Ep #3: How does Peak XV Choose Their Bets | Funding Advice, Hiring, Early Career

ELI5/TLDR

Pushpak, a 29-year-old investor at Peak XV (the firm formerly known as Sequoia India), explains how a seed VC actually picks companies: meet 300 founders a year, say yes to three. About a third of those bets are made before a single line of code exists, so the whole game is reading the team and the market. His through-line is that pedigree is just a cheap proxy for excellence, and if you can prove excellence some other way, the proxy stops mattering. The rest of the conversation is career advice for people in their early twenties and a calm, bullish take on AI.

The Full Story

How the job actually works

The numbers do most of the explaining. Each investor meets roughly 300 teams a year and ends up backing three or four. That’s a 1% yes rate, and 99% of the work is building a case for why to say no.

“We meet 300 companies, but we just need to say yes three times. And so 297 of those are no’s.”

The job splits into four buckets: sourcing (cold-meeting founders, including ones who wrongly assume they’re too early), evaluation (a one-to-two-week sprint talking to the team, customers, and competitors), portfolio support (helping with product, hiring, fundraising after the check), and thesis-building (sitting with a category until you have an opinion about where it’s going). Evaluation is the intense part. Sourcing is constant. The other two are where time goes once you stop being only a deal machine.

What they’re buying when there’s nothing to buy

If a third of investments happen before there’s a product, what is there to evaluate? Pushpak’s answer is deliberately unglamorous: team and market. Market means “we don’t know if this works, but if it does, it’s enormous.” Team is where it gets specific, and his honest position is that there’s no formula. They back teenagers and veterans, solo founders and groups of four. Forcing a checklist onto that, he argues, is how you miss things.

The pedigree question gets the sharpest answer in the episode.

“Pedigree is a proxy for excellence. And excellence matters. But if you can demonstrate excellence in any other capacity, that is totally fair game.”

His example: a pre-seed check into a second-year dropout from a tier-three engineering college who’d spent two and a half years buried in robotics research nobody assigned him. No credentials, but a clear, earned view of what robotics needs for its “ChatGPT moment.” That self-directed obsession is the tell. He calls it sweat equity, the founder version of putting your own money in. You didn’t risk capital, you risked years of thinking, and that alone signals you might win.

Two traits draw him in. The first is depth, founders who are “historians” of their category, who know every move that’s been tried and why it failed, who can zoom from the abstract (why aerospace compliance demands a different kind of software company) down to the granular (what a geometric kernel is and why nobody’s improved it in 20 years). The second is the pairing of earnest and ambitious, someone who’ll admit they don’t come from the space or have the money, then claim they’ll change the world anyway. He calls that combination seductive.

How to get a check (and how to scare one off)

Asked for a literal roadmap, his step one is “put in the work” on why you’re building what you’re building, which he says is shockingly rare. Most people, in his telling, start up as an act of defiance, the quit-the-job-and-build-a-startup aesthetic, rather than out of genuine conviction. Step two is quiet confidence that you’re building this regardless of whether Peak shows up.

The red flag is the inverse:

“Total red flag if you are coming to us to seek validation and to say that if you fund me then I will start up.”

On who should raise at all, he’s refreshingly against his own product. Every raise gives away a slice of the company and adds stakeholders with their own interests, so if your business doesn’t need the capital, don’t take it. But for ideas that genuinely require a team, an R&D cycle, a go-to-market push, venture is, in his words, a beautiful instrument, because failure is priced in. The company dies, you shake hands, nobody comes after you. The opposite of a bank loan.

The AI section, which is most of it

Pushpak is not an AI bubble believer. His argument has two legs. First, even if valuations are frothy and money is flowing absurdly, the technology is genuinely unprecedented, “geniuses in a data center,” as he quotes Dario Amodei. Second, and the line that does the work:

“Today is the worst that it’ll ever be.”

Last year’s hallucination complaints have already faded; the curve is exponential and accelerating, so even if it’s a bubble, it doesn’t change what you should do.

On jobs, he takes an abundance view that cuts against the usual fear. Are portfolio companies hiring fewer engineers? The opposite, he says, because once you can build more, you want to build more. Hiring is the number-one board-meeting topic, and the complaint is always “we can’t ramp fast enough.” The reframe he and the host land on is brain-and-hands: AI can be both the hands (execution) and the brain (thinking), so the people who win are the ones who step one level above and orchestrate agents, setting evals, guardrails, creative direction, mapping it all to business context. He sees the same shift inside companies, like GTM engineering, technical people building systems that let three salespeople do the work of ten.

Career advice and the boss

For freshers, he separates “wealth creation” from “setting yourself up for wealth creation,” and says the second is the right question early. Three axes: work under genuinely excellent people (once you’ve tasted excellence, mediocrity becomes physically unbearable), pick roles with diversity rather than a narrow execution lane, and join something that’s actually winning, because watching a company go from 25 to 500 people rewires how you understand risk and success.

On his own boss, Rajan, three things: contagious energy (“bring your best self to work every day”), being a genuine champion of his people, and direct communication with no politicking. The empowerment one lands hardest:

“I trust you. I back you. Go out and make mistakes and we’ll figure it all together. That is extremely rare.”

He frames the stress of the job through the long feedback loop, you find out if you were right in five or ten years, and a lot of it is luck you don’t control. His coping mechanism is athletic: obsess over the “game tape.” The next match is far off, but you can still rewatch every founder meeting and ask what you misread.

Key Takeaways

Seed VC funnel: ~300 founders met per investor per year, 3-4 checks written. A 1% hit rate, and roughly a third of bets are pre-product.
At seed, the only two things being evaluated are team and market. “Large market” means small odds of working but enormous if it does.
Pedigree is treated as a proxy for excellence, nothing more. Prove excellence another way and credentials stop mattering.
The strongest founder signal is self-directed, unassigned deep work in a category, “sweat equity,” because no one made you do it.
Two favored founder traits: encyclopedic depth in a category, and the earnest-plus-ambitious pairing.
Biggest red flag: a founder who’ll only build if funded. Conviction has to precede capital.
Don’t raise if you don’t need to, every raise dilutes ownership and adds competing interests. Raise only when the idea structurally requires it.
VC failure is priced in by design (no recourse), which is what distinguishes it from debt.
The four parts of the job: sourcing, evaluation (the intense one), portfolio support, thesis-building.
Analyst hiring screens for three things: raw smarts (tested via deliberately topic-hopping interviews), genuine enthusiasm (vibes, not box-ticking), and “smell of money” (first-principles feel for what makes a business, not an idea).
New interview tactic: have candidates screen-share and research live with Claude or ChatGPT, to watch how they frame follow-up questions. Curiosity is the signal.
AI stance: not a bubble; “today is the worst it’ll ever be.” Engineer demand is rising, not falling, under an abundance mindset.
The durable skill is orchestrating agents, abstracting one level above being the brain or the hands.
Career advice for freshers: optimize for excellent people, role diversity, and joining something that’s winning.

Claude’s Take

This is a clean, articulate primer on early-stage VC from someone clearly good at his job, and the host keeps it moving. The genuinely useful, non-obvious bits are the inversions: don’t raise money if you don’t have to, conviction must come before the check, and pedigree is just a stand-in you can route around. Those are worth keeping.

The discount is that almost everything here is the standard, optimistic VC catechism, delivered fluently. “Today is the worst it’ll ever be” is a real Sam Altman / accelerationist talking point that conveniently makes every current limitation disappear, and the abundance-means-more-hiring claim is the position you’d expect from someone whose portfolio’s value depends on it being true. None of it is wrong, exactly. It’s just frictionless, and the long feedback loop he describes so well means none of these confident takes can be scored for years. The “game tape” framing is the most honest moment, an admission that a lot of this is unfalsifiable in the moment and partly luck. A 7: high signal-per-minute for anyone curious how seed checks get written, lightly cut with house optimism.