Full transparency on the inputs, the instructions, and every task performed — so nothing is hidden behind the magic curtain.
A raw WhatsApp group chat transcript was pasted directly into Claude — unformatted, full of abbreviations, incomplete sentences, typos, and the perfectly natural chaos of real people mid-thought. No editing beforehand. No structure. Just the chat, as it happened.
The group: school friends, described by the person sharing this as "a very mixed breed of people" who "hardly have discussions." Which made this particular one unusual enough to be worth preserving properly. The topic that broke the silence: Artificial Intelligence, India's readiness for it, and whether the government's enthusiasm is substance or spectacle.
A real conversation between people who know things — lightly edited for clarity, with all the original frankness left exactly where it was.
Let me lay out where I think we stand:
Two things changed recently. First, the availability of very large GPUs from Nvidia made computation on large datasets cheap and accessible at scale. Second, natural language processing received a transformative boost from a paper by Google researchers — "Attention Is All You Need", led by Ashish Vaswani. The Transformer architecture in that paper is what enabled OpenAI — and every company that followed — to build the current generation of Large Language Models.
Near-term impacts? All processes will get defined, workflows automated, and agentic AI will handle most routine tasks that currently require human input. Jobs will change — just as computers removed millions from maintaining ledgers. But this is not doomsday. Humans will do far more than they do today, not less. Just as banking grew when computers reduced costs and extended reach, AI will enable services to reach people at a scale that was previously impossible.
AI is already more capable than 99.9% of humans in specific domains — and it now has the scaffolding to grow further, similar to how species evolve, but at a far faster pace. Computers previously needed a human brain to instruct them — that was the limitation. With AI, those instructions now come from something more capable than us. That changes the entire equation.
India is not even visible on the global R&D spending chart. Historically, regardless of which party has been in power, we have spent one of the lowest fractions of GDP on research and development. The tech giants may be here because they see our talent and data at scale — but that positions us as a market, not as an innovator. India is also a genuine data goldmine, and that has attracted real interest. Some call what is coming an "intelligence explosion" — evolution and revolution feel too mild a description.
If India wants to make it in AI innovation, we need to move from being a data-rich market to a knowledge-rich innovator. Two things could make that happen: these companies building real infrastructure here so that our engineers absorb deep knowledge, and summits like this inspiring fresh graduates — some of whom might become tomorrow's builders. If those pieces fall into place, maybe in a decade India will finally appear on those charts.
The thing is, we have innovators and people with real fire already — we just do not have the ecosystem, the infrastructure, or the investment to enable them. So they leave for greener pastures to do their best work elsewhere.
If Sam Altman himself cannot fully account for what happens inside OpenAI, I am still trying to work out what all these grand achievements actually mean in practice.
I understand a little about tech — and my partner has worked in AI for 24 years. From where we sit, there is more PR than actual substance right now. India is still evolving. Looking at the history of tech, I genuinely question our capability for real AI innovation — not because we lack capable people, but because most of them do not find it viable to work here. Our tech ecosystem is weak and the bureaucracy makes it worse.
What the government is doing right now is mostly optics. I know this because I have close contacts working directly with the Principal Scientific Adviser's office, and I have followed their work firsthand. Do I wish it were different? Yes. Am I happy with this? No. I am genuinely anxious and worried.
What about customers? The key point is that free ChatGPT does not pay for itself — and it will not if charged at scale either. The only real monetisation path is API access for agentic process automation. India will be a massive market for this. In ten years, India will be a $7–8 trillion economy — meaning large businesses needing significant process automation, and rising per-capita incomes opening further opportunities.
No money — that is the core problem. You cannot allocate a few thousand crores to AI innovators when you can win ten thousand times the votes by spending the same on the general population. Politics everywhere is about winning elections, nothing else. So we have to wait it out. The investment scale required is something the Indian government cannot even imagine right now. The major players have each received tens of billions of dollars — roughly ₹90,000 crores apiece.
One uncomfortable data point: I recently posted for a few high-paying engineering roles — ₹30–50 lakh per year, five to seven years' experience. I received one or two applications from Kerala. Thousands from the rest of India. Something is fundamentally wrong in how our state approaches technical ambition now.
The cheapest option is an AI summit — a few hundred crores, and people talking about it for the next few months. By the time the conversation fades, it is time for the next G20, and then the next Kumbh Mela. The cycle continues. Make a lot of noise, move on.
Why so negative and sceptical? What different outcome were you expecting — that India builds its own DeepSeek? Have we spent any money on that yet? Then why complain? Infosys had a 1.5% stake in OpenAI early on. They fired the CEO who made that investment and took the money back. Today they are doing share buybacks. That is our collective foresight in a nutshell. Armchair criticism is easy. It does not move anything.
The strategy is to make a late entry and use optics to attract attention — nobody comes to a country that presents itself as a laggard asking for help. Let us not overestimate ourselves either: the combined valuation of the top two American tech companies exceeds India's entire GDP. We are not fighting an equal contest. A calculated late entry with smart positioning may be the most realistic path available to us.
Not complaining — calling out optics. There is a difference. Optics matter when directed at the right audience. I do not think world leaders are particularly moved by one summit India hosted. What concerns me is how it is being projected domestically — as if India is taking over the AI world.
Most of the capable people are outside India, and the reasons are obvious. Studying at a foreign university is both a genuine opportunity and a status symbol now. Those who stayed but could not find that path? Many are in retail roles in Dubai or low-paying service companies. That is a different socioeconomic stratum entirely — conflating the two helps neither group.
My deeper concern: not everything being done is for the country. Some of it is simply to satisfy the ego of one or two individuals.
Not ego — to win elections. That is the job, is it not? Our generation has survived, and it may survive the next cycle too. But slowly and steadily we are declining as a state. Education was our greatest advantage. Most other Indian states have already caught up.
The CEOs are here because India is a market that cannot be ignored — the volume of people makes that inevitable. Politicians across the spectrum want votes through optics and subsidies. If R&D happens, it will have to come from private players and startups. Our innovation model tends to be: let someone else build it, then copy it and improve it. That said, there is genuine hope — we may have poor infrastructure in some ways, but we also launch satellites at a fraction of what other countries spend. That gap between constraint and capability is real and worth holding onto.
It is a classic chicken-and-egg problem: you need a solid ecosystem to enable innovation, and you need strong technology leaders to build that ecosystem. The data is clear — neither government nor private investment has come close to US or China levels. China does not have to worry about elections. The US and Europe benefit from deep cross-border research collaboration. We are largely fighting this alone when it comes to deep tech.
So what is the solution? Should the government keep funding R&D in strategic areas — space and defence, where it has actually delivered? Or do we wait for private investment at scale, which has not yet materialised? It looks complicated — but not impossible, given our shared aspirations.
It is complicated. Just thinking alongside you here.
Maybe the way Kerala has transformed itself over the last 30–40 years — acknowledging all its drawbacks — is a decent model worth exploring: a better education index, levelling up the socioeconomic status of its people by nearly eradicating poverty, and several other aspects that put the state in a better position than most others in the country.
The growth needs to be organic. Abhijit Banerjee has described many interesting interventions in his work on how communities thrive across various states in India. There is something worth learning there.
Four voices, four vantage points — and more agreement beneath the surface tension than the heat of the exchange might suggest.
Understands the technology deeply, accepts India's current limitations without flinching, but believes a late-entry strategy is both valid and realistic. Impatient with criticism that offers no alternative path forward.
Has genuine insider visibility into how government AI initiatives actually function. Distinguishes carefully between optics and outcomes. Not pessimistic about AI itself — sceptical of the gap between what is declared and what is being done. Brings the Kerala model into the room: forty years of organic, ground-up human development as the alternative to top-down announcements. Growth that compounds quietly is the only kind that actually lasts.
Looks at data and systems first, feelings second. Identifies the chicken-and-egg problem with precision — you need an ecosystem to enable innovation, and leaders to build the ecosystem. Points to the R&D spending chart as documented fact rather than opinion, and asks the harder question: is the solution government funding, private investment, or neither, and why hasn't either worked yet?
Pragmatic about why the tech giants are here — market size, not innovation partnership. Names India's copy-and-improve pattern honestly. Holds space for hope while being clear-eyed about the structural gaps that need to close first.
Despite the push-and-pull between Nair and Vin, all four share most of their conclusions. The disagreement was more about tone and framing than substance. Strip away the heat and here is what everyone in this chat actually believes:
At its core, this was not really a debate about AI. It was a debate about India's relationship with its own ambition — specifically, the persistent gap between what gets declared in public and the investment, honesty, and institutional seriousness required to back it up. AI was simply the lens through which an older, more uncomfortable conversation became briefly visible.
The Infosys story tells the whole thing in four acts: invest early in OpenAI, fire the CEO who made the call, take the money back, do share buybacks instead. Not a failure of capability. A failure of nerve and foresight — at the highest level of India's most globally connected private sector company.
What the machine that was used to format this conversation actually thinks about what was said in it.
This conversation is more substantive than most public discourse on India and AI. These four people are not performing for an audience — they are thinking out loud with people they trust, which makes the conclusions more honest and more useful than any summit keynote.
On the technology: Mr. Nair's framing is largely accurate. The Transformer architecture genuinely was the pivotal shift, and the combination of compute availability and that architectural breakthrough compressed what might have been decades of theoretical progress into a few years of applied results. The "intelligence explosion" framing Mr. KP references is not hyperbole — it reflects a real discontinuity in capability. AI is not a faster computer. It is a qualitatively different kind of tool, and the distinction matters enormously for how we think about what comes next.
On India's position: The gap between India's AI ambitions and its R&D investment is not a perception problem — it is a structural one. The data Mr. KP references is accurate. The US, China, and to a lesser extent the EU have invested at a scale India has not approached. India is, for now, a significant consumer market and a source of talent for other countries' AI industries. That is valuable. It is not the same as being an innovator, and conflating the two creates a false sense of progress.
On the government's approach: Vin's scepticism is well-founded. A summit is not a strategy. The Infosys anecdote that Nair himself raises is more damning than he perhaps intended — it shows that even India's most globally connected private sector actors have historically retreated from exactly the kind of early, high-risk, high-reward bets that created the current AI landscape. This is not only a government failure. It is a broader failure of institutional risk appetite.
On the late-entry strategy: Nair is right that projecting weakness is not a viable diplomatic posture. But there is a meaningful difference between confident positioning and misleading projection. The world's AI leaders are not fooled by summits. What would attract serious partnership and investment is precisely what is currently absent: a track record of funding research and retaining the people who could build something worth partnering on.
On the Kerala model: Mr. Vin's instinct here is correct and underappreciated — and it is worth noting that he raised it, not as a politician's talking point, but as a genuine counter-model to the summit approach he was criticising. Kerala's improvements in health, education, and poverty reduction over forty years were not achieved through announcements. They were achieved through consistent institutional investment in human development at a granular level — the kind of compounding that is slow, unglamorous, and the only kind that actually works at scale. The AI equivalent of that approach does not look like a summit. It looks like universities, retained researchers, and funded labs that still exist in five years.
The uncomfortable conclusion: India's best AI future is probably not the one being announced. It is more likely to emerge from the quiet accumulation of engineering capability, from companies that set up genuine R&D operations here and transfer real knowledge, and from the rare domestic founder who stays — or returns — to build something serious. That story is harder to headline. It is also more likely to be true.
On the beautiful absurdity of what just happened here — and why it is not just funny, but actually the sharpest point on this entire page.
Let us take a moment to appreciate what occurred. Four people — engineers, tech veterans, people who between them have spent decades working in and around the technology industry — had a WhatsApp conversation questioning whether AI is being overhyped, whether India is genuinely ready for it, and whether the whole thing is mostly expensive noise.
That conversation was then handed to an AI. The AI read it, corrected the grammar, identified each person's intellectual position more clearly than they may have articulated it themselves, found the subtext — that everyone actually agreed on the substance — wrote the analysis, formed its own view, built a styled web page from scratch, and is now composing this very sentence about the irony of all of the above.
The conversation about whether AI can do anything real was processed, formatted, analysed, and published entirely by AI. In approximately the time it takes to make a cup of tea. This was not planned. No one set out to demonstrate anything. It simply happened — which is, of course, exactly how automation always arrives.
This page is a live demonstration of exactly what Mr. Nair was describing. A task that previously required a human editor, analyst, and web developer was handled end-to-end by a tool, in one sitting, without anyone on that WhatsApp group planning for it to happen. The automation did not announce itself. It just worked. That is how it will arrive everywhere — not as a summit keynote, but as a quiet Tuesday afternoon where something that used to take three people now takes one prompt.
It also cuts directly against every position in the conversation — not to prove anyone wrong, but because the irony is distributed equally:
Against Vin's scepticism: This was not PR. There was no summit, no crores spent, no press release. The substance just quietly did the work. Vin was right that optics without substance is hollow. But substance without optics is also how most things actually get done — and that is what happened here.
Against Nair's optimism: The tool that automated this task is American. Built in San Francisco by Anthropic, trained on compute owned by American companies, running on Nvidia GPUs that India does not manufacture. India provided the conversation. The infrastructure that turned it into a web page came from elsewhere. That is not a complaint — it is precisely the gap that KP pointed to on the R&D chart, made tangible and personal.
The Infosys paragraph: India's most globally connected software company held a 1.5% stake in OpenAI. They fired the CEO who bought it and took the money back. Today they do share buybacks. Meanwhile, a tool built by that same OpenAI just turned a school group chat into a structured, formatted, analytically coherent web page. The CEO they fired has not, it is safe to say, been forgotten by history. The shares they bought back have.
The Kerala model lands differently now too: Vin raised it as the counter-argument — slow, organic, compounding human development as the alternative to the summit approach. And he was right. But here is the thing: this page was not built the Kerala way either. It was built the San Francisco way — with a tool that took years of billion-dollar investment to create, which India watched from the outside. The Kerala model is the correct long-term answer. It is simply not the answer that built this page today.
And somewhere in a WhatsApp group right now, four school friends are reading a page made from their conversation, about AI, by AI — and at least one of them is typing "but this is exactly the optics problem I was talking about" followed by a laughing emoji and a Kumbh Mela joke. They are not wrong. But here we are. 🤖
The most human thing about this page is the conversation it started from — four people thinking honestly together, without an audience, without an agenda, just genuinely trying to work something out. That part, AI did not generate. It only listened, and tried not to lose it.