In this environment of fear, excitement and FOMO, I believe that the answer lies in AI Literacy. Much like you had to be literate in language to qualify for any job or be digitally literate to work in a white-collar job, you will have to be AI Literate to earn and thrive in this AI Age.

AI Literacy is not upskilling, or training; it is about rewiring your Operating System.

I have worked in this area for a few years now, and distilled here are my Three Laws of AI Literacy™:

First Law: In the Age of AI, everyone need not be an AI Expert, but everyone needs to be AI Literate.

AI literacy is not the ability to build models, but the ability to work with AI responsibly, effectively, and safely.

With digital literacy, most workers did not become software engineers, but everyone had to learn to use computers, search, spreadsheets, and email to participate fully in modern work. With AI, where everyone will need to work with models, agents, tools, bots, robots, and applications to participate in work of any kind.

The literate individual understands what AI can and cannot do, how to frame problems into prompts and workflows, how to verify outputs, and how to manage privacy, IP, and compliance risks.

As answers become a commodity, the AI literate person knows how to articulate the right question.

Organisations that treat AI as a specialist-only capability create bottlenecks, uneven adoption, and shadow usage. By contrast, broad literacy distributes competence across teams and makes AI a shared language and an everyday tool.

Second Law: The definition of Literacy will change from reading, writing, and arithmetic to these plus how to work with AI in everything you do, at work or otherwise.

Literacy has always evolved with the dominant tools of knowledge.

Reading and writing enabled participation in bureaucracies, science, and civic life; numeracy enabled commerce and engineering; and digital literacy became essential as information moved to screens and networks.

AI literacy is the next logical extension as it increasingly mediates how information is created, searched, summarised, and acted upon.

Working with AI includes:

a) practical competence (delegating tasks, iterating, using AI for analysis and creativity),

b) epistemic competence (distinguishing plausible text from reliable truth, demanding sources, cross-checking), and

c) ethical competence (fairness, privacy, disclosure, avoiding harmful or deceptive uses).

As AI becomes embedded in societies, industries and life itself, citizens will need AI literacy to avoid manipulation, understand automated decisions, and exercise agency.

A company or society that treats AI literacy as optional will widen inequality in both opportunity and voice.

Third Law: All enterprise and educational investments in AI models, tools, and agents will not land to provide the desired outcomes, unless all employees or students are AI Literate.

AI becomes a capability, only when people become capable.

History shows that capital investment in hardware and software (the ‘silicon’) rarely yields transformative ROI without a commensurate investment in human capital (the ‘wetware’”). AI initiatives fail less from the wrong models or insufficient data or incorrect algorithms, and more from insufficient integration into real work and workflows.

The value of AI depends on how people ask questions, define success, curate inputs, validate outputs, and redesign processes around new capabilities.

Without literacy, employees underuse or misuse tools and students treat them as shortcuts, undermining learning rather than deepening it.

Literacy creates the human scaffolding that makes AI investments effective and compounded with shared best practices for prompting, evaluation, documentation, and escalation as well as the right governance for sensitive data and regulated decisions.

Literacy reduces ‘pilot purgatory,’ where impressive demos and MVPs never translate into scaled models and changed behaviour. It enables reliable adoption across functions, not just in innovation teams.

You did not hear me say it, but the last week was a slow one for AI! The holidays may have wound down, but AI’s consolidation wars are heating up! While OpenAI quietly perfected its audio future with Jony Ive, Nvidia made its biggest bet ever on Groq, and Meta pounced on Manus. This week’s “slow” news? Anything but.

OpenAI decided to take a break this year, unlike the last days of 2024 when they launched 12 new products in 12 days! Having said that, OpenAI seems to have doubled down on audio/speech in the last few days. It is taking steps to improve its audio AI models, in preparation for its eventual release of an AI-powered personal device, said a person with knowledge of the effort. The device is expected to be largely audio-based, said three people with knowledge of it. The device being built with Jony Ive—the legendary British designer who spent nearly three decades at Apple creating the iconic iPod, iPhone, and iPad, before founding his design firm LoveFrom in 2019—is going to be one of the big news stories of 2026. OpenAI acquired Ive’s startup io for $6.4 billion in May 2025, bringing the designer who revolutionized consumer technology into the AI hardware race. The first prototypes are already finished, and Sam Altman has promised a device that brings “peace and calm” rather than the frenetic smartphone experience—a screenless, audio-first gadget that could redefine how we interact with AI. The company is aiming to release the new audio model in the first quarter of 2026, the person with knowledge of the effort said. I have always believed that speech or audio is the right interface for Generative AI, as speech can convey not only information, but also context, tone, emotion, urgency, and many other subtle inflections which mere text cannot. Plus, voice will bridge the AI digital divide in countries like India, where digital literacy stands at just 38% overall and 25% in rural areas, making voice-based technology a necessity rather than a convenience for millions who interact more naturally in their native languages than through text.

Meanwhile, ‘slow’ in AI is relative. NVIDIA did not wait for the Holidays to end to usurp Groq. It swooped down to strike a roughly $20 billion non-exclusive deal to license Groq’s ultra-low-latency AI inference technology and hire key Groq talent, including founder Jonathan Ross and president Sunny Madra. Groq remains an independent company and continues operating its GroqCloud services, while Nvidia integrates Groq-style inference IP into its AI platform to boost real-time, cost-efficient inference workloads and protect its dominance against emerging accelerator architectures. Groq’s Language Processing Unit (LPU) architecture enhances Nvidia’s GPU-dominant stack with specialized inference capabilities optimized for ultra-low latency and efficiency. Jensen Huang saw the writing on the wall of model pretraining moving more and more to inferencing, and he intends to be a leader in that too. The controversial arrangement transfers Groq’s key assets, IP, and talent (like founder Jonathan Ross) to Nvidia while keeping Groq nominally independent, sidestepping Hart-Scott-Rodino Act filings required for traditional acquisitions. Regulators view this as a “stealth acquisition” or “acquihire loophole,” similar to Microsoft-Inflection, maintaining a fiction of competition despite effective neutralization of a rival.

Meta made its own moves, gobbling Manus AI, the Chinese-Singaporean company which had created a viral sensation with its multi-agent demos. The $2 billion deal at the turn of the year marked one of its largest deals after WhatsApp and Scale AI. Founded by Chinese entrepreneurs with roots in China but relocated operations, Manus specializes in semi-autonomous “general-purpose” AI agents that handle complex tasks like coding, data analysis, research, planning, and web interactions with minimal user input—earning its name from Latin for “hand.” The acquisition accelerates Meta’s shift from foundational models like Llama to full AI agent platforms integrated across Facebook, Instagram, WhatsApp, Meta AI, and enterprise tools. Manus, which hit $100-125M ARR in under a year serving millions via subscriptions, brings proven revenue and 147T+ tokens processed on 80M+ virtual computers. Manus operates semi-independently from Singapore, retaining its app/website sales while Meta scales the tech globally; all Chinese ownership ties were severed to address security concerns. CEO Xiao Hong joins as VP, emphasizing unchanged workflows amid Meta’s AI monetization push.

Subscribe now

Happy New Year!

1. IPO Boom: Anthropic at about $300 billion, SpaceX at $800 billion, Sam Altman dreaming of a $1 trillion OpenAI IPO. If 2025 saw big IPOs in the Indian tech space, 2026 is where the Big Boys of AI and tech will rock the NASDAQ. While Anthropic with its focused enterprise and code generation approach, and Musk’s SpaceX sending and bringing rockets down like clockwork are shoo-ins for the IPO bacchanalia, the OpenAI one will be looked at with most interest. Google has stolen OpenAI’s crown in the latter half of 2025, and GPT 5.2, OpenAI’s latest model, has not exactly set the house on fire. So, unless he pulls a rabbit out of the hat, the $1 trillion valuation looks like a road too far…

2. Nvidia continues to rule: Bubble or no bubble, Jensen Huang and Nvidia will continue to rule the AI world. There are barbarians at the gate, with Google’s TPUs promising real competition, and the Chinese building up their own tech and being directed to shun the A200s of Nvidia. But even if Nvidia’s market share were to come down, they would still have a slightly lesser share of an expanded market. Jensen is adept at creating new markets – robotics and physical AI will be the next, a new way of computing, and of keeping both Trump and China happy. The FT Man of the Year 2025 will continue to be King in 2026.

3. Physical AI (Robotics): 2026 will be the year AI gets a body. Humanoid robots will leave R&D labs and enter warehouses and factories for pilot deployments at scale. The first ones are already out, except they look like a car – autonomous cars, led by Waymo! The company is providing over 250,000 paid trips each week. But what’s happening in 2026 is mind-blowing. Waymo is planning to expand to more than 20 new cities, and here’s the kicker—London will be their first international market, with testing also underway in Tokyo. Thus, physical AI isn’t coming. It’s already here, driving you to dinner.

4. There will be a correction in AI valuations. There is lots of substance in the AI boom, but there is way too much hype. The signs are there – cross holdings, lots of debt being raised, AI-washing by a bunch of get-rich-quick players, enterprises taking their own time adopting AI. However, this bubble will pop with a less satisfying sound than the dotcom one or the 2008 financial one. The BigTechs at the centre of this boom have robust businesses spewing out gobs of cash, and they can afford a downturn or two. The technology is real, and the fastest growing ever. The pace of innovation is furious, and there is more depth here with infra to models to applications. But, yes there will be a pop, and it will be a good thing. (See my article here).

5. The ‘Reasoning’ Web: Search engines will fully transition to ‘Answer Engines.’ (as coined first by Perplexity). Web traffic to static sites will be affected as AI browsers synthesize answers rather than providing blue links. The WWW will need a new business model. The WWW was built on clicks; the Reasoning Web is built on synthesis. We need to figure out how creators get paid, or we’ll kill the very content these AI systems need to train on.

6. Quantum AI Glimmers: We will see the first proof-of-concept “Quantum Machine Learning” algorithms running on early fault-tolerant quantum computers, promising exponential speedups. While quantum computing stocks have been on a tear, the reality check is coming. Nvidia CEO Jensen Huang said “very useful” quantum computers are probably about two decades away. Google’s Sundar Pichai pegged it at 5-10 years minimum, comparing it to where AI was in the 2010s. The quantum computing market is projected to reach just $4 billion by 2030, compared to AI revenue forecast of $390 billion in 2025.

7. The “Human Premium”: As AI content floods the internet, “Verified Human” content will become a luxury good. Brands will market “Human-Made” as a premium differentiator. As AI becomes more powerful, the value of humans will increase. As Sangeet Paul Choudary says eloquently, “Humans will become a luxury good” as the world distinguishes true human connection and creativity from the AI Slop inundating our neural pathways.

8. AI in Governance: Corporate Boards will begin using “AI Board Members” (observer seats) to provide data-driven, unemotional risk analysis during strategic meetings. AI literacy will become a KPI as Boards, regulators and education ministries will start asking for measurable AI literacy programmes, not just generic “digital skills” talk.

9. AI will continue to impact jobs. We saw the first signs in 2025 where entry level cognitive jobs were most impacted. This will continue and accelerate in 2026, as more sectors beyond software and customer service start getting impacted. This will create societal dissatisfaction and a backlash. It will also lead to structural changes in companies (Humans + agents), education (more of humanities) and the way young people look at jobs going forward.

10. Military AI Escalation: The wars of 2025 were fought with more drones than humans. Operation Sindoor, Ukraine v/s Russia, to an extent in the Middle East. This trend will accelerate in 2026 and beyond as autonomous drone swarms and robot soldier armies will become the standard deterrent in geopolitical conflicts, sparking a new Geneva Convention discussion on LAWs or “Lethal Autonomous Weapons.” Companies like Anduril and Chinese drone makers will gain.

11. Cognitive Atrophy Fears: A major public health crisis will be declared regarding “Cognitive Atrophy” in children—the loss of critical thinking skills due to over-reliance on AI tutors. This will give rise to “Disconnect” Movement: A counter-culture movement will rise, rejecting AI-mediated interactions and prizing “analogue-only” spaces and communities.

12. Agents will move from “copilot” to “colleague”: 2026 is likely when agentic AI quietly runs whole workflows—closing tickets, reconciling invoices, triaging code—under light human supervision. Think about it: an AI agent that monitors your inbox, spots a customer complaint, pulls up their purchase history, cross-references inventory, drafts a response with a discount code, gets approval from a manager-agent, sends the email, and logs everything in your CRM. All before you’ve finished your morning coffee. The question isn’t “will this happen?”—it’s “how fast can companies adapt their culture to work alongside non-human colleagues?”

13. Smart glasses will go from novelty to normal-ish: Ray-Ban Meta-style glasses plus Apple/Google/Chinese rivals will start to feel like “the new AirPods”: not universal, but no longer weird. More devices will come from the much ballyhooed OpenAI – Jony Ives one to something from Amazon’s buy of Limitless.

14. Meanwhile in India: India will host the big AI Impact Summit in February. This is a certainty, not a prediction. But I believe there will be announcements around some kind of DPI for AI, a Digital Public Infrastructure protocol to democratize access to compute and models, similar to UPI for payments. This might happen at the Feb 2026 AI Impact Summit. On a personal note, it will be the beginning of realising my JanAI dream.

15. Sovereign Cloud Operational: India’s GPU-based sovereign cloud will go fully live, mandating that sensitive government and banking data be processed on domestic AI infrastructure. Government portals, PSU banks and regulated sectors are likely to pilot IndiaAI-backed sovereign models for chatbots, summarisation and translation, reducing over-reliance on foreign APIs. Perhaps this is more of a fond hope, but if it happens it augurs well for the country. We cannot let all our AI be managed from outside the land.

16. Vernacular Content Explosion: The internet will finally sound like India. 90% of new content consumed in India will be in regional languages, generated and dubbed instantly by AI. The choice of interface will be voice, as India becomes the biggest voice driven AI market in the world. It will also be the market with the highest numbers of subscribers for all the big global AI players (other than the Chinese).

17. IndiaAI GPU clusters will finally feel real: As IndiaAI compute capacity ramps, more Indian startups, researchers and MSMEs will actually train and fine-tune on domestic GPU clouds at subsidised prices. States will compete openly as AI destinations for these GPUs and others. Following Andhra Pradesh’s push, expect Telangana, Karnataka, Maharashtra, Gujarat and others to announce their own AI roadmaps, sandboxes and incentive schemes to attract the mega investments.

Subscribe now

1. Rise of Agentic AI: The buzzword shifted from ‘Generative’ to ‘Agentic.’ AI stopped just chatting and telling you what to do and started doing it—executing complex, multi-step workflows autonomously across enterprise software. OpenAI Browsing and Shopping Agents, the agentic browsers from Perplexity (Comet) and OpenAI (Atlas), and Agentic Workflows from n8n, Zapier, and Make, to name a few, created waves as conversations in enterprises shifted to building AI agents from adopting AI tools. More new buzzwords like Service as a Software being the new SaaS and A-commerce (agentic commerce) started being whispered. (My article on A-commerce here)

2. Claude claimed the coding crown: Anthropic’s Claude 4/Opus 4 positioned itself as the best coding and long-context reasoning model, pushing assistants closer to genuine “co-worker” status. While Google and OpenAI hogged the airwaves, Anthropic and its thoughtful CEO Dario Amodei quietly donned the mantle of the serious, enterprise-focused AI company, amidst all the hype. As software coding became the first real, at-scale use case of Generative AI, it was Claude’s models that became the benchmark to emulate.

3. Going beyond the Pilot: 2025 was the “Put Up or Shut Up” year. C-suites slashed vanity AI projects, focusing strictly on use cases with proven, measurable ROI. Enterprises want to scale; they have waited three years now. Surveys showed AI adoption becoming the default: over three-quarters of firms used AI somewhere, and many began measuring real revenue and cost impact, not just pilots.

4. Coding is Dead; Long Live Coding: Vibe coding with Claude Code, Lovable, Replit, Cursor, Github copilot etc. made AI generate over 40% of new enterprise code. The role of “software engineer” seems to be shifting to “systems architect” and “AI supervisor.” Or even the “HR for software agents” as Jensen Huang memorably put it in his 2025 CES address.

5. The “Atoms for Algorithms” Movement: Big Tech signed historic deals with nuclear energy providers to power gigawatt-scale data centers, validating nuclear as AI’s battery. Microsoft struck a landmark $1.6 billion deal with Constellation Energy to restart Pennsylvania’s Three Mile Island Unit 1 by 2027. Amazon moved aggressively with $1 billion in commitments, including a 1.9-gigawatt power purchase agreement with Talen Energy’s Susquehanna plant. Meta committed to sourcing 1.1 gigawatts from Constellation Energy’s Clinton Clean Energy Center in Illinois. Google partnered with startup Kairos Power to deploy small modular reactors. Meanwhile, China went all in, building more nuclear capacity than the rest of the world combined: as of 2025, China had 32 reactors under construction—more than half of the total worldwide—and 102 total nuclear units (operating, under construction, and approved) with 113 GW of installed capacity, ranking first globally in overall nuclear scale. My take on this is that AI is bringing nuclear power out of the cold, and even if the AI bubble sputters out, as some pessimists claim, it will leave behind a rejuvenated, strong nuclear energy infrastructure.

6. 2025 belonged more to China than any other country in the world. China focused on applications, while its US competitors trumpeted AGI. It created massive clean energy capacity, to make energy ‘free’ or a non-issue for its AI companies, while the US scrounged for power. It made AI chips a national mission, and is fast catching up with the US. More than anything else, it quietly became the leader in open source AI, with models like DeepSeek V3/R1 and Qwen 2.5 Max which are credible frontier-class alternatives. History tells us that the one who owns open-source rules the world, and China is best positioned here (my article on China winning AI).

7. Deepfake Democracy: 2025 elections globally faced a tsunami of hyper-realistic deepfakes, forcing social platforms to implement “provenance” watermarking standards that largely failed to stem the tide. India, UK, US, and many others fear AI subverting democracy, even as AI Slop became mainstream. I believe countries have been slow to counter this insidious threat with legal and societal measures, and hopefully 2026 brings in a raft of them.

8. The Loneliness Cure: AI Companions saw massive adoption in 2025, sparking fierce ethical debates about human-AI attachment and psychological impact on social skills. Apps like Replika, Character.AI, and newcomers like Nomi reached tens of millions of users, with some spending hours daily in conversations with AI personalities. The phenomenon raised questions if AI companionship was filling genuine human connection voids. Mental health professionals noted both therapeutic benefits for lonely individuals and concerning patterns of withdrawal from real-world relationships. Now major tech platforms have quietly integrated such features, while regulators struggle to define appropriate boundaries for these intimate digital relationships (see my article on Social AI).

9. The Data Wall: We officially ran out of high-quality public human text. Labs pivoted aggressively to “Synthetic Data”—AI teaching AI—to continue model scaling. This represented a fundamental shift in AI development methodology. Where previous models trained on human-written text from books, websites, and conversations, frontier labs now generate vast synthetic datasets using their own models, then train next-generation systems on this machine-created content. The approach raises fascinating questions about whether this creates a “Habsburg AI” problem (like the genetic issues from royal inbreeding), where models become increasingly detached from human patterns of thought and communication. Some researchers worry about model collapse—where quality degrades over generations of synthetic training. (refer to my Habsburg Internet article for deeper analysis).

10. The King is Dead, Long Live the King: 2025 started with most experts writing Google off as too slow and a has-been in the AI race, even as OpenAI, Anthropic, Microsoft and Meta were taking leaps forward. The year is ending with Google donning the crown of AI King. Google has always been the ONLY full stack AI player (chips to data to models to research to applications). It has unrivalled access to the Internet data, crushing distribution through its mail, search, mobile and video properties, the best lab in the business with Demis Hassabis and DeepMind, and now it has arguably the best AI model (Gemini 3), product (NotebookLM), and image/video generator (Nanobanana Pro). With the DoJ unshackling its regulatory chains, the AI war is Google’s to lose. What a difference a year makes!

11. AI Funding, Cross-Deals and Data Centres Propped Up the US Economy: By 2025, AI startups captured roughly a third of global VC funding, with valuations and round sizes staying frothy despite macro-jitters. The real story was AI infrastructure becoming an economic pillar. Tech companies were projected to invest $250 billion in AI infrastructure in 2025 alone, with Microsoft allocating $80 billion. This spending spree created a cascade effect: semiconductor manufacturers expanded fabrication capacity, cloud providers built gigawatt-scale data centers, and energy companies invested billions in nuclear and renewable power generation. The construction boom generated hundreds of thousands of jobs—from chip fab technicians to data center electricians to nuclear engineers. Wall Street analysts noted that AI infrastructure spending had become a measurable component of US GDP growth. By year’s end, projections suggested trillions in cumulative AI infrastructure investment by 2030, fundamentally reshaping global capital flows.

12. Smart glasses quietly became the next screen and saved Meta’s AI blushes. Even as its LLaMA open source model sank into relative irrelevance, and its AI star Yann LeCun moved away from the mothership, it is the Rayban AI glasses that became a sleeper hit, hinting at AI-first wearables as a post-smartphone category. Zuckerberg dangled hundred million dollar salaries and billions of dollars of acquihire money to corral the world’s AI talent. He has been successful in getting an AI A-team, now his Meta Superintelligence Labs has to deliver. He is doubling down on devices, with the acquisition of Limitless. Meanwhile, there are more devices coming in, the most eagerly awaited one being the OpenAI- Jony Ive one.

13. Meanwhile in India: Mission “Make in India” AI: The IndiaAI Mission (₹10,000+ Cr) disbursed its first major tranches in 2025, funding local GPU clusters and subsidizing compute for domestic startups like Sarvam AI, which focuses on India-centric language models, and Gnani.ai, specializing in voice AI for Indian languages. The mission provided critical infrastructure access to hundreds of Indian AI startups that previously couldn’t afford expensive compute, democratizing AI development across the country.

14. 2025 brought the AI world to India in unprecedented fashion. OpenAI opened its first India office in New Delhi, with a focus on education initiatives. Anthropic also announced plans for its own office in Bangalore, targeting enterprise customers. Perplexity gave its paid model free to Airtel’s 350+ million users, instantly creating India’s largest AI user base. Then came the blockbuster: Google’s $15 billion announcement for a 1GW AI data center in Visakhapatnam in partnership with Adani and Airtel. Satya Nadella swung by, promising investments in cloud and AI infrastructure. Amazon upped the ante with data center commitments across multiple Indian states. The big tech AI announcements will fundamentally reshape India’s position in the global AI infrastructure landscape.

15. Chip Fabs Rising: Meanwhile, Tata Electronics commenced production runs at its semiconductor fabrication facility in Gujarat, marking India’s tangible entry into the chipmaking supply chain essential for AI sovereignty. Micron broke ground on its $2.75 billion assembly and test facility in Gujarat. The government noted, “With a vision to build a self-reliant semiconductor ecosystem by 2030, the government has rolled out dedicated incentives under the India Semiconductor Mission, complementing the broader PLI (Production-Linked Incentive) scheme framework.”

16. DPDP Rules 2025 Made Data the New Compliance Battlefield: The Digital Personal Data Protection Rules fully operationalized India’s data law, forcing AI firms to re-engineer consent, retention and training data practices. The rules created particular challenges for AI companies using web scraping, as they had to demonstrate a legitimate basis for collecting and processing Indian users’ publicly available data. Some complained the regulations created barriers to innovation, while privacy advocates argue they were essential safeguards.

Subscribe now

When Sonali Bendre went on Instagram recently to say that a naturopathic protocol built around autophagy helped her through cancer, the internet did what it is best at. Everyone became an autophagy expert, with admiration, outrage, and scholarly pieces about quackery.

She clarified that she wasn’t offering medical advice, only sharing a very personal journey, but the word had already escaped into the wild. Autophagy suddenly entered the vocabulary of wellness influencers and WhatsApp forwards.

Underneath all the noise, though, autophagy itself is real biology; it is our cell’s house-cleaning system. Under stress from fasting, infection or damage, cells start breaking down misfolded proteins and defective organelles, recycling the parts to stay alive and reduce harm.

In nature, autophagy (from the Greek auto meaning ‘self’ and phagy meaning ‘eating’) is a survival mechanism. This process is so central to cellular health that Yoshinori Ohsumi got the 2016 Nobel Prize for uncovering its mechanisms. Done right, autophagy keeps the organism resilient, but if done wrong, it can contribute to degeneration.

From Cellular Survival to Digital Cannibalism

Now shift from cells to silicon. In AI, a disturbingly similar word has appeared: Model Autophagy Disorder, or MAD. In this case, it’s not cells but AI models eating themselves.

Over the last couple of years, AI researchers have asked a deceptively simple question: what happens if you train AI models not on fresh human data, but on data generated by earlier models, also called ‘synthetic data’? The answer, across different model families, whether it be image generators, text models, or diffusion systems, is disturbing.

At first, the models quietly lose diversity. Rare and unusual cases vanish from their outputs. If you keep repeating the process, the models eventually ‘collapse’, with their outputs converging to a narrow, bland, sometimes outright broken set of patterns, bordering on complete gibberish.

A recent UCLA essay that coined the term MAD calls this a self-consuming loop. In a sense, the model eats its own tail, like an ouroboros, the ancient symbol of a snake eating its own tail, until there is almost nothing left. The mathematics behind it is subtle, but the basic mechanism is quite intuitive.

A large model is first trained on messy, high-entropy human data like web pages, books, code, social media, and transcripts. From all of these sources, its neural network internalizes an incredibly complex probability distribution: which words tend to go together, what objects co-occur in images, which patterns are rare but possible.

When you prompt the model, it doesn’t reproduce all that complexity. It gives you a smoothed sample: a polished, higher-probability, statistically simpler remix of the data it has seen.

Now imagine training a new model on those polished samples instead of the original mess. Not only have you changed the dataset, but you have narrowed it considerably. The new model can only learn from the ‘already simplified’ view of reality produced by the old one.

Its own outputs are then even more compressed, even more average. Repeat this over several generations and you get an inexorable contraction of the distribution and eventual collapse. The snake keeps eating itself until there is only a small ring of tail left.

What Happens When AI Eats Itself

To understand the severity, we must look at the symptoms observed in experimental setups where AI models were force-fed their own generations over five loops. The results were catastrophic, manifesting in distinct patterns of degradation.

One disturbing example was the phenomenon of dissolving digits. When researchers asked models to generate handwritten numbers similar to the famous MNIST dataset, the models initially performed well.

However, after a few generations of inbreeding, the numbers morphed into indecipherable scribbles. The distinct structure of a ‘7’ or the loops of a “3” dissolved into a muddy average, losing all semantic meaning.

A similar collapse occurred in image generation, where faces did not just become unpleasant, but they became disturbingly uniform, eventually converging on a single, distorted visage - an ‘average’, a statistical ghost. In another experiment, researchers found that the ability of GPT-4 to generate executable code dropped from 53% to a shocking 10% when trained on synthetic data loops. The logic simply fell apart.

This is where the metaphor of statistical inbreeding becomes useful. When European royals, like the Habsburgs, married within their own bloodlines to “keep the power pure,” they didn’t preserve their greatness; they amplified their flaws.

The famous “Habsburg Jaw“ was a physical manifestation of a gene pool that had become a closed loop. Training models on their own outputs is the data equivalent: each generation has less genetic variety, fewer long-tail events, more fragile structure.

MAD is the Habsburg Internet.

AI models work by probability. They desire the ‘centre’ of the bell curve, which is often the most likely answer. When a model trains on the output of another model, it amplifies these central probabilities and trims the ‘tails’ of the distribution. But it is in those tails that contains the rare, the quirky, the creative, and the outliers. With each generation, the variance drops. The model becomes more confident, but less diverse. It is inbreeding data until the digital equivalent of haemophilia sets in with the slightest error becoming fatal to the logic of the system.

In theory, the fix sounds easy. Do not inbreed, keep adding fresh human data. In practice, that is where the problem starts. Several analyses have pointed out that frontier models have already consumed much of the high-quality public text and code on the internet. AII companies have been rummaging through every data cupboard they can find, but most of them are bare now.

So, they have started ‘stealing’. . OpenAI reportedly used its Whisper system to transcribe over a million hours of YouTube video and feed those transcripts to its models. Training of image and video models like Sora has raised concerns that large amounts of copyrighted content from platforms such as YouTube and TikTok may have been scrapped without clear consent.

Now, if a new model scrapes the web, it is likely scraping an article written by ChatGPT, summarized by Claude, and illustrated by nanobanana. It is a Hall of Mirrors, disfiguring the output every time.

We have seen this movie before - with ‘free’ social media. As the oft-repeated saying goes: if you are not paying, you were the product. In the AI era: if you’re not paying for the AI, you and your behaviour are its training data. That would still be tolerable if our behaviour remained independently human.

But we are already letting AI shape what we read, watch, write and buy. Recommendation systems nudge our playlists and viewing lists, Copilot-like language models co-author our emails and reports, and AI-generated “slop” fills more of the web every week.

Put these two loops together and you get a nasty feedback system. First-generation models train on a mostly human internet. Those models then influence human expression, which becomes more AI-shaped. Everything starts averaging out as styles converge, images standardise, and code snippets, boilerplate and even legal clauses start to look alike.

The next generation of models trains on this partly synthetic, partly AI-influenced corpus. They are not just learning from us, but also from a world already bent towards them. The human is technically ‘in the loop’, but in reality we are amplifying the loop that leads to model autophagy.

AI learns from a human world that has effectively learned from AI. This is the dangerous feedback loop. If the model says “the sky is neon green” often enough, and humans start repeating it because they trust the model, the next model will treat ‘neon green sky’ as an established truth!

The risks are not just aesthetic. If models collapse, they don’t just produce boring images and generic prose, but they lose touch with the rare, unexpected, minority pattern. Those are exactly the edge cases that matter most when you use AI to triage patients, screen loans, or detect fraud.

Bhasmasura’s Digital Curse

Can we prevent MAD? Well, people are trying. Researchers are trying to develop tools to measure the geometry of data manifolds and detect when the training distribution is drifting away from properly diverse human data. The UCLA group uses the shower metaphor. You turn the tap towards hot, scald yourself, correct back towards cold, overshoot again, then finally settle into a comfortable middle.

Training pipelines can do something similar by monitoring how synthetic and narrow the data has become and actively dialling back, injecting fresh, well-sourced human data as an anchor. Other approaches focus on cleaning human data rather than replacing it, and on tracking provenance so that we know what fraction of a dataset is AI-generated and can cap its influence.

But all of this is futile if the social and economic incentives point in the opposite direction. If every publisher is rewarded for flooding the web with cheap AI content, if every platform optimises for engagement rather than depth, and if every organisation treats AI primarily as a way to remove humans from the loop rather than upgrade them, then the technical fixes are akin to a shower thermostat in a house ablaze.

This reminds me of the original MAD, Mutually Assured Destruction, in the nuclear arms race. That surprisingly effective doctrine said that if both sides know that pressing the button destroys everyone, that knowledge will paradoxically keep the peace.

Model Autophagy Disorder is a quieter, subtler version of the same structural risk. This is not as much about blowing up the world, as it is about blurring it. If Model Autophagy Disorder is not checked, we do not risk a nuclear winter, but a true winter. A dystopian future where history, culture, and creativity are smoothed over into a dull beige, hallucinated, statistical average.

Indian mythology speaks of the story of Bhasmasura, the demon who was granted the power to turn anyone to ash by touching their head. In his hubris, he was tricked into touching his own head, consumed by the very power he sought to wield. AI, in its hunger for infinite data, risks becoming Bhasmasura, burning itself out by consuming its own tail.

Subscribe now

Dharma is one of those words that refuses to sit quietly in a dictionary. It can mean duty, law, righteousness—but underneath all that, it points to something deeper: the essence of a thing, and the order it belongs to. The dharma of fire is to burn, of water to flow. Their dharma is not a moral rulebook; it is their nature, the pattern they express in the world.

It has been translated as duty, righteousness, law, religion, teaching—but all of these are only partial reflections. From the root dhṛ, “to hold or sustain,” dharma is that which keeps things from falling apart: the pattern, the essence, the order within which life, society, and even thought can make sense.

In the Bhagavad Gita, Dharma is not just a philosophical abstraction, but a here-and-now problem and dilemma. Arjuna stands on the battlefield paralysed between conflicting obligations. Krishna does not give him a generic moral checklist or a black-and-white answer. Instead, he points him to his svadharma, or his own particular duty and essence as a warrior, a leader, and a part of a larger cosmic order. “Better one’s own dharma, even if imperfect, than another’s perfectly performed.” Dharma here is both what you are and what you must do, given who you are and where you stand.

In Buddhism, Dharma takes on another nuance. Not only is it the Buddha’s teaching but also the law of reality itself - the way things truly are beneath our stories and illusions. To ‘see the Dharma’ is to see clearly: impermanence, interdependence, the patterns out of which our experience is built. In some Buddhist schools, dharmas are literally the smallest units of reality, the building blocks of mind and matter.

From Ancient Wisdom to Artificial Intelligence

Across these traditions, dharma fuses three ideas: essence, order and responsibility. It is the nature of a thing and the web of relations and dependencies it lives in.

We are now unleashing Artificial Intelligence into every facet of human life. AI is not just a clever piece of software running in a data centre. It is starting to shape how societies work, how geopolitics plays out, how economies allocate power and wealth, how culture is produced, how language and knowledge are created and consumed. It is beyond a tool, more like a system or even an organism.

So the question that interests me is: what is the dharma of AI?

At a narrow technical level, AI’s dharma seems simple: recognise patterns, predict the next token, or optimise a workflow. But as AI is plugged into education, warfare, finance, governance, social media and everyday life, something much larger emerges. We get an AI-shaped order: new ways of deciding who gets a loan, what news we see, how we work, how we wage war, how we imagine success, even how and with whom we fall in love.

Like Arjuna, we now stand in the middle of a field shaped by forces we helped create but do not fully understand. We cannot walk away from AI, it is here now, whether we like it or not; but, nor can we blindly surrender to it. Our challenge is to understand its dharma and our own: what is the essence of this technology, what kind of world does it tend to produce, and what are our particular responsibilities— as CEOs, policymakers, engineers, workers, home makers students, citizens? As homo sapiens sapiens.

There is also a geopolitical dharma at play. For the first time, the balance of power between nations is being reorganised not just by land, labour or oil, but by access to data, algorithms and compute. A handful of countries and companies control the largest models and the biggest data centres. The rest provide data, markets, talent, regulation—and often bear the downstream effects. The Global North writes most of the code and the narratives, but it is the Global South that lives much of the reality.

India’s Moment: Beyond Silicon Valley

Can India and other emerging economies articulate a distinct Dharma of AI, one that reflects our pluralism, our development challenges, our sensitivities about work, education and dignity, rather than merely importing Silicon Valley’s dreams and Brussels’ anxieties? Dharma in this sense is not only an ethical question, but is a strategic one: how do we align this general-purpose technology with the actual lives and aspirations of billions of people across the whole globe, especially the forgotten portions of it?

Buddhism’s sense of Dharma, as the law of reality and the path to seeing it clearly, matters here too. AI is surrounded by hype and fear: Godlike saviour on some days, apocalyptic threat on others. To work with it wisely, we need something like Dharma vision: the ability to see what these systems really are, how they are trained, where the power sits, what incentives drive them, what they can and cannot do. Without that clarity, we either worship or panic. Hopefully, by knowing the essence we will decide and choose more wisely.

AI’s Mirror: Seeing Ourselves Clearly

There is also another subtler “dharma” at play: what AI reveals about us as humans. When a model predicts our clicks better than we understand our own habits, when it fakes a video more convincingly than our memory can resist, when it drafts better emails than we care to write, it holds up an uncomfortable mirror. What is the essence of human judgment, creativity, responsibility, once machines can approximate them?

The Dharma of AI is my attempt to explore all of this: not just AI’s ethics, but its essence and influence on work and business, on society and geopolitics, on culture, story and meaning. I will write about how AI is changing offices and factories, elections and media, classrooms and households. I will bring in mythology and history, boardrooms and bazaars, India and the wider world. Sometimes I will be optimistic, sometimes sceptical, but always trying to demystify rather than shroud AI further in mystery.

This is not a doctrinal blog that claims to know the one correct path. It is, in the Buddhist sense, a Dharma path of inquiry: a continuing effort to see more clearly what AI is, what world it is quietly building around us, and what roles we might each play within that new order.

If dharma is that which holds things together and reveals what they truly are, then this blog is my ongoing inquiry into what AI really is becoming among us – and what kind of world it is quietly building.

Subscribe now

Ok, Sam Altman and OpenAI are back to headline Weekend Warp again. But, not in the way Altman would prefer.

Stung by the nanobanana-powered strike back by Google, and a focused profit-led strategy by Anthropic, Sam Altman declared a Code Red at OpenAI (Read more). In a memo to his team on Monday, he said “We are at a critical time for ChatGPT,” acknowledging that more work was needed to enhance ChatGPT’s speed, reliability and personalization features. To focus on this “code red” effort, OpenAI will be delaying work on advertising, AI agents for health and shopping, and a personal assistant called Pulse (Read More). The company is also planning daily calls among those responsible for improving ChatGPT—a level of urgency that speaks volumes about the competitive heat OpenAI is feeling. Ironically, Google had declared a Code Red earlier, when even Sergey Brin came back to literally code his company out of AI wilderness. I have always maintained that the AI war is Google’s to lose (see my Mint article on that). They are the only full stack play out there - from NVIDIA-threatening chips, the best AI lab with Deepmind, the best engineers led by Nobel Laureate Demis Hassabis, dominating distribution, and, most importantly, the best world data with search, YouTube and everything else. They own the Internet, and it is now up to them to ‘own’ AI too.

One of the ways OpenAI is coming back is through a flurry of announcements. If GPT-5 was code-named Strawberry, the next one, it seems, is called Garlic. According to The Information, OpenAI is working on a new model codenamed “Garlic” (Read more). The breakthrough? OpenAI has apparently figured out how to infuse a smaller model with the same amount of knowledge it could previously only get from developing a much larger model. So less GPUs, less cost—a critical advantage when you’re burning through capital at OpenAI’s rate. Even more urgent: GPT-5.2 is being fast-tracked to launch as early as December 9 (More here)—next week!—originally scheduled for mid-December but pulled forward to counter Google’s momentum. The model wars are heating up, and the release cadence is getting frantic!

Another announcement is where Jony Ive and Sam Altman say they finally have an AI hardware prototype. For context: In May 2025, OpenAI acquired Ive’s design startup “io” for $6.4 billion (Read more), bringing the legendary Apple designer (who created the iPhone, iPod, and iPad) on board to reimagine what an AI-first device could be. The partnership was shrouded in mystery until last week’s reveal at the Emerson Collective Demo Day. The Verge reports (more here) that “little has been revealed so far about the OpenAI device in development, but it’s rumored to be screen-free and roughly the size of a smartphone.” Altman described the design as “simple and beautiful and playful,” (Read More). It would still take about two years more to come out though, so it’s a while before we get our itchy hands on it.

Bonus: Anthropic meanwhile is sitting pretty. Its founder Dario Amodei sneakily suggested at the NYT DealBook Summit that “Anthropic does not need a Code Red.” (Read more). This is because they are super focused on GenAI for the Enterprise, and spending less while creating more. “We have a little bit of a privileged position where we can just keep growing and just keep developing our models,” Amodei said (read here), adding that Anthropic has issued no “code reds.” Investors are warming up to this disciplined strategy more than OpenAI’s random walks, and it’s possible that Anthropic will IPO sooner than OpenAI, with valuations projected to exceed $300 billion. (More here)

Double Bonus: Perplexity created its own news this week, with founder Aravind Srinivas announcing that the football GOAT Cristiano Ronaldo is an investor and the face of Perplexity. Ronaldo posted on X (Check here) saying he was “proud” to announce his investment, describing curiosity as “a requirement for greatness.” Srinivas welcomed the partnership on Instagram (More here), calling it an “elite collab” and praising Ronaldo as “the GOAT—he’s relentless and constantly researching the latest techniques.” The partnership includes launching the “Ronaldo Hub”—a custom AI assistant featuring unseen images from Ronaldo’s personal archive, career statistics, interactive goal replays, and curated Q&A about his two-decade career. With Ronaldo’s millions of social media followers, this gives Perplexity massive distribution in Latin America, the Middle East, and Asia. (Side snarky comment: Perplexity, by the way, still hedges on who is the GOAT - Ronaldo or Messi. If Srinivas was Musk, he would have fixed that algorithm by now 😉)

In this week’s AI Tool of the Week, we look at a technology designed for one of the biggest challenges of the AI era: knowing what is real and what is generated. Google DeepMind’s SynthID adds invisible watermarks to AI-created images, giving users a way to verify origin and maintain trust in visual content.

Full use case here

Subscribe now

[This was an article published earlier on my blog]

AI has been with us for more than sixty years but is only in the post-ChatGPT years that we have started hearing a lot of talk about the ‘age of AI.’ The emerging narrative is that the age of AI is dawning upon us, and we humans need to figure out how to survive and thrive in this age where AI dominates. I have been as guilty of this refrain, arguing that as AI becomes more powerful, we humans will have to rediscover our human traits of curiosity, passion, and connection, rather than becoming a more AI like automaton. While I still stand by this assertion, I want to question this narrative of the ‘age of AI’ - the idea that the machines are rising, and our struggle is to preserve our relevance as humans.

What if we have got it the wrong way around?

This may be the age of powerful large language models and advanced ‘thinking’ and ‘reasoning’ algorithms, but it is still, fundamentally, the age of humans. We create the AI models and derive meaning, write the guardrails and regulations, and face the consequences, if we do not do it right. AI does none of these things independently. To imagine a future where AI ‘replaces’ us is to misunderstand both the role of technology and the adaptability of human beings.

Instead, we might be better served to ask a more provocative, but more relevant, question: How should AI behave in the age of humans?

By flipping the frame, we recognize a deeper truth. It is AI, which is the new kid on the block; we humans have been inhabiting it for millions of years. Just as a new entrant into a community or club needs to learn the rules of coexistence, so too must AI systems learn to live within the bounds of a human world. Thus, the real question is not about how we adapt to AI—but how AI must adapt to us.

In that spirit, I have attempted a Manifesto on how an incoming AI can be relevant in the Age of Humans

• Be Trustworthy: If AI is to be truly become a citizen of the world of humans, it must be trustworthy, law-abiding, and a contributing citizen, aligning with human values and norms. Humans are building these models, and we must ensure that they should be truthful, not prone to fabrication or misinformation. Much like humans, they might pretend some of the time but should learn to say “I don’t know” when appropriate and provide transparency in how they arrive at their answers. We cannot afford black boxes all the time.

• Be Fair: Moreover, AI must strive to be fair and unbiased. In a human world, fairness is not optional—it is essential. AI must be designed to recognize its limitations and adjust for historical, societal and cultural contexts that affect biological humans. No AI can be completely unbiased, like no human can. However, as the best of us try to listen to all sides and arrive at a view, so should AI.

• Be Kind to the Environment: We now know that the carbon footprint of training the world’s AI is not trivial - requiring gallons of clean water and gigawatts of electricity, while spewing out country-level CO2. As AI becomes more ubiquitous, sustainability must be built into its architecture with smaller models, greener energy, cleverer algorithms, and on-device compute. If AI are to cohabit with humans, it must have a planet to do so.

• Be Dignified: This is a subtler principle - AI must preserve human dignity. It should not manipulate emotions, nudge behaviours covertly, or remove choice under the guise of convenience, everything where social media failed at. Even as AI Agents come in and live with us in our teams at work or with our families at home, their role be to augment human agency and dignity, and not to erode it.

• Be Humble: Finally, AI must learn to live with humility. AI must be able to adapt to the changing without overconfidence, and have the humility to be updated, corrected, and overruled by human judgment. This becomes more important if Artificial Super Intelligence (ASI) emerges; it is only its built-in humility which could make us co-exist and thrive.

In this coming age of humans with AI, both of us do not compete for supremacy, but collaborate in partnership. Humans should recognise that AI can be an extraordinary assistant and a second brain for us, a gift of superintelligence if harnessed right. AI, on the other hand, must accept that human judgment, with all its flaws and wisdom, remains at the centre of the system. This is not a romantic view of human exceptionalism, but a clear-eyed recognition that technology, even if highly intelligent, does not replace the complexity of human life.

So, the real question may not be how we humans survive the rise of AI. It is how AI learns to thrive among us humans on our terms. The other question, more troubling, is that it is us humans that will or will not shape this future coexistence, and therefore what is of more concern is not artificial intelligence, but human stupidity.

Subscribe now

A shell-shocked Sam Altman and OpenAI tried capturing the initiative last week. And failed.

As I had written last time, the AI winds seem to be shifting Google ‘s way – with Gemini 3, nanobanana pro, and its TPUs (Tensor Processing Units) finally giving some solid competition to NVIDIA‘s monopolistic GPUs. This is now reflected in their mutual stock prices as Alphabet Inc.’s stock surge is reshaping the list of the world’s most valuable companies (Read more). The Mountain View firm jumped 2%, pushing its valuation to $3.9 trillion. Since mid-October, the shares have climbed 37%, adding roughly $1 trillion. The market cap gap between these two AI titans has collapsed: Alphabet now sits about $300 billion shy of Nvidia’s $4.2 trillion market cap. Alphabet also seems to be eating ChatGPT’s market share as Google Gemini has more than doubled its web traffic share over the past year, climbing from 5.6% to 13.7%, while ChatGPT’s dominance eroded from 86.6% to 72.3% (Read more). And that has Altman worried. It reflected in an internal memo he sent his team (Read more) acknowledging the “rough vibes” and noted they “need to stay focused through short-term competitive pressure”. However, Altman is ready to face Google’s challenge -- OpenAI is apparently working on a new language model codenamed “Shallotpeat” to fix pre-training bugs and catch up.

This is perhaps why OpenAI made a couple of announcements last week, more features than big developments though. One was its launch of group chats in ChatGPT, letting up to 20 people join the same conversation with the AI. The feature is now available to all logged-in users worldwide after a short test earlier this month. OpenAI sees this as a way to plan things together—dinners, trips, shared projects—with ChatGPT helping in the background. ChatGPT is trained to follow the flow of the conversation. It will only step in when it thinks it’s helpful, but you can tag “ChatGPT” if you want a direct reply. This is quite cool, and should be giving WhatsApp Groups some reason to pause? (see more here). The other one was a bigger announcement in my view: It has rolled out an upgraded version of its ChatGPT shopping assistant, adding research capabilities designed to offer users more detailed and personalized product suggestions. The update arrived in time for the Black Friday sale, one of the year’s biggest retail events, and aimed at expanding the tool’s ability to search for well-matched products and refine recommendations through conversation. Bigger, because it adds fuel to the inexorable move from e-commerce to a-commerce (or agentic commerce), something which I wrote on recently here (Here is my ET article).

Closer home, all eyes on Vizag. Last month, Google raised eyebrows (and land prices) as it committed about $15 billion over 2026–2030 to build its first India AI hub and a gigawatt-scale data-centre campus there, in partnership with Bharti airtel and AdaniConneX (Adani + EdgeConneX). This would entail an investment of $5 billion by Adani Group, through AdaniConneX. Not to be left behind, Reliance’s Digital Connexion – a JV of Reliance Industries Limited, @Brookfield and Digital Realty – has announced around $11 billion (about ₹98,000 crore) capex by 2030 to build another 1 GW “AI-ready” data-centre cluster in Visakhapatnam on roughly 400 acres! (see more here). Vizag seems to be emerging as the AI hub, or at least the data centre hub, of India, pushed by its Chief Minister Chandrababu Naidu Nara, who wants to do a ‘Hyderabad’ to Vizag. And Naidu knows a thing or two about building tech hubs. In the late 1990s, as Chief Minister of Andhra Pradesh, Naidu transformed Hyderabad from a second-tier Indian city into a global tech powerhouse. His tactics were audacious: he did a 45-minute chat to convince Bill Gates to open Microsoft‘s first development center outside the U.S., built HITEC City in just 14 months, and marketed Hyderabad with the slogan “Bye-bye Bangalore.” President Bill Clinton visited HITEC City in March 2000, calling Naidu’s government “smaller, far less bureaucratic, and far more market-oriented.” Now, Chandrababu is attempting the same magic with Vizag. Where HITEC City cost $350 million, Google’s Vizag AI hub represents $15 billion. Naidu promises what took 25 years in Hyderabad can happen in a decade with “quantum-speed execution.” His son Lokesh Nara, now IT Minister, declared: “Vizag emerges as the Data Capital of India!”

Will Chandrababu pull it off again? Time will tell, but the MoUs are signed and the shovels are ready!

In this week’s AI Tool of the Week on Mint, we’re looking at a feature that helps job seekers turn text-heavy CVs into visual resumes and slide decks — without needing any design tools.

If your résumé looks like everyone else’s, this update may be the edge you need.

Read more.

Subscribe now

[This article was first published on aiandbeyond.ai]

Last weekend, the spotlight was on Manus, a Chinese AI super-agent reportedly capable of handling up to fifty complex tasks simultaneously. Videos of it seamlessly managing transactions while conducting deep research, launching websites, and even planning vacations went viral. This development isn’t just a wake-up call for AI labs like OpenAI—it also raises unsettling questions for us humans. Are we already being outpaced by AI?

Parmy Olson recently wrote about a Microsoft-CMU study examining how 319 knowledge workers interacted with AI. A surprising takeaway was that as they placed more trust in AI for skills like writing, analysis, and evaluation, they practiced those skills less themselves. Over time, they became passive recipients of AI output, rarely questioning its accuracy or refining their own judgment. This fuels a deeper anxiety—if AI keeps advancing, will we gradually lose not just our jobs but also our ability to think critically and creatively?

Fear of technological displacement is nothing new. Socrates worried that writing would weaken human memory. When calculators emerged, people feared they would erode arithmetic skills. Computers were once predicted to render millions of knowledge workers obsolete. Yet, despite these fears, technology has always integrated into our lives, shifting how we work rather than eliminating our purpose. Now, with frontier AI labs racing toward AGI—Artificial General Intelligence—the stakes feel higher. Futurist Ray Kurzweil initially predicted that AGI, or the Singularity, would arrive in 2045. The acceleration of generative AI has prompted some to revise that timeline, suggesting it could happen as soon as 2026, or within the next five years.

In reality, we may not need to wait until 2045—or even 2026. AI is already reshaping our world, not in a sudden, dramatic takeover but through a slow, incremental process. Imagine the proverbial frog in a pot of gradually warming water, unaware of the rising temperature until it’s too late. That’s us with AI.

Fifteen years ago, we memorized phone numbers; now, our phones do it for us. We used to navigate from place to place using memory or maps; now, we blindly follow GPS directions. My phone reminds me when to leave for my physiotherapy sessions, and soon, an autonomous car might pick me up without my intervention. As smart homes evolve, AI-powered appliances will anticipate and fulfill our needs—our blender preparing our protein shake, our microwave and fridge coordinating meals. The convenience is undeniable, but the trade-off is clear: we risk losing basic skills and autonomy in the process.

So, what can we do as AI becomes increasingly powerful, independent, and intelligent? Three things:

1. Contain AI: We must establish rules and safeguards to ensure AI remains beneficial and aligned with human values. Mustafa Suleyman, Microsoft’s AI chief, highlights this in his book—traffic, for example, could be chaotic and deadly without signals and driving laws. Similarly, AI needs regulatory frameworks to prevent unintended consequences.

2. Become AI-literate: AI is here to stay, and we must learn to navigate it. Just as previous generations adapted to new languages and mathematical systems, we must integrate AI fluency into our skillset. This means understanding how AI works, using generative AI tools effectively, and collaborating with AI as we would with human colleagues.

3. Emphasize human strengths: As AI takes over more tasks, our uniquely human abilities become even more critical. Skills like curiosity, compassion, language, logic, intuition, and collaboration will define our relevance. We must learn to curate and interact with AI thoughtfully—asking the right questions, applying judgment, and refining AI-driven solutions. Our ability to work with others, exercise ethical reasoning, and explore creative alternatives will be indispensable. The more AI advances, the more human we must become.

The rise of AI doesn’t have to mean the decline of human agency. But if we want to remain active participants in our future, we must consciously shape our relationship with AI rather than passively letting it shape us.

Subscribe now

Despite Nvidia defying gravity once again, this week belonged to Google. And how.

The 800 pound gorilla is so back, and it is pounding the others into the dust. The triumphal launch of Gemini 3, along with some eyepopping upgrades to NotebookLM and nanobanana, finally showed what Google was capable of in AI. Gemini 3tops the LMArena Leaderboard with a breakthrough score of 1501 Elo. It demonstrates PhD-level reasoning with top scores on Humanity’s Last Exam (37.5% without the usage of any tools) and GPQA Diamond (91.9%). It also sets a new standard for frontier models in mathematics, achieving a new state-of-the-art of 23.4% on MathArena Apex. Beyond text, Gemini 3 Pro redefines multimodal reasoning with 81% on MMMU-Pro and 87.6% on Video-MMMU. It also scores a state-of-the-art 72.1% on SimpleQA Verified, showing great progress on factual accuracy. If this is all LLM gobbledygook (or Googledygook, for that matter to you), it basically means that Gemini 3 has blown past every other model in multiple performance parameters. This also means Gemini 3 Pro is highly capable at solving complex problems across a vast array of topics like science and mathematics with a high degree of reliability. (see more here) NotebookLM, which I have always argued should be the hero product at Google, is now also powered by Gemini 3 and has some amazing new infographic and other capabilities; see what it did to my LinkedIn profile, for example!!

And, if that was not all, the world is going gaga over the stupendous capabilities of nanobanana, which seems to leave even the might Sora 2 in the dust (see more here). Well, I supposed Sam Altman is having some sleepless nights, even though he was gracious enough (along with Elon Musk) to post a congratulatory message to Google (see here). I have always maintained through numerous articles of mine is the Google is the ONLY full stack AI player out there, owning everything from chips to apps, along with the best datasets, and it is a matter of time that they get their act together (more here).

Well, the act is together.

The Google launch overshadowed even the bubble-defying results of NVIDIA, as it singlehandedly quashed the AI Bubble fears, at least for a day, before they tentatively emerged again. It just reported another blockbuster quarter. Revenue hit about $57 billion, up more than 60% from a year ago, and profit was roughly $32 billion, about two-thirds higher than last year. Its quarter-on-quarter (sequential) revenue growth of 22% in Q3 2026, increasing revenue from about $46.7 billion in Q2 to $57 billion in Q3. Most of the money comes from data centers, where big tech companies use Nvidia’s GPUs to run AI models. The company expects sales to climb further to around $65 billion next quarter, suggesting demand for its AI chips is still very strong.This is a stupendous growth from a company that large, and CEO Jensen Huang crowed about the fact that everyone is talking about an AI bubble, but they are seeing something else. The markets again displayed their nervousness the day later, reflecting in Nvidia stock plummeting again. Strange are the ways of the market, more driven by sentiment than by performance. Maybe, someone noticed that the aforementioned Gemini 3 was built entirely on Google’s own Tensor Processing Units (TPUs), not on Nvidia chips or rented infrastructure.

See more here

Meanwhile, causing perhaps further consternation to Altman, Nvidia teamed up with Microsoft and arch-rival Anthropic to launch their own alliance in the sea of alliances that the AI world is seeing, most of them largely stitched by OpenAI. They formed a strategic alliance worth up to $15 billion in investments—Nvidia up to $10 billion and Microsoft up to $5 billion. Anthropic will scale its Claude AI models on Microsoft Azure using Nvidia’s cutting-edge AI hardware, committing to spend $30 billion on Azure compute capacity. This collaboration includes co-designing AI model architectures and optimizing performance and efficiency together. Microsoft is also expanding Claude’s integration across its Copilot product family, ensuring wider enterprise access, while Nvidia and Anthropic collaborate on future GPU designs tailored to Anthropic’s workloads. (See more here). So, finally, Microsoft is diversifying decisively away from OpenAI. I believe this is a more natural fit for them, given that Anthropic is sharply focused on the enterprise customer, which is Microsoft’s sweet spot too.

Bonus: the restless Jeff Bezos launched a new AI company, Prometheus, funded it with a measly $6.5bn and installed himself as the co-CEO. I guess he was missing all the AI action. (more here)

In this week’s AI Tool of the Week, we’re looking at a feature that changes how you navigate, explore, and stay aware on the road — without ever touching your phone.

It takes Google Maps beyond coordinates and distances, and turns it into something far more natural and context-aware.

Read more

Subscribe now

Most of us first encountered artificial intelligence in fiction—whether through the relentless pursuit of the Terminator, the moral dilemmas of Asimov’s Three Laws, or the charming automatons of Japanese animation. These imagined machines shaped our earliest expectations of AI, long before it took its real-world form (https://bit.ly/3EPXA4f). When AI finally arrived in earnest, it did not come as humanoid robots but as invisible algorithms. It slipped into our lives through recommendation engines, transformed biology with DeepMind’s protein-folding breakthroughs, and then spoke with near-human fluency in ChatGPT.

Yet, curiously, AI now circles back to its robotic origins. For years, the dancing machines of Boston Dynamics captured public fascination on YouTube, but today, the robotics industry is experiencing a surge in investment. Nvidia has backed Field Robotics, valued at $2 billion. Skid, another startup, recently secured SoftBank funding at a $4 billion valuation. Meanwhile, Physical Intelligence, with its pi-zero robot, is valued at half that. The movement back toward embodied AI was underscored by Nvidia’s founder, Jensen Huang, in his CES keynote, where he outlined AI’s trajectory: from Perception AI to Generative AI to Agentic AI, and finally, to Physical AI. What is striking about this progression is how closely it mirrors human development.

For the first five decades of AI, the focus was on perception—teaching machines to recognize patterns, interpret sensory data, and make sense of raw information. This stage is akin to infancy, where a human baby learns to distinguish faces, voices, and objects in a world of overwhelming sensory input. Predictive AI—the kind that powers facial recognition and personalized content recommendations—serves as the eyes and ears of machines. It identifies, categorizes, and anticipates, laying the groundwork for more advanced cognition.

Then came Generative AI, the counterpart to early childhood creativity. Much like a child who starts drawing, storytelling, or playing music, AI evolved to generate content rather than just interpret it. Neural networks began producing text, art, videos, and even code. Where Perception AI passively observed, Generative AI became an active participant in creation. It democratized digital artistry, allowing anyone to craft images, compose music, or write poetry with an AI collaborator at their side. This was a leap from mere recognition to imagination.

The next phase—Agentic AI—mirrors adolescence, a time of learning to reason, make decisions, and take initiative. Unlike its predecessors, Agentic AI is not just a tool but an active problem solver, capable of setting goals, making strategic decisions, and adapting to its environment. It moves AI from passive assistance to proactive engagement, akin to an intern who anticipates needs and acts accordingly. In this phase, AI is no longer just following commands; it is thinking through solutions, optimizing workflows, and providing intelligent responses with contextual understanding.

The final frontier is Physical AI—the adulthood of artificial intelligence. It is here that intelligence, once confined to data and algorithms, manifests in the real world. Robotics infused with AI become not just reactive tools but independent actors, making real-time decisions in complex environments. Think of autonomous surgeons performing intricate operations, self-driving vehicles maneuvering through unpredictable cityscapes, or robotic farmers tending to crops with surgical precision. This is intelligence given form—machines that do not just compute but act, shaping the world around them in tangible ways.

Jensen Huang envisions Physical AI as a multi-trillion-dollar industry, one that surpasses even the wildest expectations set by previous AI breakthroughs. But beyond the economic impact, this evolution tells a deeper story: AI is not replacing human potential; it is amplifying it. Just as the smartphone expanded human communication and knowledge, AI is extending our cognitive, creative, and physical abilities.

The roadmap Huang outlined is more than a technological forecast—it is a reflection of human progress itself. Intelligence, once the sole domain of biological beings, is now emerging in machines that can see, create, think, and act. AI is not an alien force; it is intelligence taking a new form, moving forward on a trajectory that, in many ways, feels familiar. Just as we grow, so does AI—learning, evolving, and stepping out into the world.

Subscribe now

So, who knew that the simple and staid Sundar of Google had got Elon- envy?

Even as Elon sends a rocket up almost every other week, loaded with STARLINK satellites and more, Google wants to send an entire data centre there! Google’s answers to NVIDIA’s GPUs, called TPUs, or Tensor Processing Units, floating around in space! The idea behind Project Suncatcher (as opposed to Moonshots), as the project is called, is to harness the nearly unlimited power of the sun to power data centres. By now, we all know that the biggest limiting factor to build more and more ginormous data centres is the scarcity of clean, plentiful power. Google decided to take a leaf out of the starship book and go directly to the source. Microsoft had earlier gone the other way – below the ocean, so that the white hot chips would get cooled automatically by the water, saving massive amounts of energy in the process. As Pichai proudly proclaimed posting: “Project Suncatcher is exploring how we could one day build scalable ML compute systems in space, harnessing more of the sun’s power (which emits more power than 100 trillion times humanity’s total electricity production).Like any moonshot, it’s going to require us to solve a lot of complex engineering challenges. More testing and breakthroughs will be needed as we count down to launch two prototype satellites by early 2027, our next milestone of many. Excited for us to be a part of all the innovation happening in (this) space!”

More here

Well, even as the rockets went up, stocks went down. Wall Street has been holding its breath on AI’s colossal capital expenditure, even greater hype and some disappointing noises by brave enterprise customers saying that stuff still does not work as well as they thought. However, Big AI has been bucking these trends so far, throwing out trillion dollar valuations for its companies. While the Big Crash is still awaited, a small one did happen last week. Big tech stocks experienced a significant downturn recently, with the tech-heavy Nasdaq posting its worst week since April 2025. Major technology companies, the “Magnificent Seven” were especially affected, leading to broader market losses. Nvidia dropped over 9.5% in the past week, its sharpest decline since April 2025.; Meta and Microsoft both lost more than 4% over the week, declining 2-3% just on Friday; Tesla shed about 3.5% Friday morning, and Alphabet Inc. and Amazon each fell more than 1% for the week, though Apple slightly bucked the trend with a modest gain of 0.2%. Sign of things to come, the small tremors before a major earthquake? Or just some profit booking? Time will tell…

More here

Meanwhile the reigning Emperor of AI – no, not Sam Altman, but Jensen Huang – startled everyone as he went on to declare to the Financial Times that “China has won the AI race.” He cited China’s lower energy costs and fewer regulatory hurdles as key advantages. He is right, at least in the energy part. China is producing more clean energy than the rest of the world, and the US policies are not helping its own companies. Probably, Huang was frustrated since Trump has blocked selling the best Nvidia GPUs to China. In any case, he later softened his remarks, clarifying on social media that “China is nanoseconds behind America in AI” and emphasized the importance for the U.S. to “race ahead and win developers worldwide.” (Read more here). Whatever Huang thinks, I am swinging to the view that China is better positioned to own the future, with strong leadership in clean energy, EVs, Drones, Robotics, and, yes, AI. I wrote on that recently for The Economic Times (here)

Pleased to share that my book “Winning With AI: Your Guide To AI Literacy” coauthored with Anuj Magazine and published by Juggernaut Books is now available wherever Amazon is!

Love it or hate it, it is clear that AI is a generational technology, which will impact and transform everything – jobs, business, education, life, society, geopolitics, the planet, even the human species.

In a panel yesterday, I spoke about how everyone does not need to be an expert in AI, but definitely needs to be literate in it. The definition of literacy will change from reading writing, arithmetic to all of these plus how you use AI innately and instinctively in everything you do.

The best way to get AI Literate is to pick up our book, which makes you one in five easy steps.

Buy Now

In today’s AI Tool of the Week, we’re looking at a breakthrough for analysts and investors who spend hours combing through earnings call transcripts — Perplexity Finance.

It lets you query, compare, and extract insights directly from company earnings calls — from strategic priorities to AI investments — all in seconds. No more digging through 50-page documents to find what matters.

Read more

Subscribe now