How to 10x Your Value in the A.I. Era | ft. Kunal Shah

ELI5 / TLDR

The internet made information free; AI makes intelligence free. The catch: the same phone that can make you smarter is mostly being used to waste time, and the income gap between people who use these tools well and people who don’t is about to get enormous. Kunal Shah’s argument is that nobody — not the government, not your company, not your parents — is coming to fix this for you. You have to treat yourself like a stock with a growth rate and raise it yourself, by learning fast, picking better people to be around, and being willing to sacrifice.

The Full Story

Intelligence is now in the air

The framing Shah keeps returning to: the internet “made information accessible in the air” — search anything, get anything. AI does the same thing for intelligence, the one trait we’d always reserved for ourselves as a species.

“Now you can just plug it up in the air and just suck it in and you can actually accelerate materially.”

He’s not in the AI-apocalypse camp, but he’s not relaxed either. His worry isn’t mass unemployment — the world has “interesting ways to keep people employed” (there are still ATM guards and liftmen). The worry is per-capita income. The gap between people who use AI and people who don’t will widen, and that delta is the real story.

The largest employer in the world is inefficiency

A nice line, delivered flatly:

“We forget that the largest employer of the world is actually inefficiency.”

AI eats inefficiency. As it disappears, jobs built on it get displaced or rewritten. The displacement lands hardest at the bottom rung: tasks you’d once hand to a junior — a research note, basic code, the kind of work that got outsourced to India precisely because it was delegable — are now the easiest things to automate. Which breaks the career ladder. If a 22-year-old can’t get the entry job where they’d normally hone judgment over five years, how do they ever reach the middle?

The same device makes you smart or dumb, fast

India uses ~35GB of mobile data per person per month — more than any country on earth. Most of it goes to short videos and time-pass. Shah’s point is that time is the real wealth, and the phone is a fork in the road:

“We have the same device that can make us super intelligent or super dumb very very fast — within weeks you can actually see yourself go in opposite directions.”

He throws in a linguistic observation that lands oddly hard: there is no native word for “efficiency” or “productivity” in Indian languages. The vocabulary arrived with the industrial revolution, when time became money and Western work went hourly. Indian brains, he argues, were never wired in hourly mode, so we undervalue time.

Apathy, then panic

India’s pattern with anything big, in his telling, is the COVID pattern: apathy until it’s in your face, then sudden panic, nothing in between. We’re in the apathy phase with AI because the damage feels like it’s happening “in some other market.”

His worked example of why this is dangerous: banks. Financial services are 30–40% of India’s market cap. Bank stability comes from lending; lending leans on IT and back-office (BPO) operations, which might be 30–40% of a bank’s cost base. Automate even 10–20% of that and you start hitting the safest part of the book — and nobody, he says, is pricing in that second-order chain. His broader complaint is that we don’t think past the first order; curiosity ends at a tweet.

A protected economy delays the reckoning

Why might India feel the shock later than the US? Because a large chunk of India’s profit pools are protected — by license, regulation, or government ownership. AI matters at the “global maxima” (anything that competes internationally: manufacturing, IT outsourcing) and matters less at the “local maxima” shielded by regulation. So the impact plays out slowly here — but the moment one player breaks ranks, it cascades.

That’s the game-theory bit. If one bank doubles its revenue-per-employee using AI, it dislodges everyone’s market share and forces every competitor to follow — the same thing that happened with computerisation and internet banking, only far faster this time.

The 10x species inside startups

Where he sees it already: tech startups. At his company, CRED, the share of code written by AI went from roughly 5% to 90% in a year. And a split is opening — 10% of employees are becoming “a completely different species” from the other 90%, so productive that everyone and every process around them now feels slow.

“When I speak to these 10x people right now and ask them what can I do to make your life better, they’re saying: can you remove my meetings? Can I stop having people report to me?”

The trap for big companies is what he calls the India traffic problem: a few cars capable of 1,000 km/h stuck behind everyone doing 10. Put high-output people inside low-output processes and you see no gain — which is why incumbents are at a disadvantage to firms built entirely around the fast people.

Nobody is coming

The repeated, slightly uncomfortable theme: high agency. Stop waiting for a government scheme or a company workshop. Everything is “up there”; if you’re genuinely keen, you’ll teach yourself. The shortage isn’t access — India has the cheapest data and the highest smartphone penetration — it’s self-motivation and discipline. We’ve gotten “addicted to somebody else solving this for us.”

He pushes a harder edge too: progress needs pain. Countries evolve through setbacks (his example: post-war Japan). Successful parents struggle to raise successful kids because they remove the exact thing — struggle — that made them. Some protection is non-negotiable (60–70% of the population needs support), but the educated, privileged 20–30% with access to these tools should be taking far more risk and building “substance,” not chasing shortcuts. And substance, he insists, requires sacrifice.

Treat yourself like a stock

His closing advice to a hypothetical 24-year-old, mostly in investor language for a Groww crowd:

Become extraordinary at learning anything fast. If a skill takes you three years, you won’t make it; assume you can learn most specific things in days or weeks.
Audit who you spend time with. You’re the average of five people — think of them as stocks with growth rates. If their XIRR is low, yours will be too. India’s social glue makes it hard to “move on” from people who aren’t growing, and that’s a quiet ceiling on you.
Treat yourself as a stock with a growth rate that only you can raise. Or as an app: fix the bugs, ship features, release often, keep evolving.
Learn technology from people younger than you; learn values from people older than you. Don’t mix up the two.

He ends with the crocodile-shark-crab analogy: the species that survived 100M+ years unchanged share three traits — they can throttle their metabolism at will (survive on little), they have a very high strike rate (rarely attack, but convert when they do), and extreme adaptability. The apex-predator playbook for an AI era: be able to live lean, be selective and lethal when you pick a shot, and adapt relentlessly. And — said at his own event — “don’t waste time at events.”

Key Takeaways

Internet made information free; AI makes intelligence free — the first time a uniquely human trait became commoditised “in the air.”
The real risk from AI in India isn’t unemployment but stalled per-capita income growth; the world keeps people employed, but the income gap between AI-users and non-users widens.
“The largest employer in the world is inefficiency” — AI removes inefficiency, so jobs built on it disappear or get rewritten.
Outsourced/delegable work is most exposed precisely because it was delegable; entry-level roles are hit first, breaking the ladder juniors use to build judgment.
India uses ~35GB mobile data per person per month, the highest in the world — mostly on short video; the same device can make you smart or dumb within weeks.
No native word for “efficiency” or “productivity” exists in Indian languages; the concepts arrived with the industrial revolution and hourly Western work.
India’s response pattern to big shifts is “apathy then panic,” with nothing in between (the COVID pattern).
Financial services = 30–40% of India’s market cap; bank stability rests on lending, which rests on IT/back-office (also ~30–40% of cost) — a prime AI target whose second-order risk to the index is unpriced.
Regulation/licenses/PSU status protect a large share of India’s profit pools, so AI’s domestic impact lags — but it bites hard in globally competitive sectors (manufacturing, IT outsourcing).
Game theory: once one player doubles revenue-per-employee via AI, it dislodges rivals’ market share and forces the whole sector to follow — like computerisation and internet banking, but faster.
At CRED, AI-written code went from ~5% to ~90% in a year; 10% of startup employees are becoming a “different species,” finding everyone around them slow.
The “India traffic problem”: high-output people stuck in low-output processes show no gain — incumbents are disadvantaged versus firms built entirely around fast people.
Female labour participation is a structural drag — Shah’s rough guesses: women are <10% of loan-takers, <20% of credit-card holders; no country has gotten rich on one gender working (absent oil).
“You are the average of the five people you spend time with” — treat them as stocks with growth rates; refusing to move on from non-growing relationships caps your own growth.
Treat yourself as a stock (raise your own growth rate; nobody else will) or an app (fix bugs, ship features, release often).
Learn technology from younger people; learn values from older people.
Survival traits of 100M-year species (crocodile, shark, crab): adjustable metabolism (live lean), high strike rate (selective + lethal), extreme adaptability — the apex-predator template for the AI era.

Claude’s Take

This is Kunal Shah doing what Kunal Shah does: dense, quotable, slightly provocative, optimised for the clip. Most of it is genuinely useful framing — “the largest employer is inefficiency,” the bank second-order chain, the India-traffic metaphor for why incumbents stall, and “treat yourself as a stock” are all sticky and mostly correct. The 5%-to-90% code stat at CRED is the single most concrete data point and the most interesting, even allowing for the fact that “code written by AI” is a slippery measure (autocomplete counts).

Where to keep a hand on your wallet: the high-agency gospel does a lot of heavy lifting and quietly waves away structure. “We have the cheapest data, so there are no excuses” is a debater’s move — access to a smartphone is not access to the meta-skill of knowing what to learn or the slack to learn it, and he half-concedes this when the interviewer pushes. The “progress requires pain, stop protecting people” thread is the kind of thing that sounds bracing from a billionaire founder and lands differently for someone one layoff from trouble; he hedges it (60–70% need support) but the edge is doing rhetorical work. The crocodile-shark-crab bit is fun but is pop-biology dressed as strategy — survivorship-flavoured, literally.

The female-participation point is the most important structural claim in the talk and gets about ninety seconds; the numbers are his own admitted guesses. Treat them as directionally real, not citable.

Net: a 7. High signal-per-minute, a few genuinely portable mental models, no real depth or pushback — it’s a keynote, not an argument. Worth the 34 minutes for the framings; don’t mistake the aphorisms for analysis.