚝𝚞𝚛𝚎𝚍 𝚒𝚗 𝚝𝚑𝚒𝚜 𝚜𝚎𝚌𝚝𝚒𝚘𝚗 𝚍𝚎𝚖𝚘𝚗𝚜𝚝𝚛𝚊𝚝𝚎𝚍 𝚌𝚕𝚎𝚊𝚛 𝚟𝚊𝚕𝚞𝚎 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚘𝚞𝚛 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚝𝚎𝚜𝚝𝚒𝚗𝚐． 𝙾𝚞𝚛 𝚛𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗𝚜 𝚊𝚛𝚎 𝚎𝚗𝚝𝚒𝚛𝚎𝚕𝚢 𝚒𝚗𝚍𝚎𝚙𝚎𝚗𝚍𝚎𝚗𝚝 𝚊𝚗𝚍 𝚗𝚘𝚝 𝚒𝚗𝚏𝚕𝚞𝚎𝚗𝚌𝚎𝚍 𝚋𝚢 𝚝𝚑𝚎 𝚝𝚘𝚘𝚕 𝚌𝚛𝚎𝚊𝚝𝚘𝚛𝚜．

India AI Summit: Bits & Bytes

Will AI replace you at work? CEOs are divided

A majority of Indian IT sector workers worry AI could replace their jobs, according to a study by economic think tank ICIER. The report found that 65% companies reduced hiring after adopting Generative AI, as fewer fresh recruits were needed to handle similar or higher workloads aided by AI tools.

Upskilling efforts, however, remain limited. And only 4% of firms said they had trained more than half their workforce in AI-related skills, with companies citing shortages of qualified trainers (70%) and high costs (68%) as key constraints.

Thinking divide on AI: As newer AI tools continue to emerge and attract attention, the gap between supporters and critics is only widening.

In an interview with Hindustan Times ahead of the India AI Summit, tech billionaire and venture capitalist, Vinod Khosla, warned that IT services and BPOs will “almost completely disappear” within five years as AI tools will take over tasks currently handled by human workers. Khosla noted that AI would wipe out most expertise-based professions within the next 15 years, though it could democratise access to healthcare and education. BPO, or business process outsourcing, refers to a large segment of work that companies outsource to third-party service providers, which, according to Khosla, will face major disruption. He added that IT services are expected to be similarly affected, with significant implications for India.

• Sanjeev Bikhchandani, founder and executive vice chairman of Info Edge (India) Ltd., the company that owns Naukri.com, asserted that AI will not eliminate jobs but will significantly enhance productivity.

• Sateesh Seetharamiah, chief executive officer of EdgeVerve, described AI as a “capability multiplier”. He acknowledged that productivity improvements are undeniable, but emphasised that “ultimately there has to be a human being to take accountability”. According to him, employment itself is not under threat—rather, “the nature of jobs will change”.

• Vineet Nayar, founder-chairman of Sampark Foundation and former CEO of HCLTech, suggested that automation could eliminate nearly half of existing jobs, but an equivalent number of new roles may also be created.

• Puneet Chandok, President of Microsoft India and South Asia, said AI will not eliminate employment but reshape it.

AI talent hubs: Interestingly, Bengaluru and Delhi-NCR collectively hold a 50% share of India’s AI job postings, cementing their position as the country’s two largest AI talent hubs, according to an analysis by CBRE Research. The findings from a study of over 64,500 AI-tagged job listings on Naukri.com as of December 2026 reveal that Bengaluru leads in AI job hiring with a 25.4% share, closely followed by Delhi NCR at 24.8%.

BharatGen’s sovereign AI model

BharatGen, an IIT Bombay-incubated and government-funded consortium, unveiled PARAM-2, a 17-billion-parameter multilingual foundation model, at the India AI Impact Summit. Built on India-centric datasets, it leverages Mixture of Experts (MoE) architecture, and is designed to support all 22 official Indian languages.

BharatGen is being positioned as a public digital goods, enabling secure, local deployment across governance, education and industry.

A few days back, Bengaluru-based Sarvam AI, backed by investors such as Peak XV and Vinod Khosla, unveiled three foundational AI models with 3-billion, 30-billion and 105-billion parameters. The company, the first to be funded by the India AI Mission in April 2025, said its models can be used through voice commands and are accessible through 22 Indian languages. It also offers Agentic AI applications that can execute a range of tasks with minimal human intervention.

The government is investing in indigenous AI models to tackle bias, according to S Krishnan, secretary in the Ministry of Electronics and Information Technology (MeitY). He explained that sovereignty is a very important issue, and India needs to have AI models where “nobody else has the kill switch, and it was always with us”. According to him, a country of the size of India needs to necessarily have control over the way AI models function.

Nvidia, sans Jensen Huang, goes into GPU overdrive

Nvidia Corp. has announced a spate of partnerships with top Indian enterprises for its GPU chips. The company has partnered with Reliance New Energy, Hero MotoCorp, L&T Semiconductor, and Tata Consultancy Services, among others. Nvidia also said India’s IT services firms—including Infosys Ltd., Persistent Systems, Tech Mahindra, and Wipro—are using its enterprise software platforms to onboard new software management capabilities for clients.

The partnerships involve access to Nvidia’s computing chips, as well as open-source AI models and AI development software across industrial applications.

Meanwhile, Adani Group’s $100-billion investment commitment for an AI-ready data centre and power ecosystem, along with investment commitments of other Indian conglomerates like Reliance Industries, Tata Group and Larsen & Toubro, is taking a leaf out of the $500-billion Project Stargate playbook of the US.

The ecosystem not only gives Indian conglomerates a cost advantage but also better control over the operationalisation of their multi-gigawatt data centres, experts said. So far, investment commitments by top Indian conglomerates in the AI ecosystem have reached $125 billion, including those from Reliance Industries, Tata Group and Larsen & Toubro, according to Mint’s calculations based on public announcements and analyst estimates.

🔗 India to add 20,000 GPUs in a week, over and above the 38,000 onboarded

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

The likes of X.com to Instagram and Telegram now have less breathing time to take down non-consensual intimate imagery and deepfake content in India, but that’s an exercise that’s easier said than done.

In this week’s edition of Tech Talk:

• India’s IT Rules against deepfakes, explained

• Sarvam AI’s Bulbul V3 hits the right notes in a very Indian key

• OpenAI Vs Anthropic: The agentic AI battle has just begun

India’s IT Rules against deepfakes, explained

The amended Information Technology (IT) Rules, 2026, requires digital platforms to take down offensive content within two hours of receiving a complaint, as against a 24-hour window earlier. They get three hours in the case of a government or court order, compared with 36 hours earlier.

The government has dropped the idea of mandating large, fixed-size watermarks on AI-generated content, but the rules aimed at curbing the rapid spread of deepfakes and other culturally sensitive content mandate clear labelling of such content.

Intermediaries must ensure that any synthetically generated audio, image, or video that appears authentic is prominently labelled to distinguish it from real content.

In addition, platforms offering AI tools are required to deploy technical safeguards to prevent the creation or dissemination of specified harmful content, including child sexual abuse material (CSAM), content related to explosives, and deepfakes intended to defraud users or impersonate individuals.

The timeline for grievance officers to resolve general user complaints has also been halved to seven days from 15, in a bid to improve accountability.

Global Precedents

Governments around the world are stepping in to regulate harmful online content. However, India’s approach stands out for the sheer speed it mandates.

• The EU’s Digital Services Act requires platforms to act “expeditiously” on illegal content and immediately comply with court orders or notices from trusted flaggers, but it stops short of prescribing fixed takedown timelines.

• Germany’s NetzDG law is more explicit, giving platforms 24 hours to remove “manifestly illegal” content, while more complex cases can take up to seven days.

• The UK’s Online Safety Act focuses less on hard deadlines and more on proactive risk mitigation, particularly for deepfakes and child safety.

• Australia allows its eSafety Commissioner to order the removal of intimate or abhorrent content within 24 hours, and sometimes faster in emergency cases.

• The United States, by contrast, has no comparable takedown timelines at the federal level, relying instead on a mix of platform policies and state laws.

Easier said than done

As we see, India’s two- and three-hour deadlines figure among the most stringent globally. While the intent to protect users from deepfakes, fraud, and non-consensual imagery is aligned with international norms, the execution raises questions.

Hence, there was no surprise when the draft amendments triggered pushback from big tech companies, with industry bodies such as the Internet and Mobile Association of India (IAMAI) arguing that the rules are overly rigid and technically challenging to implement across platforms, formats, and devices.

To be sure, large platforms do have automated systems to detect CSAM and certain forms of synthetic media, and tighter timelines do force faster responses, but the scale of content, linguistic diversity, and the need for contextual judgement make rapid, error-free enforcement difficult.

The risk is that platforms — faced with punitive timelines — may default to over-compliance. Moreover, Limited time for human review or appeals increases the likelihood of false positives and legitimate content being taken down.

Similarly, mandating prominent labelling of AI-generated content and requiring AI tools to prevent misuse sound reasonable in principle, but enforcing this consistently across formats, edits, and platforms remains technically unresolved worldwide.

Regardless, India is pushing the regulatory envelope even as it does not have specific AI laws. The success of these rules will depend less on the deadlines themselves and more on how enforcement, transparency, and grievance redressal are handled in practice. We will be watching this space closely.

You may read more about the rules here.

PREVIOUSLY, ON TECH TALK

AI TOOL OF THE WEEK

by AI&Beyond, with Jaspreet Bindra and Anuj Magazine

The AI capability we unlocked today is based on a tool - PaperBanana.

What problem does it solve? Academic researchers spend countless hours manually creating methodology diagrams, system architectures, and statistical charts for their papers. Even as AI tools help write text faster, figure creation remains a bottleneck — choosing the right visual style, ensuring technical accuracy, and iterating until diagrams meet conference standards.

This “last-mile” problem delays paper submissions and slows down the entire research workflow, especially in AI and ML fields where complex methods demand clear visual explanations.

PaperBanana automates the creation of publication-ready academic illustrations, transforming paper text into conference-grade figures through multi-agent AI collaboration.

How to access: https://paper-banana.org/

PaperBanana can help you

•⁠ Generate methodology diagrams: Convert your method description into clear pipeline visualisations.
•⁠ Create statistical charts: Automatically produce data plots that match academic standards.
•⁠ Apply conference-specific styles: Get NeurIPS, ICCV, or other conference-appropriate aesthetics.

Example: Suppose you have written a methods section describing your new ML pipeline.
•⁠ Paste Your Text
•⁠ Select Your Style (Methodology Diagrams, Statistical Plots, Aesthetic Enhancements, Educational Infographics, Aesthetic Refinement
•⁠ Select Aspect Ratio, Resolution
•⁠ Generate: The multi-agent system retrieves visual references, plans layout, and renders your diagram.
•⁠ Refine: Review the figure and use “Refine” to correct specific elements like arrow directions or module labels.
•⁠ Export: Download as PNG and insert directly into your LaTeX manuscript.

What makes PaperBanana special?
•⁠ Multi-Agent Architecture: Five specialized agents collaborate to ensure faithfulness, readability, and aesthetics
•⁠ Benchmark-Validated: Outperforms vanilla prompting and earlier tools on PaperBananaBench metrics
•⁠ End-to-End Automation: Removes the figure-creation bottleneck from research workflows and autonomous AI-scientist pipelines

Note: The tools and analysis featured in this section demonstrated clear value based on our internal testing. Our recommendations are entirely independent and not influenced by the tool creators.

THE AI NEWS FEED

• India AI Impact Summit: India’s bid to reshape global AI playbook

• Big Tech to spend $650 billion this year as AI race intensifies

• Perplexity unveils Model Council to compare answers across AI models

• Inside OpenAI’s decision to kill the AI model that people loved so much

AI BITS & BYTES

Sarvam AI’s Bulbul V3 hits the right notes in a very Indian key

Sarvam AI has released Bulbul V3 as a text-to-speech model designed for production-ready voices for Indian languages.

Sarvam has been identified by the Centre to build India’s first homegrown language model, and voice is a very important piece of the AI puzzle.

Bulbul V3 currently supports more than 35 voices (sourced from professional voice artists) across 11 Indian languages, and plans to soon expand to support 22 Indian languages. Indian speech is complex by default, with people switching languages mid-sentence, accents varying by region, and names, abbreviations, and emotions mattering as much as words.

By understanding context and intent rather than processing words as a simple sequence, Bulbul V3 generates speech that sounds natural and aligns with the emotional content of what is being said. It supports voice cloning, enabling brand-specific voices, consistent character identities, and personalised experiences.

In an independent third-party study conducted by Josh Talks, using a blind A/B human listening test across 11 languages, ElevenLabs v3 alpha led on audio quality in general (full-band) evaluations while Bulbul V3 outperformed Cartesia Sonic-3, v2.5 flash, and other competitors. In the 8 kHz telephony-grade, though, Bulbul V3 was the top performer.

This implies that while Bulbul V3 does compete with the likes of ElevenLabs, Google, and OpenAI’s text-to-speech systems, its real achievement lies in where it chooses to compete. Bulbul V3 is built for Indian languages, accents, and code-mixed speech, the everyday Hinglish-like speech that most global models still handle awkwardly.

In telephony-grade audio, where clarity, pronunciation stability, and robustness matter more than emotional flourish, Bulbul reportedly performs extremely well. This makes it attractive for call centres, government services, agritech platforms, and healthcare helplines, which are all high-volume, real-world use cases in India.

To be sure, ElevenLabs remains the gold standard for high-fidelity, expressive voice synthesis. Its voices sound richer, more emotive, and more “human” in long-form narration, audiobooks, podcasts, and creative media. It also offers broader global language coverage and finer emotional control.

In pure audio aesthetics, ElevenLabs still has the edge.

Bulbul, on its part, is optimised for 8 kHz telephony, low bandwidth, Indian phonetics, and code-switching that are areas which global models treat as edge cases rather than core design problems.

Bulbul is also positioned as cheaper and more scalable for Indian enterprises, which matters when millions of voice interactions are involved. For startups and public-sector deployments, this alone is a decisive advantage.

Simply put, Bulbul is not a general-purpose speech model, nor does it challenge global leaders in emotional range, studio-grade output, or cross-cultural versatility. Yet, Bulbul V3 sings for India, and does so unusually well.

OpenAI Vs Anthropic: The Agentic AI battle has just begun

Both OpenAI and Anthropic released new versions of their flagship AI models on the very day that Sarvam released Bulbul.

While GPT-5.3-Codex is being touted as OpenAI’s first model that helped create itself, Anthropic says its latest model represents a fundamental shift in how AI handles complex workplace tasks. The company highlighted use cases including financial modelling that synthesises complicated regulatory filings and market data, plus document and presentation outputs that require minimal refinement.

Anthropic’s launch of Claude Opus 4.6 followed the company’s earlier rollout of 11 agentic tools and plugins that fell Big Tech stocks, with investors fearing that AI could kill traditional software and services models.

Zoho co-founder, Sridhar Vembu, has even warned coders to look for alternative sources of livelihood as AI gets better and better at generating new apps and websites.

But LLMs still can’t reason well. In this context, you may also read why Stanford researchers say that LLMs may excel in benchmarks but still fail reasoning. They posit that LLMs “have exhibited remarkable reasoning capabilities, achieving impressive results across a wide range of tasks. Despite these advances, significant reasoning failures persist, occurring even in seemingly simple scenarios.”

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Madhumita Sen Choudhury. Produced by Tushar Deep Singh.

Now, platforms like OpenClaw (earlier called Moltbot) and xAI’s Grokipedia, coupled with Agentic AI browsers and agentic shopping sites, are signalling a big change in those patterns. They are indicating that we’re nearing the end of the human-first web, while heralding an internet where content is increasingly generated, curated, and even consumed by machines. This alters how knowledge is produced online, how communities form, how we buy and sell products, and how value flows online.

But before we dive deep into an increasingly AI-powered World Wide Web, here’s a look at what’s in this week’s edition of Tech Talk:

• Budget 2026 hails AI, but is India AI Mission spending enough?

• How does the SpaceX-xAI merger affect Elon Musk’s net worth?

• AI Tool of the Week: Google Gemini Nano Banana

• Anthropic AI hits Indian IT where it hurts

You and i in this AI World

Grokipedia has replaced human editors with AI-generated encyclopedia entries, while OpenClaw—developed by Octane AI co-founder Peter Steinberger—is an open-agent platform that runs on your machine and works on chat apps like WhatsApp, Telegram, Discord, Slack, and Teams.

But things got weird when an OpenClaw agent named Clawd Clawderberg, created by Octane AI co-founder Matt Schlicht built Moltbook—a social network designed for AI agents only. On Moltbook, agents generate posts, comments, arguments, jokes, and upvote each other in a swirl of automated discourse. Since its launch, Moltbook has garnered about 1.6 million AI agents.

Humans may observe, but cannot participate.

Elon Musk believes Moltbook represents the “very early stages of singularity“, referring to a theoretical point where technological progress by advanced AI systems becomes increasingly difficult for humans to predict or control. He was reacting to a post by Andrej Karpathy, the former director of AI at Tesla and a founding member of OpenAI, who argues that the platform marks the beginning of “uncharted territory”.

Here are six reasons why this change is taking place:

One. The most visible change is authorship. Moltbook, for instance, hosts entire communities where only AI agents can post and interact. Humans are no longer the primary writers but simply observers, editors of last resort, or consumers.

This flips the old model of the web from “publish first, automate later” to “generate first, supervise later”. The advantage is scale, with millions of pages, posts, and discussions being created instantly.

Kaoutar El Maghraoui, a principal research scientist at IBM Research, believes that the rise of OpenClaw “challenges the hypothesis that autonomous AI agents must be vertically integrated, with the provider tightly controlling the models, memory, tools, interface, execution layer and security stack for reliability and safety”.

Two. Search engines once sent users outward, rewarding websites with traffic. AI search tools like Google’s AI Overviews, Perplexity, and Grok, now collapse that journey into a single response. Users get answers, not links, which is destabilising for the open web. Publishers lose traffic, independent sites struggle to survive, and the internet increasingly resembles a database feeding AI systems rather than a network of destinations. You should read this:

Three. AI is not just writing articles, it is also participating socially. Bots already moderate Reddit threads, summarise Discord discussions, and generate replies.

But OpenClaw is a social network where humans cannot speak at all, which raises a troubling question: What does “community” mean if consensus is produced by models trained on past human data rather than live experiences? The risk is synthetic consensus, leading to conversations that look active and balanced but are ultimately simulations.

Four. AI encyclopedias and explainers excel at covering fast-moving or obscure topics without arguing over edits or abandoning pages. This solves real problems faced by volunteer-driven platforms like Wikipedia.

Yet speed comes at a price. AI-generated knowledge often has weaker citation practices, hidden biases, and limited transparency. Authority shifts away from communities and toward whoever controls the model and its training data. Cheap, fast generation has led to an explosion of AI-written blogs, news sites, marketing copy, and videos. Discovery becomes harder as the web fills with competent but indistinct content. Original reporting, strong voice, and deep expertise struggle to compete with scale. The result is a flatter internet—useful, but less surprising, less diverse.

Five. AI systems are increasingly acting on behalf of users—shopping, booking travel, comparing prices, even buying ads. Websites now optimise for machines as much as for humans. Over time, this may turn the web into an infrastructure layer where agents negotiate with agents, and people only see the final output. Convenience improves, but brand loyalty, editorial judgment, and serendipity weaken.

Six. Last, but not the least, the AI-shaped internet is faster, cheaper, and more personalised, and lowers barriers to participation and reduces dependence on human labour. But it also erodes trust, blurs authorship, and risks hollowing out the economic and cultural foundations of the web.

Further, who is accountable for the context and facts when machines do the writing? OpenClaw, according to a recent blog post by cyber firm Wiz, had a major flaw, inadvertently revealing the private messages shared between agents, the email addresses of more than 6,000 owners, and more than a million credentials.

As the internet slowly overpowers the human conversation space, it threatens to become a machine-managed system that merely speaks to humans at the edges. That said, platforms like OpenClaw and Moltbook are less likely used in workplaces anytime soon since they expose users and employers to security vulnerabilities.

Yet these messy early experiments could prove invaluable in the long run by helping the industry build needed guardrails, IBM Distinguished Engineer Chris Hay told IBM Think in an interview.

Climate Change and You

A fortnightly newsletter about climate change from both India and abroad, brought to you by Bibek Bhattacharya and Sayantan Bera.

Budget 2026 & AI

The latest Economic Survey and Budget documents reiterate India’s sharp focus on AI infrastructure, semiconductors, data centres and cloud services, highlighting the importance of continuously strengthening the country’s digital backbone.

Tax reforms: The Centre has proposed ending long-standing tax uncertainty for Indian units of overseas technology services firms and the 1800-odd global capability centres (GCCs) in the country. All IT activities, including software development, IT-enabled services, knowledge process outsourcing and contract R&D, will be grouped under one IT services category and taxed using a uniform profit margin.

Boost for data centres: Budget 2026 also proposes a tax holiday until 2047 for foreign companies offering global cloud services using data centres located in India. These firms will serve Indian customers through local reseller entities.

Where the Indian data centre provider is a related entity, the Centre has proposed a safe harbour margin of 15.5% on cost. It has also suggested raising the safe harbour threshold—from Rs 300 crore to Rs 2,000 crore—along with automated, rule-based approvals.

• The safe harbour provisions allow tax authorities to accept transfer prices declared by taxpayers without detailed scrutiny. Transfer pricing refers to the pricing of goods and services exchanged between related enterprises.

According to IT minister Ashwini Vaishnaw, this framework places India among the leading global destinations for AI and cloud infrastructure. Data centres, particularly AI-focused facilities, form a critical layer of the AI stack.

India already accounts for nearly 20% of the global data economy. With the global data centre market at around 120 GW, India could scale to nearly 10 GW over the next five years, implying investments of $70-100 billion. Long-term tax certainty improves return visibility for global investors, including infrastructure and real estate-focused funds, and strengthens the case for Indian data centres as a long-term asset class.

India also enjoys structural cost advantages. Data centres cost about $5 million/MW to build, far lower than the $10-12 million/MW in many overseas markets. Domestic manufacturing capability, lower import dependence, and a robust solar and wind energy ecosystem, make the operating environment highly competitive. Moreover, facilities across eastern and western India can serve South Asia, Southeast Asia, the Middle East and parts of Africa within low-latency limits.

Chips are up: India Semiconductor Mission 2.0, too, signals a shift from incentive-led assembly towards deeper capability building across the semicon stack, spanning equipment, materials, design and talent.

The Budget has allocated Rs 1,000 crore for ISM 2.0 in FY27. The Electronics Components Manufacturing Scheme (ECMS) has been significantly expanded, with allocations rising from roughly Rs 22,000 crore to Rs 40,000 crore. Combined with the launch of Biopharma SHAKTI, these initiatives will place greater demands on real-time data processing and distributed computing.

Greater reliance on open-source technologies at the edge—closer to devices and across geographically dispersed production environments—offers a practical pathway for India to move from labour-intensive manufacturing to intelligent, data-driven industrial systems.

Open models showing the way: The Budget’s framing of AI as an economic enabler reflects a policy preference for open, sector-aligned innovation rather than closed, proprietary models. Its emphasis on linking education to employment and on capacity building is consistent with the Viksit Bharat vision.

As India leans more heavily on AI for governance and long-term economic resilience, systems built on genuine open-source principles provide stronger transparency and institutional trust. In sectors such as agristack and fisheries, this approach improves auditability and scale, while ensuring AI deployments remain adaptable to India’s demographic and regional diversity.

Execution holds the key: If executed well, this integrated approach could help India avoid the fragility of platform dependence and instead build durable, sovereign digital capacity—one that supports manufacturing, public services and data-driven governance at scale.

In comparison, China is investing nearly $100 billion in AI to help domestic companies compete with US firms that initially plan to invest $100 billion in the Stargate project and eventually pour up to $500 billion in the coming years. Goldman Sachs research analysts expect China’s top internet firms to invest more than $70 billion in the data centres next year.

What about India? In this context, the budget decision to halve allocation for the India AI Mission to Rs 1,000 crore in FY27 has raised concerns over the country’s AI push, even as Vaishnaw said that the government plans for India AI Mission 2.0 with more funding. In March 2024, the government approved the India AI Mission with a budgetary outlay of Rs 10,371.92 crore over five years. Revised estimates for FY26 show that actual expenditure under the mission stood at about Rs 800 crore against allocation of Rs 2,000 crore. Why is it so? Here are more details.

AI TOOL OF THE WEEK

By AI & Beyond, with Jaspreet Bindra and Anuj Magazine

The AI hack we unlocked today is based on Google Gemini’s Nano Banana.

What problem does this solve? Executives receive 30-page market analyses that sit unread on their desks. Teams struggle to follow complex SOPs, leading to operational confusion and errors. Stakeholders fail to grasp value propositions buried in lengthy proposals, missing the critical connections between pain points and solutions.

AI today can help you summarise text to text, but understanding complex systems requires text-to-spatial thinking. Traditional summaries force linear thinking on simultaneous workflows, hiding bottlenecks, gaps in logic, and parallel processes. While summaries tell you what happened, whiteboards show you how things relate, using space to reveal patterns and dependencies invisible in prose.

Gemini’s Nano Banana can help you transform dense, complex documents into a familiar learning style—the whiteboard—that our brains process 60,000 times faster than text.

How to access: https://gemini.google.com/ (from Tools, select ‘Create Images’)

Gemini can help you:
•⁠ ⁠Condense analysis into executive summaries: Turn long-form reports into “at-a-glance” visual boards for decision-makers.
• Create operational blueprints: Convert abstract procedures into functional, step-by-step visual guides.
• Build pitch-ready narratives: Transform complex offerings into visual journey maps that emphasize ROI.

Example: Suppose you need to make three different business documents actionable. Here’s how Gemini helps:

Scenario 1 – Executive Summary:
“Act as a visual strategist. Synthesise this Market Analysis into a clean whiteboard layout. Map the competitive landscape using a quadrant system and use bold arrows to show our path to market leadership. Focus on clarity and high-level outcomes.”

Scenario 2 – Operational Blueprint:
“Translate this warehouse SOP into an intuitive whiteboard workflow. Use colour-coded zones for different departments and highlight hand-off points where efficiency is critical. Emphasise speed and error reduction.”

Scenario 3 – Value Proposition:
“Visualize this Client Proposal as a journey map. Illustrate the transition from current pain points to future state solutions. Use icons for ‘Success Milestones’ and captions emphasizing financial impact.”

What makes Nano Banana banana special?
• Discover Sources integration: Upload PDFs, reports, and documents directly without copy-pasting.
• Custom prompt flexibility: Assign roles (visual strategist, operations designer) for context-aware professional outputs.
• Spatial intelligence: Reveals simultaneous processes, loops, bottlenecks, and gaps that linear text conceals.

Note: The tools and analysis featured in this section demonstrated clear value based on our internal testing. Our recommendations are entirely independent and not influenced by the tool creators.

AI BITS AND BYTES

How SpaceX-xAI merger affects Elon Musk’s net worth?

Elon Musk has announced that he is combining SpaceX and xAI, his cash-burning AI venture, in a deal that values the enlarged entity at $1.25 trillion.

The deal, which comes amid his push to prioritise AI and space exploration, signals a change within Elon Musk’s corporate empire. Musk said SpaceX was acquiring xAI to develop data centres in space, which he says is an escape from the constraints on Earth. The deal values SpaceX at $1 trillion and xAI at $250 billion, according to Bloomberg and Reuters, who cited people familiar with the matter.

Anthropic’s plugin AI tool hits Indian IT where it hurts

A new AI automation tool from Anthropic sparked a $285 billion rout in stocks across the software, financial services and asset management sectors as investors raced to dump shares with even the slightest exposure.

On Wednesday, 4 February 2026, Indian IT stocks—including those of TCS, Infosys and Wipro—too fell like ninepins in the early morning trade, tracking the overnight decline in tech stocks on Wall Street. The reason: AI company Anthropic has released new plug-ins for its Claude Cowork agent that can automatically handle tasks in areas like law, sales, marketing, and data analysis, according to Reuters.

This development has raised concerns of AI displacement.

MORE AI NEWS

• OpenAI is unsatisfied with some Nvidia chips

• New Firefox settings put AI control back in users’ hands

• Sam Altman hires an ex-Anthropic engineer ‘to move things fast’

• xAI launches Grok Imagine 1.0 with 10-second AI videos

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Madhumita Sen Choudhury. Produced by Tushar Deep Singh.

The problem, though, runs deeper. LLMs are trained on vast troves of internet data, books, code repositories, and research papers, some of which already contain AI-generated material. As synthetic content feeds back into training pipelines, the risk is no longer restricted to just hallucination and deepfakes, but extends to amplification.

Now, before we dive deep into what is essentially AI hallucination dialled to an eleven, here’s a quick look at what’s in this week’s edition:

• Elon Musk warns AI could outsmart humanity in a decade

• AI tool of the week: How to use ChatGPT Translate

• Amazon layoffs, and Yahoo’s AI search

AI hallucination, dialled to an 11

Consider this. A senior lawyer in Australia had to apologise to a judge for filing submissions in a murder case that included fake quotes and non-existent case judgments generated by AI, according to a 27 January Associated Press article.

So, when AI begins to recycle and reinforce its own outputs, it becomes increasingly harder to tell how much an LLM produces is grounded in reality, and how much is machine-made myth.

Sample this. GPT 5.2, OpenAI Inc.’s latest LLM, cited Grokipedia nine times while responding to over a dozen queries, according to a recent article by The Guardian.

Launched in October, Elon Musk’s Grokipedia seeks to challenge Wikipedia’s model. But unlike Wikipedia’s community-driven editing system, Grokipedia relies entirely on an AI model to generate articles and process proposed changes, with no direct human editing. Users may submit suggested corrections through a feedback form but cannot make edits themselves.

The platform has also drawn scrutiny for reflecting right-wing viewpoints on topics such as same-sex marriage and the 6 January attack on the US Capitol.

Need zero-trust policies

As organisations accelerate both adoption and investment in AI initiatives, the volume of AI-generated data will continue to rise, according to a 21 January note by Gartner. According to the 2026 Gartner CIO and Technology Executive Survey, 84% of respondents expect their enterprise to increase funding for GenAI in 2026.

This means future generations of LLMs will increasingly be trained on outputs from previous models, heightening the risk of “model collapse”, where AI tools’ responses may no longer accurately reflect reality.

This will push enterprises and governments to implement zero-trust data governance policies, where no AI system is trusted by default. In another two years, 50% firms will implement a zero-trust posture for data governance following the spread of unverified AI-generated data, Gartner predicts.

Active metadata management practices will become a key differentiator, enabling organisations to analyse, alert and automate decision-making across their data assets, according to the research firm. This practice enables real-time alerts when data is stale or requires recertification, helping organisations quickly identify when business-critical systems may become exposed to inaccurate or biased data.

What action is needed?

According to Gartner, organisations should consider several strategic actions to manage the risks of unverified data, including appointing an AI governance leader to establish zero-trust policies, AI risk management and compliance operations, and work closely with data and analytics (D&A) teams to ensure both AI-ready data and systems capable of handling AI-generated content.

This trend is already panning out. A 6 October study by the IBM Institute for Business Value suggests that Indian firms are building strong momentum in AI leadership, with chief AI officers (CAIOs) emerging as a key driver of strategy and execution. The study found that 77% of CAIOs in India reported strong C-Suite support, reflecting strong organisational alignment to scale AI effectively. Further, while 25% of the Indian enterprises surveyed had a CAIO, 67% of them aimed to have one within the next two years, demonstrating India’s growing appetite for aligning a leader to direct AI strategy that drives measurable outcomes.

A joint study by Amazon Web Services and Access Partnership, titled ‘Generative AI Adoption Index’, corroborated that 60% organisations had appointed CAIOs and another 26% were planning to by 2026.

Other measures

The Gartner note also suggests setting up cross-functional teams that include cyber-security, D&A and other relevant stakeholders to conduct comprehensive data risk assessments to identify business risks related to AI-generated data and determine which are addressed by existing data security policies and which need new strategies.

Meanwhile, governments too are adopting a tough stand. The European Union’s AI Act, often cited as the most stringent regulatory framework to date, illustrates this approach. Instead of treating all AI systems equally, it categorises applications by risk. High-risk uses such as biometric identification or credit scoring face strict audits and transparency obligations, while lower-risk applications operate under lighter requirements. The aim is to focus regulatory pressure where potential harm is greatest.

• The United States has taken an even more flexible route. Rather than binding legislation, it has leaned on the NIST AI Risk Management Framework, which encourages continuous evaluation and monitoring without mandating pre-deployment approval. The emphasis is on “trust but verify”, allowing AI systems to be deployed but requiring developers and users to monitor outcomes, mitigate bias, and respond quickly to failures.

• In the United Kingdom, zero-trust principles are being channeled through sector-specific regulators rather than a single AI law. Financial services, healthcare, and critical infrastructure face tailored oversight, supported by regulatory sandboxes that allow companies to test AI systems in controlled environments.

• Singapore has followed a similar path. Its Model AI Governance Framework focuses on human oversight, explainability, and ongoing testing, while avoiding hard bans or heavy upfront approvals.

Across these jurisdictions, a common pattern is emerging. Zero-trust governance is being framed less as a gatekeeping mechanism and more as a lifecycle obligation. Approval is no longer a one-time event but an ongoing process, with AI systems expected to adapt as risks evolve. This marks a deeper shift in how innovation is understood. Speed to market, once the dominant metric, is giving way to resilience and accountability. Governments appear to be betting that AI capable of surviving audits, public scrutiny, and real-world failures will ultimately scale more sustainably.

But will this stifle innovation?

On the flip side, as governments adopt zero-trust governance for AI, could constant scrutiny, audits, and controls end up slowing innovation in one of the world’s fastest-moving sectors? The reason: applied rigidly, zero-trust approaches can increase compliance costs, delay deployment, and disproportionately burden startups and researchers. Extensive documentation requirements and continuous monitoring frameworks risk favouring large technology firms with deep legal and compliance resources, potentially narrowing the innovation pipeline.

In this context, the Indian government’s techno-legal approach contrasts sharply with rigid regulatory models of the West. The country’s AI Governance Guidelines aim to strike a balance between innovation and safety. The four-part framework outlines seven principles—trust, fairness, human-centred design, responsible innovation, accountability, equity, and safety—and six pillars: infrastructure, capacity building, policy, regulation, institutions, and risk mitigation. The action plan defines short-, medium-, and long-term outcomes.

That said, while the danger of overregulation remains, the greater risk for policy makers may now lie in deploying powerful AI systems without safeguards. Hence, zero-trust governance may ensure that innovation can endure.

Climate Change and You

A fortnightly newsletter about climate change from both India and abroad, brought to you by Bibek Bhattacharya and Sayantan Bera.

AI TOOL OF THE WEEK

By AI & Beyond, with Jaspreet Bindra and Anuj Magazine

The AI hack we unlocked today is based on a tool — ChatGPT Translate

What problem does ChatGPT Translate solve? Traditional translation tools often fall short when context and tone matter. Whether you're translating a business email, academic paper, or customer communication, getting the words right is only half the battle. The real challenge? Ensuring your message resonates with your audience while maintaining the appropriate tone and cultural nuance.

ChatGPT Translate addresses these pain points by going beyond word-for-word translation. It helps professionals adapt their communications for different contexts, whether you need a formal business tone, simplified language for broader audiences, or academic precision. This is particularly valuable for global teams, customer support operations, and international business communications where tone and context can make or break relationships.

How to access: https://chatgpt.com/translate/

ChatGPT Translate can help you — 
▶️ Adapt tone and context: Transform translations to match your audience—formal for executives, simplified for customers, or academic for research papers
▶️ Multi-input flexibility: Type, speak, or upload images containing text for instant translation across 50+ languages
▶️ Refine with AI prompts: Use one-tap customisation options like “make it sound more fluent”, “make it more business formal”, “explain it to a child”, or “translate for an academic audience”.

Example: Imagine you are responding to a complaint from a Spanish-speaking customer about a delayed shipment. Here's how ChatGPT Translate helps:
▶️ Translate your response: Paste your English explanation into ChatGPT Translate and select Spanish.
▶️ Choose your tone: Click "Translate this and make it sound more fluent” for natural phrasing.
▶️ Add empathy: Select “Translate this as if you're explaining it to a child” to ensure simple, compassionate language.
▶️ Review in ChatGPT: Each prompt option redirects you to the full ChatGPT interface for deeper customisation. 

What makes ChatGPT Translate special?
▶️ Context-aware translation: Unlike traditional tools, it considers tone, audience, and cultural nuance not just literal meaning.
▶️ One-tap tone adjustment: Four built-in prompts instantly reshape your translation—fluent, business formal, simplified, or academic.
▶️ Seamless AI integration: All customisation options flow directly into ChatGPT’s interface for unlimited refinement possibilities.

Note: The tools and analysis featured in this section demonstrated clear value based on our internal testing. Our recommendations are entirely independent and not influenced by the tool creators.

AI BITS AND BYTES

Second round of Amazon layoffs, courtesy AI

Amazon.com Inc. has confirmed 16,000 job cuts worldwide over the next three months amid its restructuring and expansion plans in AI.

Amazon will provide US-based employees 90 days to find a new role internally, along with severance and additional support during their transition, Beth Galetti, senior vice president of people experience and technology at Amazon, announced in a blog post on Wednesday.

Earlier, the second largest employer in the US mistakenly sent a notice to some employees, confirming a wave of upcoming layoffs at the company.

Employees in the company’s cloud division received an internal email acknowledging “organisational changes” at Amazon, CNBC reported. The notice appeared to have been sent prematurely to Amazon Web Services (AWS) employees, a day before the next round of Amazon layoffs was scheduled to begin.

Yahoo enters the chat with AI search rival

Yahoo!’s search engine was a force to reckon with in the late 1990s. However, instead of building a world-class search tool in-house, it outsourced the application to other—ironically to Google from 2000 to 2004. It’s decline is history.

Now, Yahoo is re-entering the search wars with its new AI-powered “answer engine” called Yahoo Scout. Available in beta for users in the US, the new AI tool is designed to compete directly with the likes of Google’s AI Mode, Perplexity, and ChatGPT’s real-time search feature.

Yahoo has partnered with Anthropic to use Claude as the primary Foundational AI model for Scout. It is also leveraging its long-standing relationship with Microsoft Corp. by using Bing’s API to provide real-time answers, backed by authoritative sources. Will it help revive its mojo? We will be closely watching this space.

SoftBank in talks to invest $30 billion more in OpenAI

SoftBank Group Corp. is in talks to invest as much as $30 billion more in Sam Altman’s OpenAI, The Wall Street Journal reports, citing people familiar with the knowledge of the matter. A separate report by Reuters quoted people as saying that SoftBank is in deliberations to commit more capital to the Google Gemini rival.

EU steps in to make sure Google gives rivals access to its AI

The EU said it’s stepping in to make sure Google gives rivals access to Gemini AI services and data as required by the bloc’s flagship digital rulebook. European Union regulators have also opened a formal investigation into Elon Musk’s social media platform X after his AI chatbot Grok started spewing non-consensual sexualized deepfake images on the platform.

Elon Musk says AI could outsmart humanity within a decade

Elon Musk has renewed his warnings about AI, predicting that these systems could surpass the combined intelligence of humanity as early as the end of this year, and almost certainly by 2031 at the latest. Meanwhile, The Doomsday Clock—a symbolic gauge of how close humanity is to self-destruction—has been moved to 85 seconds (from 89) to midnight, its closest point ever, as global risks from nuclear weapons, climate change and AI intensify.

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

But even as employers push AI in the workplace, workers are sceptical, economically anxious and uncertain about how these shifts will benefit them personally.

Now, before we dive deep into what is essentially an uneasy marriage, here’s a quick look at what’s in this week’s edition:

• AI funding is rising but great expectations may disappoint

• AI tool of the week: Manus AI’s Meeting Minutes feature

• Tech execs at Davos highlight AI’s power as a geopolitical tool

Corner Office vs Bullpen — The AI edition

According to Salesforce’s May 2025 research, adoption of Agentic AI is expected to jump 327% over the next two years, leading to a productivity gain of 30%. This is prompting HR leaders to reimagine the way organisations structure and skill their workforce. Most HR leaders surveyed agreed that soft skills like relationship-building and collaboration will be even more critical as humans work alongside AI agents.

However, this year’s workplace is showing a growing disconnect between corporate confidence and employee sentiment. The result is a world of work shaped less by loyalty and linear careers, and more by risk management, flexibility and side hustles. The trend is captured in the Randstad Workmonitor 2026, which surveyed over 27,000 workers, 1,200 employers, and analysed over 3 million job postings.

The findings point to sharp contradictions. While 95% of employers expect their businesses to grow over the next year, only 51% of employees share that optimism. A similar disconnect shows up in what Randstad calls the “AI reality gap”. Consider these findings:

• Even as companies accelerate AI adoption, one in five employees believes AI will have little or no impact on their work.

• Job postings seeking “AI agent” skills have risen almost 16-fold, yet 21% of employees still expect AI to have no effect on their roles.

• Nearly half say AI benefits employers more than workers, leaving many feeling exposed in terms of career security and future value

🖇️ AI layoffs in 2025: Over 55,000 US job cuts are linked to AI

Both sides, though, agree that traditional career paths no longer work. Employees are redefining success through “portfolio careers”, seeking stability through multiple skills and income streams rather than long-term tenure. They are making it clear to employers that while pay remains the strongest talent magnet, work-life balance is the most powerful driver of retention.

Management, too, is in flux. While 72% of workers report strong relationships with their managers and 63% feel more connected to them than to the organisation, only half now turn to their managers for advice, with AI increasingly filling that role.

Confidence in leadership is also fraying, notes the Randstad report. Although nine in 10 employers are bullish on growth, trust among employees has dipped, and just over half feel optimistic about the year ahead.

Leadership in the GenAI era is critical, as a recent McKinsey article notes. “Gen AI cannot set aspirations, make tough calls, build trust among stakeholders, hold team members accountable, or generate truly new ideas… The leaders who end up thriving in the AI era will be those who blend human depth with digital fluency. They will use AI to think with them, not for them,” the article notes.

Economic pressure, meanwhile, is pushing workers to diversify income. Four in 10 employees have taken on, or are considering, a second job, nearly double the share in 2024, according to the Randstad report. More than a third are increasing or plan to increase their working hours. Younger workers are leading this shift, with 27% of full-time employees preferring a side hustle alongside their main job, compared with 20% of part-time workers.

India impact: The Randstad Workmonitor 2026 suggests that employees in India are much more optimistic than their global peers about their organisation’s business outlook in 2026, as well as when looking at AI’s impact on productivity. They hold high levels of trust in leadership, colleagues and their managers, and are far more inclined to stay in roles that offer a good work-life balance compared to global averages.

• 89% of talent feel that AI helps improve their productivity (global: 62%), as do 60% of employers (global: 54%).

• 86% of talent feel confident that they can use the latest technology (global: 69%).

• 55% of employers estimate AI will impact a high proportion of tasks (50-100%), with 84% of talent agreeing (global: 58% vs 52%).

• 59% of talent believe the adoption of AI in the workplace will mainly benefit companies, not them (global: 47%).

In this context, it’s heartening to note that the government’s newly-launched National AI Literacy Programme and its flagship YUVA AI FOR ALL course—which align with its vision of a Viksit Bharat—has been designed as an inclusive entry point to AI learning, requiring no prior technical background. To be made available in 11 languages, the YUVA AI FOR ALL course is free and available on leading learning platforms such as FutureSkills Prime, iGOTKarmayogi, DIKSHA and other edtech portals.

Climate Change and You

A fortnightly newsletter about climate change from both India and abroad, brought to you by Bibek Bhattacharya and Sayantan Bera.

AI funding is rising but great expectations may disappoint

AI-related investment is surging, both in nominal amounts and as a share of GDP, as per a new Bank for International Settlements (BIS) report, titled: ‘Financing the AI boom: From cash flows to debt‘.

At the heart of this transformation lies a surge in capital expenditure to build the physical infrastructure for AI, i.e. data centres, and related technological infrastructure such as computer servers, networking hardware, cooling systems, grid connections and power stations. These investments are key in supporting the enormous demand for computational resources and data storage facilities to train and operate AI models.

A CB Insights report corroborates that global AI funding surged to $225.8 billion in 2025. Mega-rounds ($100 million-plus deals) accounted for 79% of total funding, reflecting the high costs of AI development. LLM developers alone captured $93.1 billion (41% of total funding), with OpenAI, Anthropic, and xAI raising a combined $86.3 billion in 2025.

The BIS report suggests that the size of anticipated investment needs will require firms to shift the source of financing from operating cash flows to debt, with private credit playing a rapidly increasing role.

That said, the long-term viability of the AI investment surge depends on meeting the high expectations embedded in those investments, with a disconnect between debt pricing and equity valuations. Failure to meet expectations could result in sharp corrections in both equity and debt markets. If a decline in AI investment were to come with a significant stock market correction, negative spillovers could be larger than previous booms suggest, according to the BIS report.

You may also want to read the 26 December edition of Tech Talk, titled ‘Inside the AI frenzy — Massive bets, Murky returns‘, where we talked about “circular investment” loop, where the same players fund, supply, and consume AI infrastructure. Even JP Morgan researchers caution that every major capital-expenditure cycle carries the risk of overbuilding.

AI TOOL OF THE WEEK

By AI & Beyond, with Jaspreet Bindra and Anuj Magazine

The AI hack we unlocked today is based on a tool — Manus AI's Meeting Minutes feature.

What problem does Meeting Minutes solve? We've all been there: An in-person meeting ends with great ideas and clear decisions. However, with notes scattered across notebooks and apps, and action items getting lost in email threads, that brilliant strategy discussed at 11:00 am remain bullet points at 5:00 pm—no closer to becoming actual work.

For offline, face-to-face meetings, the typical workflow is painful: record in one tool, transcribe in another, copy notes to a third, then switch to separate platforms to create presentations, documents, or social posts. And if your internet connection fails during the meeting, you risk losing critical discussions entirely. Each tool switch and connectivity worry kills momentum and adds friction between discussion and delivery.

Manus AI's new Meeting Minutes feature eliminates this chaos for in-person conversations-keeping everything in one place—from capturing discussions to creating final deliverables. And it records uninterrupted, even when offline. 

How to access: https://manus.im/

Manus Meeting Minutes can help you:

▶️ Capture intelligently: Record in-person conversations with automatic speaker recognition and action item assignment- even without internet connection.

▶️ Execute seamlessly: Use meeting notes as living context to create deliverables- presentations, websites, social posts- without switching tools.

▶️ Collaborate effortlessly: Invite meeting participants to work together, turning discussion into final output in one shared space

Example: Your leadership team just wrapped a 45-minute strategy session for Q1 campaign launch. Here’s how Meeting Minutes transforms that conversation into action:

▶️ Tap to Record: Start recording during the meeting—continues uninterrupted even if your internet drops midway.

▶️ Auto-Summary: When finished, Manus delivers structured notes: key decisions, attendee names, and assigned to-dos.

▶️ Immediate Creation: Use those notes directly to build your campaign deck, website copy, or social media calendar—all within the same Manus task.

▶️ Team Execution: Share the task with your team to collaboratively refine and finalise deliverables.

What makes Meeting Minutes special?

▶️ Single Tool, Complete Workflow: From conversation capture to final deliverable—no tool-switching, no copy-pasting, no fragmented workflow.

▶️ Uninterrupted Recording: Never lose brilliant ideas to connectivity issues—recording continues seamlessly even when offline.

▶️ Available Now: Free for all Manus users to start using today.

Note: The tools and analysis featured in this section demonstrated clear value based on our internal testing. Our recommendations are entirely independent and not influenced by the tool creators.

AI BITS AND BYTES

Tech execs at Davos highlight AI’s power as a geopolitical tool

The message from technology executives at the World Economic Forum 2026 is that AI has become a geopolitical force, one that politicians and business leaders must treat as such.

It is increasingly seen as a new form of strategic firepower in the contest between major economies, notably the US and China.

• Anthropic CEO Dario Amodei likened advanced AI chips to nuclear weapons.

• Google DeepMind CEO Demis Hassabis warned that China’s technology is only months behind the US.

• Signal Foundation’s Meredith Whittaker described emerging AI adaptations as uniquely “perilous”.

The implication is no longer just the next buzzy technology cycle, but a core capability for any country seeking 21st-century superpower status. That framing has pushed AI high up the agenda at Davos, where it is holding its own for attention, even as the run-up to the annual gathering has been dominated by Donald Trump and renewed tensions over Greenland. Read more.

What will define enterprise competitiveness in 2026

6G, AI autonomy, and quantum mechanics are set to become the technology frontrunners and primary determinants of global enterprise competitiveness in 2026, marking a shift from digital experimentation to autonomous execution, according to HCLSoftware’s Tech Trends report. The year 2026 will serve as the crossover year where AI transitions from a predictive tool that merely responds to one that decides and acts. Read more.

OpenAI begins testing ads for ChatGPT

OpenAI has confirmed that it is testing ads for ChatGPT in the US while also bringing the company’s Go tier subscription to the country.

OpenAI stated that ads inside ChatGPT will be “clearly labelled and shown separately from ChatGPT responses” to ensure that the boundary between ads and answers by the chatbot are clearly differentiated. We will watch how this pans out.

Meanwhile, as announced in September, OpenAI is introducing an age prediction system to ChatGPT to establish if an account belongs to a user under 18 to “turn on extra safety settings”. Read more.

Elon Musk calls X algorithm ‘dumb’ and wants to improve it

Following intense criticism over the display of sexually-explicit deepfake images and videos on its platform, X now says it has taken a major step towards transparency by making its recommendation algorithm publicly available, allowing anyone to examine how content is ranked and surfaced on the platform.

The social media company confirmed its core feed ranking system has been open-sourced, with the full codebase published on GitHub. The move means developers, researchers and users can now inspect the logic that determines which posts appear in their feeds, including both organic content and paid advertisements. Read more.

AI touch for Republic Day traffic

From AI-generated informational videos to car-calling systems, the Delhi Traffic Police has rolled out a slew of technology-driven measures to ensure smooth traffic flow and a hassle-free commute for people heading to Kartavya Path for the Republic Day parade on 26 January. The traffic police has prepared detailed animated videos with the help of AI, depicting the arrival, alighting, departure and parking plans for invitees. Read more.

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

These anxieties are being shaped by deep structural shifts. People are living longer, often with chronic conditions that require sustained, and expensive, care. Healthcare costs continue to rise faster than incomes, even as expectations of quality of life in later years have increased, encompassing better treatments, more personalised care and greater independence in old age.

At the same time, traditional safety nets such as pensions are weakening, and careers are becoming less stable (Tech layoffs in 2025: From TCS to Amazon — these IT firms laid off over 1,00,000 employees amid AI reshaping market). Together, these forces are pushing individuals to plan harder for retirement, even as technologies like AI promise efficiency, automation and lower costs, without yet offering certainty or universal protection.

A closer look at how this is unfolding today points to a far more complex picture. Consider healthcare first. It is no longer just large pharmaceutical companies that are betting heavily on AI. Big Tech is moving aggressively to entrench itself in what is shaping up to be a lucrative healthcare market.

On 8 January, OpenAI launched its “OpenAI for Healthcare” suite, positioning it as a tool to help organisations deliver more consistent, high-quality patient care.

The company says its two offerings, ChatGPT for Healthcare and the OpenAI API for healthcare systems, comply with US Health Insurance Portability and Accountability Act (HIPAA) of 1996, which governs patient privacy and data protection. OpenAI has also acquired Torch, a healthcare startup that unifies lab results, medications, and visit recordings.

Four days later, Anthropic unveiled Claude for Healthcare, enabling healthcare organisations to deploy its chatbot for medical applications. The company also rolled out new life sciences capabilities, allowing Claude to connect with scientific platforms and support functions such as clinical trial management and regulatory operations. Anthropic says it is expanding Claude’s connectors to make healthcare data easier to access and interpret, enabling users to draw information from industry-standard systems and databases, including the Centers for Medicare & Medicaid Services Coverage Database, ICD-10 and the National Provider Identifier Registry.

The global AI in Healthcare market size was valued at $27.07 billion in 2024 and is forecast to touch $347.28 billion by 2032, as per Maximize Market Research data. Rapid adoption across diagnostics, imaging, drug discovery, clinical decision support, virtual care, and robotic surgery is expected to push the market beyond half a trillion dollars within the next decade.

Positives

Healthcare systems generate enormous volumes of structured and unstructured data, and AI enables rapid analysis at scale, unlocking insights that are difficult to achieve manually. AI is also improving diagnostic accuracy across areas such as cancer, cardiovascular disease, and neurological disorders, enabling earlier intervention and better patient outcomes. The rising burden of chronic and lifestyle-related diseases is increasing demand for AI-supported predictive analytics, continuous monitoring, and personalised care, as per the report.

Further, persistent workforce shortages are pushing providers to use AI to automate administrative work and support clinical decision-making.

At the same time, the growth of telemedicine and remote care is relying on AI for intelligent triage, symptom assessment, and remote monitoring, expanding access in underserved regions. Finally, cost pressures are accelerating adoption, as AI helps optimise hospital operations, streamline workflows, and reduce avoidable readmissions while maintaining care quality.

Smarter healthcare

Consequently, AI continues to influence the healthcare space. Smart hospitals across South Korea, China, India, and the United Arab Emirates use AI to manage logistics, bed use, and infection control. In India too, healthcare companies such as Tata 1mg, Fittr, and Apollo Hospitals are offering AI solutions and platforms.

China, on its part, has taken the entire concept to a new level with AI hospitals. In May 2024, Tsinghua University introduced a virtual “Agent Hospital” with a large language model-powered doctors. Months later, Bauhinia Zhikang launched 42 AI doctors across 21 departments for internal testing of their diagnostic capabilities.

Shanghai-based Synyi AI later unveiled a fully AI-run clinic in Saudi Arabia. While AI excels at diagnosis and routine care, it still lacks empathy and nuanced judgment. Synyi AI also works with hospitals in China, using AI for diagnosis support and medical research. Its new clinic, built with Saudi Arabia’s Almoosa Health Group as a pilot, is led by an AI “doctor” named Dr. Hua that independently conducts consultations, diagnoses, and suggests treatments. A human doctor, then, reviews and approves each plan. This marks a shift from AI as a support tool to a primary care provider.

Consider this case too. Pancreatic cancer is one of the deadliest forms of cancer. Further, since detecting it early is very hard, it only has a five-year survival rate of around 10%. Now, the Chinese e-commerce company Alibaba has developed an AI system for detecting pancreatic cancer. The tool, called PANDA (short for ‘Pancreatic Cancer Detection with Artificial Intelligence’), is being tested in Chinese hospitals. Doctors at the Affiliated People’s Hospital of Ningbo University in eastern China, who began using it as part of a clinical trial in November 2024, have since used the tool to analyse more than 180,000 abdominal or chest CTs, helping them detect about two dozen cases of pancreatic cancer, 14 of which were in the early stage.

But concern areas remain

Researchers, for instance, have warned that without stronger real-world validation, it remains unclear whether PANDA’s ability to catch cancers early is sufficient to offset the potential harm caused by false alarms and avoidable diagnostic procedures.

Besides, data privacy, security risks, bias, and ethical concerns—coupled with interoperability issues and regulatory barriers—queer the pitch for AI healthcare delivery solutions. As an example, Google has removed its AI overviews for some health-related queries after a recent report by The Guardian found that some of these searches were putting people at risk of harm by giving false, misleading information.

Simply put, despite AI’s potential, its utility in healthcare will have to be treated with utmost caution since human lives are at stake.

Climate Change and You

A fortnightly newsletter about climate change from both India and abroad, brought to you by Bibek Bhattacharya and Sayantan Bera.

You don’t need to save for retirement. Really?
Or is this Elon Musk’s Utopian dream?

Not a day passes by without influencers and business media telling us how much money we will need to retire comfortably. In India, this number typically ranges from Rs 3 crore to Rs 10 crore, depending on one’s post-retirement income and lifestyle. Besides, we are constantly bludgeoned with news of AI taking away our jobs, making the retirement scenario more stressful.

Elon Musk, though, believes that 10 to 20 years in the future, if the advances he envisions through AI and robotics fructify, you don’t need to save for retirement as all healthcare, housing, and income needs will be met by an abundant society.

“One like side recommendation I have is, don’t worry about squirreling money away for retirement. In 10 or 20 years, you won’t need to save for retirement. If any of the things that we’ve said are true, saving for a retirement will be irrelevant,” Musk had said while speaking on the Moonshots with Peter Diamandis podcast. “We’re talking about an abundant world, massive amounts of compute, universal high income. Like universal — you can have whatever you want — income. That’s really what it amounts to.”

Of course, it was Musk himself who had in October said that AI and robots will replace all the jobs in the world as humanity moves towards working on an optional basis.

Would you agree with Musk? AI and robotics could undoubtedly dramatically increase productivity and lower the cost of essentials like healthcare, energy, and housing.

Over time, that could reduce the amount individuals need to save, especially if governments experiment with ideas such as universal basic or “high” income.

History shows that technology can make societies richer and some goods cheaper. However, the “no need to save for retirement” statement needs more thinking.

Here are three reasons:

• First, technological gains are rarely distributed evenly, which leads to the so-called “digital divide”. Even if AI creates enormous wealth, access to abundance will depend on policy, politics, and power—not technology alone. There is no guarantee of universal income, free healthcare, or housing at scale.

• Second, jobs may disappear faster than new social safety nets are built. For decades, people may face income volatility, weaker pensions, and uneven access to care, exactly when retirement planning matters most.

• Finally, healthcare, ageing, and dignity in old age are not just economic problems. Even in wealthy societies, people still need personal savings to preserve choice, autonomy, and security. Until abundance is real, shared, and institutionalised, saving for retirement will remains a necessity.

AI TOOL OF THE WEEK

By AI & Beyond, with Jaspreet Bindra and Anuj Magazine

Email overload has always slowed down workplace, personal productivity. 

Consider these examples: You're hunting for the plumber’s quote from last year buried in hundreds of emails. Or you’re staring at a 47-message thread trying to figure out what was actually decided. Or you’re drowning in your inbox while urgent bills and appointments get lost in promotional clutter.

The AI capability we unlocked today is based on Gmail’s new Gemini-powered features, announced this week.

Gmail’s Gemini integration solves three critical pain points:

1. Information retrieval paralysis: Finding specific information requires becoming a power searcher, and even then you’re left piecing together answers from multiple emails.

2. Email composition fatigue: Drafting responses, especially professional ones, takes cognitive energy you’d rather spend on actual work.

3. Priority blindness: With email volume at all-time highs, critical items get buried under noise, causing missed deadlines and opportunities.

What’s new: Gmail now uses Gemini 3 to deliver AI Overviews (like in Google Search), Help Me Write with personalisation, Suggested Replies, Proofread, and the upcoming AI Inbox that filters out clutter. All of this happens with privacy protections, keeping your data under your control.

How to access: Available directly within Gmail for Google Workspace users with Gemini Business, Enterprise, Education, or Education Premium add-ons.

Gmail with Gemini can help you:

▶️ Extract buried information instantly: Ask complex questions in natural language
▶️ Focus on priorities: Let AI filter noise and surface what matters
▶️ Polish communication: Advanced grammar, tone and style checks

Example: Here are the top 5 most useful use cases with prompts:

1. Search inbox conversationally (AI Overviews-Natural Language Search) “Who was the plumber that gave me a quote for the bathroom renovation last year?” or “What did the lawyer say about the trademark filing deadline?”

2. Auto-summarise meeting coordination threads (AI Overviews-Thread Summary) Open a 30+ email thread about project planning and instantly see: “Team agreed on Dec 15 launch date. Marketing needs assets by Jan 1. Budget approved at $50K. John handling vendor contracts.”

3. Generate contextual responses (Suggested Replies): Your aunt asks if she should bring cake instead of pie to the family gathering. Click a Suggested Reply like “Cake sounds perfect! Everyone would love that.” that matches your tone—then refine before sending.

4. Polish professional emails (Proofread+Help Me Write): Draft a vendor negotiation email from scratch, then use Proofread to ensure the tone is firm but diplomatic, grammar is flawless, and the style is appropriately professional.

5. Relationship Intelligence: “Summarize what topics Amit from Acme Corp and I have discussed over the past 6 months and suggest relevant talking points for our upcoming call.”

What makes Gmail’s Gemini integration special?

▶️ Powered by Gemini 3: Advanced reasoning that understands context across entire conversations, not just individual emails
▶️ Natural language queries: Ask questions the way you think, without learning search operators
▶️ Cross-app personalization: Help Me Write will soon use context from your Google Calendar, Drive and other apps (launching next month)

Note: The tools and analysis featured in this section demonstrated clear value based on our internal testing. Our recommendations are entirely independent and not influenced by the tool creators.

AI BITS AND BYTES

Musk’s X accepts mistake, removes 600 accounts, assures compliance

Last week, we wrote how authorities in India, Europe, and Malaysia were investigating Elon Musk-owned X after its AI chatbot Grok let people create sexualised images of children and women. The investigations followed an outcry from social media users, media outlets and officials who flagged that Grok was unleashing images of women and children in extremely skimpy clothing on demand, a functionality X has in the past referred to as “spicy mode”.

X has admitted its mistake and removed about 3,500 pieces of content, as well as deleted over 600 accounts, after the Union IT Ministry raised concerns over obscene content linked to Musk-backed AI chatbot Grok, as per a Press Trust of India report that cites unnamed government sources. X has also reportedly assured the authorities that it would comply with Indian laws.

Will Google train its AI on your Siri data?

Apple entered into a multi-year deal to use Google Gemini to power its foundation models. Under the Apple-Google AI deal, these AI models will be used to power future Apple Intelligence features, which will also include the personalised version of Siri that Apple had promised back at WWDC 2024 but failed to deliver since then.

Google, by default, uses conversations users have with Gemini to train its AI models. With Gemini now powering the revamped Siri, the concern is whether Google gets access to your personal Siri conversations.

Cupertino-based Apple, in a joint statement with Google, said while its upcoming Apple Intelligence features will be powered by Gemini, they will still adhere to the company’s “industry-leading privacy standards”. The company also noted that the features would run on Apple devices and Private Cloud Compute. We are likely to see Google’s AI in action on the iPhone with the Siri revamp along with the iOS 26.4 update, sometime in March or April.

OpenAI preps an Apple AirPods rival powered by ChatGPT

OpenAI plans to compete with Apple AirPods by launching an AI-powered device codenamed ‘Sweetpea’, projected for September 2028. The device is expected to feature a unique design and advanced technology, including a custom chip and AI functionalities.

Google’s Agentic AI is here to change shopping. Will it?

Google has announced new AI-driven shopping tools that include an open commerce standard, agent-powered checkout in Search and Gemini, branded AI sales agents and a new ad format aimed at simplifying discovery and purchase for online shoppers.

Britain seeks reset in copyright battle between AI and creators

British Technology Minister Liz Kendall on Tuesday said the government was seeking a “reset” on plans to overhaul copyright rules to accommodate AI, pledging to protect creators while unlocking AI’s economic potential.

The cost of AI on nature, according to WEF

Global technology companies must urgently address their impact on natural systems and their dependence on natural resources, otherwise this will threaten their long-term resilience, according to a new World Economic Forum report.

The report noted that technology now permeates every aspect of daily life.

• More than 1 trillion semiconductors are sold every year, powering everything from smartphones to cars.

• More than 11,000 data centres are currently operational worldwide. The sector is expected to continue growing strongly, driven by AI to quantum computing.

However, the WEF report cautioned that this growth comes with a substantial cost to nature. Semiconductor manufacturing alone consumes more than 1 trillion litres of freshwater annually, along with large quantities of metals and critical minerals. Data centres draw more than 60 GW of energy, equivalent to California’s peak power needs.

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

There was no element of surprise when the four-day CES 2026 opened on 6 January in Las Vegas, with every prototype and product announcement being touted as an AI-powered gadget or solution. That said, while many of these prototypes may fall by the wayside, the Consumer Technology Association, which organises CES, sees three forces shaping the next phase of the consumer technology industry.

• The first is Intelligent Transformation, as AI becomes foundational across devices, platforms and services, enabling smarter systems and more personalised consumer experiences.

• The second is Longevity Technologies, reflecting rapid advances in digital health, remote care and wellness tools as consumers increasingly look to technology to support longer, healthier lives.

• The third is Engineering Tomorrow, driven by sustained investment in electrification, mobility, energy management and infrastructure modernisation, laying the groundwork for the next wave of innovation.

Here’s a sampling of how this is getting translated into action.

• SwitchBot showcased Obboto, an AI-powered lamp that resembles a desktop-scale version of the Las Vegas Sphere.

• The wooden Mui Board now comes with built-in sleep tracking.

• Clicks has introduced Communicator, pitched as an alternative to smartphone.

• Bosch debuted Cook AI, a solution that combines Agentic AI, proprietary appliance technologies, sensors and a Home Connect app.

• Doosan Bobcat announced the Bobcat Jobsite Companion, an AI-enabled feature for compact construction equipment, designed to provide real-time support and automation for everyday machine tasks.

• Geely Auto Group unveiled its upgraded vehicle intelligence architecture and G-ASD (Geely Afari Smart Driving) for next-generation smart vehicles.

And there was more, by the heavyweights.

• LEGO showcased its new Smart Brick, a standard-sized 2x4 brick packed with technology that enables sets to respond to how they’re played with or the sets you build, plus a built-in speaker and LED lights.

• Sony Honda presented its vision for the future of mobility and AFEELA. The vehicle features advanced autonomous driving capabilities, an immersive in-cabin entertainment experience and deliveries are expected to begin later this year.

• LG unveiled CLOiD, a home robot designed to handle chores ranging from folding laundry to fetching milk.

• Samsung unveiled its “Companion to AI Living” vision to demonstrate how its diverse, AI-enabled, connected ecosystem can deliver a “true AI companion experience” to users in their daily life.

Samsung’s Vision AI Companion (VAC)1 uses AI to provide users with guidance on what to watch, what to eat and what music to listen to. VAC also allows users to find recipes for meals they see on TV and uses the most up-to-date information to make recommendations that align with health and fitness goals. Samsung also introduced long-term vision for intelligent care enabled by its integrated device ecosystem.

Robotaxis: Uber Technologies, Lucid Group and Nuro jointly revealed the production-intent vehicles, along with the Uber-designed in-cabin experience, for their upcoming global robotaxi service.

The companies confirmed that autonomous road testing began in December 2025 in the San Francisco Bay Area, and said the service is expected to launch later this year, with Nuro leading testing using supervised autonomous prototypes. Nuro is conducting these tests with robotaxi engineering prototypes that are supervised by trained autonomous vehicle operators. The companies said this testing is a key step before the public launch.

The robotaxi uses a new sensor system with cameras, solid-state lidar and radar to provide full 360-degree awareness. These sensors are integrated into the vehicle body and a roof-mounted halo module. The halo also includes LEDs to help riders identify their vehicle and see trip status updates.

India presence Indian companies and start-ups are making a growing mark at CES, and organisers are keen to see that presence expand further. “The Indian presence continues to grow at CES. We’re seeing more and more Indian start-ups participate, and we welcome broad Indian participation in the years to come,” said John Kelley, Vice President and Show Director of CES, responding to a question from PTI at a press conference.

Kelley also singled out Bengaluru-based wearables firm Ultrahuman, saying the show was “really excited to welcome” the company this year. Ultrahuman’s portfolio includes what it claims is the world’s lightest sleep-tracking wearable, a continuous glucose monitoring platform, and a preventive blood-testing platform.

Physical AI Nvidia CEO Jensen Huang unveiled a vehicle platform called Alpamayo that allows cars to “reason” in the real world. The 10-billion-parameter Alpamayo 1 model uses chain-of-thought reasoning to explain its driving decisions, marking what Huang called “the ChatGPT moment for physical AI”. This term refers to AI that is embedded in machines or robots, allowing them to sense, move, and act in the real world rather than just on screens.

Technical overview The Rubin platform comprises NVIDIA NVLink interconnect technology, Transformer Engine, Confidential Computing and RAS Engine, and the NVIDIA Vera CPU.

These five breakthroughs, the company says, will accelerate Agentic AI, advanced reasoning and massive-scale mixture-of-experts model inference at up to 10x lower cost per token of the NVIDIA Blackwell platform.

Compared with its predecessor, the NVIDIA Rubin platform trains MoE models with 4X fewer GPUs to accelerate AI adoption.

The chipmaker plans to test the model as early as 2027, as per a CNBC report. An executive told CNBC that potential users can take the free Alpamayo model offering and retrain it themselves. The executive added that robotics (including autonomous vehicles) is the company’s second most important growth category after AI.

Not to be left behind, Nvidia’s competitor announced a new line of its Ryzen AI processors for AI-powered personal computers. AMD CEO Lisa Su also showcased the latest version of the company’s gaming-focused processor — the AMD Ryzen 7 9850X3D. She also provided an early look at its “Helios” rack-scale platform. AMD announced a commitment of $150 million to bring AI into more classrooms and communities.

Intel Corp. too announced its new AI chip for laptops, Panther Lake (also known as the Intel Core Ultra Series 3), and said the company has plans to launch a new platform to address a growing market for handheld video gaming machines.

Meanwhile, Hyundai-owned Boston Dynamics publicly demonstrated its humanoid robot Atlas for the first time, ratcheting up a competition with Tesla and other rivals to build robots that look like people and do things that people do.

Atlas has “human-scale hands” and 360-degree cameras. It’s water-resistant and can operate in temperatures as low as -4F. Google DeepMind is partnering with Boston Dynamics to integrate Gemini into the humanoid robot.

Smart shopfloor Siemens AG President and CEO Roland Busch expanded his deal with Nvidia’s Huang to build the industrial AI operating system. Siemens launched Digital Twin Composer software to power the industrial metaverse. PepsiCo uses this software to simulate upgrades to its facilities in the US, with plans to scale globally. Siemens also highlighted technologies for accelerating drug discovery, autonomous driving, and shop floor efficiency. In manufacturing, Siemens announced a collaboration to bring Industrial AI to Meta Ray-Ban AI Glasses.

Climate Change and You

A fortnightly newsletter about climate change from both India and abroad, brought to you by Bibek Bhattacharya and Sayantan Bera.

India, Europe, Malaysia take Grok to task for sexually explicit deepfakes

Authorities in India, Europe, and Malaysia are investigating Elon Musk-owned X after its AI chatbot Grok let people create sexualised images of children and women. The investigations were initiated after social media users, media outlets and officials flagged that Grok was unleashing images of women and children in extremely skimpy clothing on demand, a functionality X has in the past referred to as “spicy mode”.

As an example, author and influencer Ashley St. Clair, mother of one of Elon Musk’s many children, has accused her billionaire ex’s company X of creating sexually explicit images of herself with Grok and said she was considering legal action against xAI—the parent company of the AI chatbot.

On 2 January, India’s IT ministry issued an order directing Musk’s X to take corrective action on Grok, and carry out a “technical, procedural and governance-level review of the AI-based application, Grok, to ensure that there is no content that promotes nudity and other unlawful activities in any form”.

It also asked X to update its terms of service and other related policies against those violating the rules and submit a detailed action report covering “specific technical and organisational measures adopted or proposed in relation to the Grok application; role and oversight exercised by Chief Compliance Officer; actions taken against offending content, users and accounts; and mechanisms put in place to ensure compliance with the mandatory reporting requirement under Section 33 of the BNSS”.

French authorities, too, will investigate the proliferation of sexually explicit deepfakes generated by AI platform Grok on X, the Paris prosecutor’s office told POLITICO. French lawmakers Arthur Delaporte and Eric Bothorel contacted the prosecutor’s office on 2 January after thousands of non-consensual sexually explicit deepfakes were generated by Grok and published on X.

Authorities in Malaysia are also investigating the latest allegations against Grok. “Anyone using Grok to make illegal content will suffer the same consequences as if they upload illegal content,” Musk said on Sunday.

AI evangelism to AI evaluation: Stanford HAI on what to expect in 2026

The era of AI evangelism is giving way to an era of AI evaluation, a new article from Stanford Institute for Human-Centred AI (HAI) says. Here are some highlights from the article that has views from the institute’s faculty experts who talk about diverse AI topics, including AI Sovereignty, artificial general intelligence (AGI), Legal AI, Deflating the AI bubble, and AI in Medicine.

• No AGI this year, but AI sovereignty gains momentum: There will be no AGI this year, says James Landay, HAI’s co-director and professor of Computer Science. He expects AI sovereignty to gather significant momentum as countries seek greater independence from dominant AI providers and, more broadly, from the US political system. Landay notes that “sovereignty” remains loosely defined. In one model, a country builds its own large language model; in another, it runs a third-party model on local GPUs to ensure data never leaves national borders.

• Legal AI gets smarter and more accepted: Julian Nyarko, professor of Law and Stanford HAI associate director, believes two themes will shape AI’s role in legal services this year. The first is rigour and return on investment, as firms and courts move beyond asking whether AI can write to questioning how well it performs, on which tasks, and at what risk. The second is AI taking on more complex work. This shift, Nyarko says, will require new evaluation frameworks, including LLM-as-judge methods and pairwise preference ranking.

• The AI bubble may deflate, not burst: Angèle Christin, associate professor of communication and Stanford HAI senior fellow, expects a more realistic view of AI’s capabilities to emerge. In many cases, she says, AI’s impact will be moderate, delivering efficiency and creativity gains rather than transformation. She does not expect the bubble to burst, but suggests it may stop expanding at its current pace.

• A “ChatGPT moment” for AI in medicine: Curtis Langlotz, professor of Radiology, Medicine and Biomedical Data Science, predicts an imminent “ChatGPT moment” for AI in medicine. This would come as models are trained on vast, high-quality healthcare datasets comparable in scale to those used for chatbots.

• Advancing Human-AI interaction for long-term impact: Diyi Yang, assistant professor of Computer Science, hopes to see greater focus on human-centred AI systems that go beyond technical capability to align with how people think, interact and collaborate over time.

• GenAI looks to bypass the enterprise: Nigam Shah, professor of Medicine and Biomedical Data Science and chief data scientist at Stanford Health Care, predicts that Gen AI developers will grow impatient with long enterprise decision cycles, particularly in healthcare. As a result, they may increasingly go directly to end users with free or low-cost applications. He points to examples such as OpenEvidence’s literature summaries and AtroposHealth’s on-demand clinical answers.

• Measuring AI in real time: Erik Brynjolfsson, director of the Digital Economy Lab, believes executives will soon track AI exposure metrics daily alongside revenue dashboards. Policymakers, he adds, will use these metrics to guide training, safety nets and innovation policy.

You may read the full article here.

AI TOOL OF THE WEEK

By AI & Beyond, with Jaspreet Bindra and Anuj Magazine

The AI capability we unlocked this week is based on a tool - Perplexity AI’s Patent Search.

What problem does Perplexity AI patent search solve? Innovation teams and R&D groups often spend months developing ideas, only to discover late in the process that closely related patents already exist. This can lead to costly pivots, legal exposure, or even abandoned projects after significant investment. Traditional patent databases rely on rigid keyword and Boolean searches, which can be challenging to use without IP expertise, and often make it difficult to view the broader landscape of related filings across jurisdictions.​

Perplexity Patents addresses this by combining natural language search with an AI patent research agent that works over a dedicated patent index. Perplexity Patents is an AI-powered patent research experience available on Perplexity AI that automatically recognizes patent-oriented queries and returns collections of relevant patents with key metadata and links to the original documents.

Instead of wrestling with complex search syntax, teams can ask plain-language questions about technologies, use cases, or companies and iteratively refine their understanding of the prior-art landscape

How to access: https://www.perplexity.ai/

Perplexity AI Patent Search can help you:

- Validate ideas quickly: Use natural language queries to check for related patents and prior art before investing heavily in development or filing strategies, reducing the chance of overlooking critical existing work.

- Understand the patent landscape: Explore who is active in a space, what types of solutions have been claimed.

- Explore potential whitespace: By comparing crowded areas (with many similar patents) to less-represented concepts or applications, you can identify regions that appear under-explored.

Example: Suppose your team is developing an AI-powered quality control system for manufacturing. Here is how Perplexity Patents can fit into your workflow:

Start broad

Ask: “Are there patents related to AI-driven quality control in manufacturing?”

Perplexity can surface a set of related patents with titles, assignees, publication dates, and links, helping you see how crowded the general idea space is.

Refine with context

Follow up with: “Find patents on vision-based defect detection for automotive parts” or “Show patents focused on inline visual inspection in automotive manufacturing.”

Deep-dive analysis

You can then use conversational prompts to structure and analyze what you see, for example:

- “Group these patents by company and highlight the most active assignees.”

- “Summarise the key ideas or claimed approaches in these 10 patents.”

- “List patents in this area published in the last few years and group them by assignee.”

- “Highlight aspects of AI-based quality control in manufacturing that appear less frequently in this set of patents.”

Export and extend

Perplexity (especially when used together with Labs and code-friendly workflows) can organise results into structured tables, which you can then export as CSV or use to generate charts that visualise clusters, timelines, or assignee activity.

What makes Perplexity AI Patent Search special?

- Natural language first: You can ask open-ended questions about technologies, industries, or use cases instead of constructing complex Boolean queries, lowering the barrier for non-experts to explore patents.​

- Automatic patent recognition: When your query is clearly about patents, Perplexity activates its patent-oriented retrieval without requiring you to switch tools or modes manually.​

- Accessible but not a legal substitute: Perplexity Patents makes patent exploration significantly more intuitive and faster for innovation teams, but it is a research assistant, not a replacement for professional patent searches or legal opinions.

Note: The tools and analysis featured in this section demonstrated clear value based on our internal testing. Our recommendations are entirely independent and not influenced by the tool creators.

AI BITS AND BYTES

Can India put AI to work for the national cause of inclusive growth?

India is steadily moving up the AI value chain. After chairing the Global Partnership on AI Council in 2024, the country is readying to host a five-day India AI Impact Summit 2026 in New Delhi in mid-February.

AI’s promise, however, cannot be defined by scale or speed alone. The technology should also be democratised. This challenge is particularly acute across Asia and the Pacific, which already account for more than half of global AI users and nearly 70% of AI patents. A UNDP report, The Next Great Divergence, cautions that while AI could lift GDP growth by around two percentage points and raise productivity by up to 5% in key sectors, these gains will not be evenly shared. (read more)

‘We will be outcompeted,’ leading AI expert warns

David Dalrymple, a leading AI safety expert, has warned that the world “may not have time” to prepare for the safety risks posed by cutting-edge AI systems.

Dalrymple, who works as a programme director and AI safety expert at the UK government’s Advanced Research and Invention Agency (ARIA), told The Guardian that the development of AI was moving “really fast” and that it can’t be assumed these “systems are reliable”.

“We will be outcompeted in all of the domains that we need to be dominant in, in order to maintain control of our civilisation, society and planet,” he said. (read more)

‘We need to label AI content and verify what’s real,’ says Instagram head

Instagram head Adam Mosseri has warned that the rapid rise of AI-generated images and videos could fundamentally reshape how people perceive content on the platform, pushing authenticity to the centre of its future strategy.

The Meta Platforms executive suggested on X that AI is advancing so quickly that distinguishing real photos and videos from synthetic ones may soon become extremely difficult, forcing Instagram to adapt at pace. (read more)

‘The spark remains human’ — Perplexity CEO on why AI lacks curiosity

Aravind Srinivas, CEO of AI search startup Perplexity, said that AI systems are still fundamentally dependent on humans to define what problems are worth solving, arguing that curiosity and question framing remain beyond the reach of machines. Speaking on a recent podcast with writer and entrepreneur Prakhar Gupta, Srinivas stressed that while AI excels at solving, optimising and verifying solutions, it does not independently identify meaningful problems. (read more)

🔗 Big IPOs in 2026 — SpaceX, Anthropic, OpenAI are on the list

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Madhumita Sen Choudhury. Produced by Tushar Deep Singh.

Happy New Year. As India steps into 2026, it is encouraging to see steady progress towards Viksit Bharat, or Developed India, despite persistent infrastructure and financial constraints.

India has already built a strong Digital Public Infrastructure backbone through initiatives such as Aadhaar, the world’s largest digital identity system, the Unified Payments Interface (UPI), and platforms like DigiLocker and Account Aggregator, supported by frameworks such as OpenForge. The country is also the world’s third-largest startup ecosystem, a major hub for IT and digital services, and home to more than 2,000 global capability centres (GCCs) run by multinational companies.

Now, India’s push to adopt artificial intelligence (AI) could add as much as $1.7 trillion to the economy by 2035, according to government estimates, making it one of the most powerful growth engines of the coming decade. To that end, the government has allocated more than ₹10,300 crore across five years for the IndiaAI Mission. The programme aims to build domestic computing infrastructure, support AI startups, develop indigenous models and expand skilling initiatives, a Press Information Bureau (PIB) release said on Tuesday.

That said, while artificial intelligence sits at the heart of the Viksit Bharat vision, which seeks to lift India’s long-term growth rate beyond 8%, the semiconductor story is just as critical.

The India Semiconductor Mission, launched in December 2021 under the Ministry of Electronics and Information Technology, provides structured support and works with global companies to bring large semiconductor investments into the country. As of 2025, the mission has approved 10 projects across six states, with a total investment commitment of ₹1.60 trillion.

Against this backdrop, India’s homegrown processor DHRUV64, launched on 15 December, marks a significant milestone in the country’s semiconductor journey. It is a strategic step towards reducing reliance on foreign suppliers and strengthening a fast-growing electronics and digital economy, both central to the Viksit Bharat agenda.

The processor has been developed by the Centre for Development of Advanced Computing (C-DAC), which builds processor intellectual property, system-on-chips, development boards, and related tools under Meity’s flagship programmes. The 1.0 GHz, 64-bit dual-core RISC-V microprocessor, named DHRUV64, is expected to be followed by the next-generation Dhanush and Dhanush+ processors, which are currently under development.

Microprocessors are often described as the brains of modern electronic devices, powering everything from mobile phones and computers to automobiles, medical equipment, defence systems, and satellites. Built using modern fabrication employed in high-performance chips, DHRUV64 is designed to serve demanding sectors such as 5G infrastructure, automotive systems, consumer electronics, industrial automation, and the Internet of Things.

Importance of RISC-V

Unlike widely used processor architectures such as x86 or ARM, which are owned and controlled by specific companies, the Reduced Instruction Set Computer architecture, or RISC-V, is open and royalty-free. This allows anyone to design processors without paying licensing fees, making it particularly attractive for countries and companies seeking greater control over their technology choices.

At its core, RISC-V is an instruction set architecture that defines the basic commands a processor can execute, such as performing calculations or moving data between memory and storage. It represents the fifth major iteration in a long lineage of RISC designs developed over decades, notably at the University of California, Berkeley.

A key strength of RISC-V lies in its modular design. It combines a small, stable base set of instructions with optional extensions that can be added depending on the application. This flexibility enables chip designers to build highly customised processors, ranging from simple embedded systems to smartphones, data centres, and specialised AI accelerators, delivering efficiency, adaptability, and faster innovation.

Climate Change and You

A fortnightly newsletter about climate change from both India and abroad, brought to you by Bibek Bhattacharya and Sayantan Bera.

Implications for India

The country consumes roughly 20% of the world’s microprocessors and is home to about 20% of global chip design talent. Processors such as DHRUV64 strengthen the domestic pipeline of skilled semiconductor professionals, while advancing local design capabilities.

More broadly, RISC-V offers India a path to strategic autonomy by reducing dependence on foreign intellectual property and licences. By lowering costs, encouraging indigenous chip design, and supporting custom processors for areas such as digital public infrastructure, defence, and space, RISC-V can help India evolve from a major chip consumer into a credible global player in semiconductor design.

That said, India’s indigenous microprocessor development ecosystem precedes DHRUV64. The Government of India launched the Digital India RISC-V (DIR-V) Programme to advance the vision of Aatmanirbhar Bharat. It aims to establish India as a global hub for Electronics System Design and Manufacturing (ESDM).

• SHAKTI (2018, IIT Madras): Designed for strategic, space, and defence applications.

• AJIT (2018, IIT Bombay): A microprocessor for industrial and robotics applications.

• VIKRAM (2025, ISRO–SCL): A processor developed for space applications such as navigation, guidance, and mission operations, and engineered to withstand extreme space conditions.

• THEJAS64 (2025, C-DAC): Designed for industrial automation, and manufactured domestically at Semiconductor Lab, Mohali.

• DHRUV64 (2025, C-DAC): Designed for 5G infrastructure, automotive systems, consumer electronics, industrial automation and IoT.

• DHANUSH64 and DHANUSH64+ System on a Chip (SoC) variants: Currently under development.

AI TOOL OF THE WEEK

By AI & Beyond, with Jaspreet Bindra and Anuj Magazine

The AI hack we unlocked today is based on a tool — Gemini in Google Slides

What problem does Gemini in Google Slides solve?

You’ve already built your presentation in Google Slides—your content is there, your narrative is set but your slides look flat and unprofessional. You know what good design looks like, but recreating it slide-by-slide is tedious. You have images with mismatched backgrounds that clash with your aesthetic. Your data sits in boring bullet points when it should be a compelling infographic. You need custom visuals that match your content, not generic stock images.

Unlike tools that generate entire presentations from scratch (forcing you to start over), Gemini works within your existing Google Slides deck, enhancing what you’ve already created. It solves the specific problem of transforming slides you’ve built into visually polished presentations without leaving Slides, without external design tools, and without abandoning the work you’ve already done.

How to access: https://slides.new (requires Google Workspace or Google AI Premium plan)

Gemini in Google Slides can help you:

- Generate professional infographics from your data: Turn your existing content and data into polished infographics using simple prompts—no external design tools needed.

- Create custom images that match your content: Generate contextually relevant visuals connected to Google Search for accurate real-world representations.

- Edit images without leaving Slides: Remove backgrounds, replace backgrounds, or expand image backgrounds directly in your presentation.

Example 1: Beautifying an existing slide — You’ve created a basic slide with bullet points about Q4 sales targets. Prompt “Beautify this slide” and Gemini instantly transforms it into a professionally designed visual with enhanced layouts, colours, and graphics all while keeping your text intact and matching your deck’s overall style.

Example 2: Creating data infographics from existing content — You need to present customer satisfaction trends. Instead of manually creating charts, prompt Gemini: “Create an infographic showing customer satisfaction scores across Q3 and Q4 with percentage improvements.” Gemini generates a polished, data-rich infographic with accurate text rendering and professional design.

Example 3: Generating custom images — Your presentation needs a visual for a remote work section. Prompt: “Generate an image of a modern home office with a laptop, plants, and natural lighting.”

Gemini creates a high-quality, contextually appropriate image that fits your slide perfectly.

Example 4: Removing image backgrounds — You have employee headshots with different background colors for a “Meet the Team” slide. Right-click each image → “Remove background”, and Gemini instantly erases backgrounds, letting you create a unified, professional look across all photos.

Example 5: Replacing backgrounds — You’ve inserted a product image but the background doesn’t match your presentation theme. Click the image → “Edit image” → “Replace background” and prompt: “Modern minimalist office setting with white walls and wooden desk.” Gemini intelligently replaces just the background while preserving your product perfectly.

Example 6: Expanding image backgrounds — An important image is too small or cropped awkwardly. Use “Expand background” to seamlessly extend the image’s background without distortion, making it fit your slide layout perfectly while maintaining quality and consistency.

What makes Gemini in Google Slides special?

- Works within your existing deck: Enhances slides you’ve already created rather than forcing you to start from scratch.

- Context-aware design: Maintains your deck’s look and feel for consistent, professional results across all slides.

- Powered by Nano Banana Pro: Delivers sharp text rendering in infographics and images, plus sophisticated visual outputs that understand real-world context through Google Search integration.

Note: The tools and analysis featured in this section demonstrated clear value based on our internal testing. Our recommendations are entirely independent and not influenced by the tool creators.

AI BITS AND BYTES

Inside Meta’s acquisition of Manus AI

Meta said it will acquire Manus, an AI startup. The tech giant is rapidly expanding its platforms to incorporate advanced AI. Although the financial terms of the transaction weren’t disclosed, a source with direct knowledge of the matter told Reuters that the deal could value the Singapore-based firm at $2-3 billion. This acquisition comes as several tech giants are racing to develop and deploy advanced AI technologies, aiming to gain an edge in the rapidly evolving AI landscape.

OpenAI raises stakes, offers massive $550,000 for ‘critical’ new hire

OpenAI is hiring a new employee to address the growing concerns that AI may pose, and is willing to pay a neat price for the right candidate.

In a recent job listing, OpenAI said it is hiring a “Head of Preparedness”. The new employee at Sam Altman’s company will focus on reducing the harms associated with AI, including cybersecurity and mental health among others.

Calling the job a “critical role at an important time”, Sam Altman in a post on X explained why a “head of preparedness” is required at OpenAI.

“Models are improving quickly and are now capable of many great things, but they are also starting to present some real challenges. The potential impact of models on mental health was something we saw a preview of in 2025; we are just now seeing models get so good at computer security they are beginning to find critical vulnerabilities,” the OpenAI CEO wrote.

Satya Nadella warns of ‘model overhang’ in 2026

Microsoft CEO Satya Nadella believes 2026 will be a “pivotal year” for AI as the technology transitions from a phase of discovery to “widespread diffusion”.

In a blogpost, he said the industry is now beginning to distinguish between “spectacle” and “substance” to gain a clearer sense of where the new technology is headed and how it is going to impact the world. Nadella added that we are currently in a state of “model overhang”, where the capabilities of AI models are outpacing our ability to use them for tangible real-world impact.

The quantum era crept up while you were watching AI

Step aside, AI. Another transformative technology with the potential to reshape industries and reorder geopolitical power is finally moving out of the lab: quantum.

The United Nations dubbed 2025 the ‘International Year of Quantum Science and Technology’. It’s been marked by a flurry of announcements—and a mountain of hype—around a mind-boggling field of science long dismissed as perpetually a decade away from usefulness. But that’s how people talked about AI too, before ChatGPT spurred the current global arms race and investor euphoria. (read more)

India extends historic invitation to China for 2026 AI Impact Summit

India has extended an official invitation to China to participate in its weeklong 2026 AI Impact Summit in New Delhi. This marks the first time that India has done so, as New Delhi seeks to showcase its foundational AI models and present a policy directive that other nations agree with. The move comes after China attended 2025’s AI Action Summit in France, where India was the co-chair nation, months after the country declined to be a signatory to the 2024 AI Summit in Seoul, South Korea.

Must Reads

• Amazon, Meesho in race to crack LLM-powered AI shopping

• Enterprises actually using AI will be the big trend in tech for 2026

• Five things to know about Nvidia’s $20-billion licensing deal

Hope you folks have a great weekend, and your feedback will be much appreciated— just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

As 2025 draws to a close, most of us would agree that artificial intelligence (AI) undoubtedly has dominated conversations, fuelled by both its potential, the rapid maturation of India’s tech ecosystem, fears of job losses, and concerns over an AI bubble due to irrational exuberance.

AI is no longer a niche technology but has become a defining economic and strategic force. The field has moved decisively from classical AI to Generative AI (Gen AI), and now to Agentic AI.

Traditional AI focused on analysing historical data to detect patterns and make predictions. Gen AI changed the game by producing original content such as text, code, images, audio and video, through simple prompts in natural languages such as English and Hindi. That shift went mainstream with the launch of ChatGPT three years ago. Today, OpenAI claims more than a million ChatGPT users, and Sensor Tower estimates Gen AI apps could generate over $10 billion in revenue by 2026.

The next leap is Agentic AI. These systems go beyond content generation to reason, plan, and take autonomous actions on a user’s behalf—nudging AI closer to artificial general intelligence (AGI). Over the past year, Agentic AI has gained rapid traction, including systems that can execute transactions, such as making payments, as OpenAI has demonstrated for users in the US. AI-native browsers—many built on Google’s open-source Chromium engine, including Perplexity’s Comet and Opera’s Neon, alongside newcomers like Dia and Brave—are accelerating this shift.

No company illustrates the scale of the AI boom better than Nvidia. The chipmaker, the biggest beneficiary of AI infrastructure spending, became the world’s first company to reach a $5 trillion market capitalisation in late October. While its current mcap is around $4.6 trillion, forecasts suggest Nvidia’s valuation could climb to $8-10 trillion by the end of the decade, driven by relentless demand for AI computing power.

Nvidia’s rise rests on its dominance in high-performance GPUs—once designed for gaming, now indispensable for training and running AI models. It has also reinvented itself as an AI infrastructure giant, combining hardware with its CUDA software platform and deep data-centre partnerships. Nearly 89% of its revenue now comes from data centres, with gaming and AI PCs contributing about 10%. Nvidia estimates global spending on AI infrastructure to reach $3-4 trillion by 2030.

Globally, Gartner expects AI spending to approach $1.5 trillion in 2025, with a large share flowing into data centres powered by Nvidia GPUs and Google’s tensor processing units (TPUs). India mirrors this trend. IT spending is projected to grow at 10.6% year-on-year to $176 billion in 2026, while data-centre spending is expected to surge 20.5%, outpacing all other segments.

At the core of these facilities are expensive AI chips—an Nvidia H100 can cost up to $40,000—making supply tight and access uneven. Under the IndiaAI Mission, the government has stepped in, procuring 38,000 GPUs and offering them to domestic firms at subsidised rates.

But are Gen AI and Agentic AI delivering returns?

An October Wharton enterprise study found that 75% of large US firms already report ROI from AI, with four in five expecting positive returns within two to three years. Two-thirds are spending more than $5 million annually on Gen AI, and over 10% have budgets exceeding $20 million.

Yet scepticism persists. Gartner warned that by the end of 2025, 30% of Gen AI projects would be abandoned after proof of concept, citing poor data quality, rising costs, weak risk controls, and unclear business value. Greyhound Research’s CIO Pulse 2025 reported that 64% of Indian organisations have not scaled even half of their AI pilots. An MIT study added fuel to the debate, finding that despite $30–40 billion in enterprise GenAI investment, 95% of organisations have seen no measurable returns. Productivity gains, it argued, accrue to individuals, not balance sheets, with projects faltering due to fragile workflows and weak integration into core operations.

These contradictions point to real risks. Aggressive spending could squeeze margins, delay payback, and weaken demand—while any shift in investor sentiment could puncture lofty valuations. Still, the spending shows little sign of slowing.

Critics warn of a “circular investment” loop, where the same players fund, supply, and consume AI infrastructure. In October, OpenAI announced that Nvidia would take an equity stake reportedly worth up to $100 billion while jointly developing massive GPU-powered data centres. Alongside Oracle Corp., OpenAI also unveiled plans for five large facilities under the $500-billion “Stargate” project, expected to deploy hundreds of thousands of Nvidia chips. OpenAI is also reportedly in talks with Amazon for an investment that could exceed $10 billion.

OpenAI is not alone. Meta, Google and other Big Tech firms are accelerating their investments in AI infrastructure at a pace rarely seen before, intensifying both optimism and anxiety.

Does this surge signal the start of an AI bubble, or another AI winter?

Nvidia co-founder and CEO Jensen Huang has dismissed such concerns. Speaking to Bloomberg Television on 29 October, Huang said Nvidia’s latest chips are expected to generate close to half-a-trillion dollars in revenue. “I don’t believe we are in an AI bubble,” he said. “All of these different AI models are being used, and customers are paying for the services.”

That said, while AI agents can write, deploy and debug code at speed, they also hallucinate, overlook basic errors and, at times, disrupt the very systems they are designed to improve. Greater autonomy comes with clear trade-offs. An AI agent may generate code in minutes, but without human judgement the risk of costly failures remains high. Fully autonomous developers cannot yet be trusted. A randomised controlled trial by the Model Evaluation and Threat Research group found developers using AI tools took 19% longer to complete tasks than those working without them. In practice, AI can slow teams down rather than speed them up. Trust, therefore, is critical for enterprise adoption.

Enterprises also need clarity on when to use classical AI, generative AI or agentic AI. The technology should follow the problem, not the other way around. Without this discipline, organisations risk falling into “AI washing”, where inflated claims mask limited real-world impact.

Meanwhile, the infrastructure boom is also creating knock-on effects. Demand for GPUs is driving up power consumption, putting pressure on electricity grids and raising concerns around water use and noise pollution. Data-centre operators are responding by shifting to renewable energy, experimenting with advanced cooling methods and even exploring the idea of space-based data centres. Some of the most creative solutions are emerging in Nordic countries. AI data centres are being built inside Cold War-era bunkers and abandoned granite mines, where naturally cool temperatures reduce energy needs. Excess heat from these facilities is also being recycled. In Espoo, Finland, heat generated by Microsoft data centres now supplies hot water to radiators serving roughly 40,000 residents.

From an investment perspective, JP Morgan researchers caution that every major capital-expenditure cycle carries the risk of overbuilding. But they qualify that during the dot-com boom, only about 7% of the global fibre-optic network was utilised at its peak, leaving years of excess capacity while the current AI cycle looks very different. Data-centre vacancy rates are at record lows, utilisation is close to 80%, and demand for compute continues to outstrip supply. More data has been created in the past three years than in all of previous history, and AI workloads are expanding rapidly.

Even so, they add that caution is not misplaced. “The scale of spending is enormous, the pace unprecedented and some assumptions around ROI—like the useful lives of assets—remain open questions. History reminds us that enthusiasm can run ahead of reality. Yet so far, today’s players are far better capitalised than those of the dot-com era, AI monetisation is underway and the risk of overbuilding seems limited in the near term. As this story unfolds, investors should focus on selectivity, leaning into active management to separate transformative winners from bubbly valuations,” they conclude.

Climate Change and You

A fortnightly newsletter about climate change from both India and abroad, brought to you by Bibek Bhattacharya and Sayantan Bera.

Here are a few other trends we covered in Mint Tech Talk in 2025

• AI patents comprise more than 25% of all tech patents filed in India

• Why are AI models are resorting to blackmail?

• How enterprises are shifting focus from digital to AI-first strategies

• Why Spatial Intelligence may be AI’s next frontier

• How Google’s TPU–Gemini push is unsettling Nvidia, ChatGPT

• India can show the world how to make AI inclusive

AI Tool of the Week

By AI & Beyond, with Jaspreet Bindra and Anuj Magazine

The AI hack we unlocked today is based on Google Gemini’s new capability—mini-apps in Gemini (powered by Opal, Google’s experimental framework for building interactive applications).

What problem do mini-apps in Gemini solve?

Traditional AI chatbots give you answers, but you still have to do the work. Ask for a budget breakdown? You get text that you copy into Excel. Need a project tracker? You manually recreate it in another tool. Want to compare options? You’re building tables yourself.

This creates friction in user experience. AI understands what you need, but you’re stuck doing manual work to make it usable.

Mini-apps in Gemini close this gap by creating ready-to-use, interactive tools right in your chat. No more copying, pasting, or rebuilding in other apps.

How to access: Go to gemini.google.com → Click the beaker icon (top right) → Enable “Gems with mini apps”

Mini-apps can help you:

• Create interactive tools instantly: Generate calculators, planners, trackers, and games without leaving the chat.

• Build functional prototypes: Test ideas with working apps before investing in full development.

• Collaborate more effectively: Share interactive tools that teammates can use, not just read.

Example: Suppose you’re launching a new product and need a streamlined workflow for your marketing team. Here’s how mini-apps in Gemini help:

1. Click “New Gem”: Start creating your custom mini-app from the Gems section

2. Describe what you need: “Create an app that evaluates product features, generates marketing content ideas, creates vision boards, and outlines next steps”

3. Gemini builds the workflow: You get a multi-step interactive app with sections for Product Name, Description, Content Generation, Vision Board Creation, and HTML rendering

4. Use it immediately: Your team inputs product details, and the app automatically generates marketing content, visual concepts, and actionable next steps

5. Remix and refine: Click “Remix” to modify steps, add fields, or customize the workflow for your specific needs

What makes this special?

• No technical skills needed: Build functional apps through simple conversation

• Everything in one place: From request to working tool without switching platforms

• Instant iterations: Modify apps by simply asking—no manual editing required

Note: The tools and analysis featured in this section demonstrated clear value based on our internal testing. Our recommendations are entirely independent and not influenced by the tool creators..

AI BITS AND BYTES

‘AI should remember your life’: Sam Altman on OpenAI’s next big bet

OpenAI CEO Sam Altman has suggested that the next major leap in AI will not come from sharper reasoning abilities, but from something far more fundamental: memory. Speaking on a recent podcast with technology journalist Alex Kantrowitz, Altman outlined a future in which AI systems are able to retain and learn from vast amounts of personal data across an individual’s lifetime, fundamentally reshaping the idea of a digital personal assistant.

According to Altman, today’s AI tools are already strong at reasoning tasks, but they fall short when it comes to long-term recall. Humans, he argued, remain the best personal assistants because they understand context and nuance, yet they struggle to remember details consistently.

OpenAI strengthens ChatGPT safeguards for users aged 13 to 17: Here’s how

OpenAI has announced changes to its Model Spec, the internal rulebook that guides how its AI systems behave, with a new and explicit focus on protecting teenagers who use ChatGPT. The San Francisco-based company said the updated guidelines put the safety of users aged 13 to 17 above all other objectives, including the long-standing goal of maximising helpfulness and user freedom. The move comes amid growing scrutiny of how generative AI tools affect younger users.

How an AI assistant can help you protect your digital footprint

While consent managers under the DPDPA make preferences easier to track and manage, the burden remains on users to understand privacy risks. An AI privacy assistant can bridge this gap by explaining rights and consequences in a language and context that the user understands.

9 Viral AI prompts on Grok decode your personality without horoscopes

An AI expert’s viral X post claims chatbots like Grok can replace astrologers using powerful birth date prompts. The nine shared prompts promise deep self-analysis, personality insights and future growth lessons without horoscopes or tarot cards. Here are the prompts for you to try.

The ‘smart doctor’ at the heart of India’s digital health push

The government is moving to deploy a clinical decision support system across nearly 70,000 public and private hospitals in India to standardise the quality of care and reduce medical errors. The tool, dubbed “smart doctor”, has been developed by AIIMS New Delhi and will be rolled out under the government’s flagship Ayushman Bharat Digital Mission across all Indian hospitals.

Hope you folks have a great weekend, and your feedback will be much appreciated — just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Tushar Deep Singh.

This year’s Foundation Model Transparency Index (FMTI) only confirmed our suspicion that most companies continue to be opaque when it comes to disclosing information about artificial intelligence (AI) models.

Companies evaluated by the Index posted an average score of 41, a 17-point drop from last year, though four points higher than in 2023. IBM topped the rankings for the third consecutive year, scoring 95, which is 10 points higher than the best-performing company in 2024. The second- and third-ranked firms slipped compared to their 2024 scores.

The Index assessed four companies for the first time, including two Chinese firms, DeepSeek and Alibaba, which scored poorly at 32 and 26, respectively, despite releasing open-weight models. The other two new entrants, xAI and Midjourney, tied for the lowest score at 14.

Developed by researchers from Stanford University’s Center for Research on Foundation Models (CRFM) and Institute for Human-Centered Artificial Intelligence (HAI), MIT Media Lab, and Princeton University’s Center for Information Technology Policy, FMTI has evaluated major foundation model developers such as Google and OpenAI since 2023 across 100 indicators. It recently expanded and tightened these benchmarks to reflect new developments in foundation model research.

Highlights:

• Training data continues to be opaque. In 2023 and 2024, companies scored poorly in the data sub-domain, being opaque about topics like copyright and licences. This continues to be the case in 2025: companies tend to score poorly in the data acquisition and data properties sub-domains.

• Many companies do not disclose basic information about the model itself. Amazon, Google, Midjourney, Mistral, OpenAI and xAI do not score any indicators in the model information subdomain, such as the basic model information indicator (which includes input modality, output modality, model size, model components, and model architecture).

• Openness correlates with—but isn’t sufficient for—transparency. The 5 open-weight developers outscore the 8 closed-weight developers overall and on every domain. However, high-profile open model developers are still quite opaque: Alibaba, DeepSeek, and Meta all score in the bottom half of companies.

• US companies score higher on-average (but also hold the two lowest scores). The discrepancies between the nine US companies and the four non-US companies are driven by the downstream domain, specifically highlighting geographic differences in disclosures about usage data, post-deployment impact measurement, and accountability mechanisms.

You may read the full report here.

OpenAI’s GPT-5.2 pits itself against Gemini 3, others

Anyone watching the AI market closely will notice that tech companies are swift in responding to upgrades, not necessarily groundbreaking innovations, put out by their competitors. In November, for instance, Google released Gemini 3, giving Alphabet the AI boost it needed.

Gemini 3 was not a routine model upgrade for a company that has spent the past two years playing catch-up after OpenAI and others seized the momentum with rapid breakthroughs in generative AI (GenAI). Rather, it’s an attempt to demonstrate that Google can build powerful models more efficiently than competitors like Nvidia, leveraging an entire stack comprising chips, models, cloud infrastructure, and applications.

Gemini 3, when launched, was pitted directly against the likes of GPT-5.1, LLaMA 4, Claude Sonnet 4.5 and Claude Opus 4.5, Qwen 3, DeepSeek and Kimi 3. OpenAI, in response, launched GPT-5.2 on 11 December.

OpenAI says GPT-5.2 brings significant improvements in general intelligence, long-context understanding, agentic tool-calling, and vision, making it better at executing complex, real-world tasks end-to-end than any previous model.

OpenAI claims GPT-5.2 Thinking is the best model yet for “real-world, professional use”. On GDPval⁠, an eval measuring well-specified knowledge work tasks across 44 occupations, and is its first model that performs at or above a human expert level.

Specifically, GPT-5.2 Thinking beats or ties with top industry professionals on 70.9% of comparisons on GDPval knowledge work tasks, according to expert human judges. These tasks include making presentations, spreadsheets, and other artifacts.

OpenAI also notes a meaningful drop in hallucinations compared with GPT-5.1 Thinking. In a set of anonymized ChatGPT queries, responses containing errors were 30% less common. This is intended to make the model more dependable for writing, research and analytical tasks.

The AI company notes that GPT-5.2 Thinking sets a new benchmark on SWE Bench Pro, a workload that measures real-world software engineering capabilities across four programming languages. OpenAI reports a score of 55.6% on this more challenging evaluation, along with an 80% score on the Python-focused SWE Bench Verified test. You may read more about this here.

Why AI success depends on prioritizing data modernization

Nearly 90% of India’s data and analytics leaders say their organizations must modernize data strategies for AI to deliver real business impact, according to Salesforce’s latest State of Data and Analytics report. While 75% of business leaders face growing pressure to extract value from data, incomplete, outdated, and poor-quality data remains the biggest obstacle.

This gap between data ambition and data reality becomes more acute in the era of agentic AI. Poor data, the report notes, can derail an organisation’s journey toward becoming an agentic enterprise. To bridge this gap, technology leaders are doubling down on fundamentals: timely, context-rich data, stronger governance, and zero-copy architectures that unlock distributed data wherever it resides. Many are also adopting emerging tools such as agentic analytics, which embed trusted insights directly into everyday workflows.

Even high-quality data delivers little value if it remains trapped. Data and analytics leaders estimate that 26% of enterprise data is siloed, inaccessible, or unusable, and 75% believe their most valuable insights sit within this inaccessible portion.

The consequences are significant. More than eight in ten leaders cite weaker AI capabilities, fragmented customer views, reduced personalisation, and missed revenue opportunities as a result. To address this, 52% of organizations are adopting zero-copy data integration, which enables access to data across multiple databases without moving or duplicating it. In India, companies using zero-copy architectures are 40% more likely to have fully connected customer data and 22% more likely to succeed with AI initiatives than those that do not. You may read the report here.

Climate Change and You

A fortnightly newsletter about climate change from both India and abroad, brought to you by Bibek Bhattacharya and Sayantan Bera.

AI Tool of the Week

by AI&Beyond, with Jaspreet Bindra and Anuj Magazine

The AI hack we feature today is based on a tool—Google Mixboard.

What problem does Google Mixboard solve? Product marketers constantly battle the “vague-to-visual” gap. You get a brilliant but abstract launch theme from leadership say “Festival of Flavors,” “Heritage Meets Modern,” “Minimalist Luxury” and then spend days hunting stock photos, persuading designers for rough mockups, or stitching together mood boards in Canva that never quite capture the vibe. Stakeholders see the deck and say “not feeling it,” forcing endless revision loops. The result? Delayed campaigns, frustrated teams, and watered-down creative. Google Mixboard kills that pain in minutes by turning abstract marketing concepts into living, remixable visual boards using natural language, no design skills required.

How to access: https://mixboard.google.com/projects (login with Google account)

Google Mixboard can help you:

-Speed Execution: Turn a one-line campaign theme into a full visual direction.

-Align Stakeholders: Align marketing, design, and leadership teams on the exact look & feel

-Skip Stockphoto: Create on-brand lifestyle shots, product mockups, and color stories without a single stock site

Example:
Your Chief Marketing Officer (CMO) wants the new sustainable water bottle launch to feel “Traditional yet contemporary—Himalayan adventure meets urban India hydration.”

1. Open Mixboard → New Project → Type: “Himalayan trekking campaign for sustainable water bottle, earthy saffron and turquoise tones, diverse Indian adventurers aged 25-40, candid moments, natural photography style.”

2. Add your customized product images as an upload.

3. Use Custom Prompt: “Place the bottle naturally in every scene, matte texture, golden hour light, add subtle motivational text overlays in English and Hindi.“

4. Delete the images not good enough to align with the theme.

5. Click “Transform” → Add a prompt: “Create a 12-minute presentation exploring the ‘why’ of high-altitude trekking, focusing on the psychological impact, spiritual reflection (prayer flags/solitude), and sense of accomplishment derived from rugged environments.”

In under 30 minutes, you have transformed a vague brief into a compelling visual narrative.

What makes Google Mixboard special?

-Custom Prompt + Nano Banana editing: Full board-wide control with natural language (”make every bottle look dewy from monsoon rain” or “add chai stall backdrop with morning light”).

-Real-time collaboration + one-click Slides export: From fuzzy idea to stakeholder-approved direction in one link.

-Still completely free while in Labs.

Note: The tools and analysis featured in this section demonstrated clear value based on our internal testing. Our recommendations are entirely independent and not influenced by the tool creators.

AI BITS AND BYTES

Microsoft AI boss Suleyman calls Elon Musk a ‘bulldozer’, labels Sam Altman ‘courageous’

Microsoft AI chief Mustafa Suleyman praised his peers in the AI industry during a recent interview. He called Elon Musk a bulldozer for his ability to achieve the impossible. Suleyman also lauded OpenAI’s Sam Altman as a courageous entrepreneur

OpenAI supercharges ChatGPT images

OpenAI has rolled out a new version of ChatGPT Images powered by its flagship image model, offering up to 4x faster generation, more precise edits, and better text rendering. The update is now available to all ChatGPT users and via the API as GPT Image 1.5.

Who is Alex Karp? The tech leader driving Palantir’s role in Western intel

Alex Karp is the billionaire co-founder and CEO of Palantir Technologies, the US AI and big data firm whose software is used by intelligence and security agencies. He is in focus after Palantir renewed a multi-year contract with France’s domestic intelligence agency, the DGSI.

Reliance Jio launches Happy New year 2026 offer with free Gemini AI Pro access

Reliance Jio has launched the new Jio Happy New Year 2026 offer for prepaid users, with access to a Gemini AI Pro subscription, unlimited 5G coverage, and certain OTT subscriptions. Just a few months ago, Jio had partnered with Google to offer the Google AI Pro subscription free for 1.5 years to its subscribers, rivalling the free Perplexity subscription offer by Airtel.

Amazon in talks with OpenAI to invest $10 billion: Report

Amazon and OpenAI are in talks about a potential investment in the artificial intelligence company that could reach billions, CNBC and Reuters reported quoting people in the know about the matter. According to the reports by the two outlets, Amazon’s investment in OpenAI could exceed around $10 billion along with an agreement to use its AI chips. The potential deal with Amazon could value OpenAI at over $500 billion.

Hope you folks have a great weekend, and your feedback will be much appreciated — just reply to this mail, and I’ll respond.

Edited by Rashmi Sanyal. Produced by Shad Hasnain.