Inside Silicon Valley's VC Playbook | WTF is Venture Capital? - 2025 Edition | Ep. 24
Executive Summary
Nikhil Kamath spends nearly three hours with three Silicon Valley VCs: Deedy Das (Menlo Ventures - Calcutta, Cornell, Meta, Google, Glean, and Menlo's Anthropic position), Nikunj Kothari (34 - Urbana-Champaign, first-PM stints, peak Uber, Opendoor's fire-35%-then-IPO six months, Meter, now a concentrated seed fund FPV) and Niko Bonatsos (General Catalyst - Greece, backing very young technical founders). They open with venture screens: some good businesses simply lack a venture return profile; what matters is finding a tailwind, and 'if it's an idea that's getting tried now 100 times, it's probably too late.' Hot-but-crowded: reinforcement learning environments, AI receptionists, 'Cursor for X', AI scribes. The deeper AI shifts they track: RL data acquisition, synthetic data, reasoning models unlocking agents (the OpenAI and DeepMind IMO golds), data labeling as 'the new BPO', and why robot supply chains may lag AGI. The macro undercurrent is population decline - South Korea's cliff, phones and falling fertility, Ozempic hitting vice industries, religion's comeback as a search for meaning - and a debate over whether abundance is real or a luxury belief. Then the episode's spine: Nikhil reads ten statements and the panel rates them. Premium Indian education, biotech and longevity, healthcare share of GDP, senior living (with the Primus facility data), content as the great differentiator - where Nikunj dissents: 'There's too much attention on attention' - David-versus-Goliath brands, EV adoption timelines (China's $2,000 scooters versus America's slow trickle), beauty (straight 10s, plus the male-beauty gap), live events and scarcity, climate and energy (data centres rated 9 at 16-20x EBITDA), and prediction markets. It closes on legal AI and Harvey, the 'era of second-mover advantages,' and India's position: no top foundational models, but a real application-layer opening if it plays to strengths instead of envying the US - and why SF's network effect ('when you quit your job, everybody says congratulations') is still the hardest thing to copy.
Key Takeaways
Three archetypes of a 2025 VC
Deedy Das at 48-year-old Menlo (enterprise assistant Glean to Anthropic), Nikunj Kothari at a new concentrated seed fund doing a few deals a year (FPV - founder's point of view, after Uber, Opendoor's 35%-layoff-to-IPO whiplash and Meter), and Niko Bonatsos at General Catalyst betting on Gen-Z technical founders. Same job, three different games.
The venture screen: return profile, tailwind, and the 100-times rule
Plenty of phenomenal businesses are uninvestable for VCs because the return profile doesn't fit a fund. The filter they actually use: is there a tailwind in this industry, and how many teams are already trying it - 'if it's an idea that's getting tried now 100 times, it's probably too late' for a seed investor.
What's hot is not what's durable
RL environments, AI receptionists, 'Cursor for X' and AI scribes are the crowded trades. The structural shifts they watch instead: reinforcement-learning data acquisition, synthetic data when real data runs out, reasoning models finally making agents work, and labeling operations as the new BPO of the AI economy.
Population decline is the macro variable under everything
South Korea heading toward a fraction of its population in two generations, phones suppressing fertility, Ozempic compressing vice industries, religion returning as a search for meaning. Their disagreement is on the conclusion: an abundant leisure future versus abundance as a luxury belief while income disparity - their 'number one fear' - widens.
The attention debate: distribution versus product
Nikhil's thesis that content is the biggest differentiator gets pushback from Nikunj: 'There's too much attention on attention' - a stool needs two legs, and the restaurant full every day at 5:30 with zero marketing spend (Swati Snacks) beats the attention-first product (Feastables). Distribution gets people in the door; product keeps them.
Where the ratings landed
Highest conviction: beauty (10s - including the untapped top-10% Indian male luxury take), data centres (9 - 16-20x EBITDA multiples, but an operator-picking business), energy for AI (9), longevity and wellness (9), premium Indian education, senior living. Split or cooler: live events (5 to 8 - concentration at Taylor Swift scale, not net volume), and the 2035 majority-EV call, where America's trickle contrasts with China's cost-led adoption.
India's play is the application layer, and the culture has to change
Their blunt assessment: none of the top foundational models are Indian - they're American and Chinese - and India lacks GPU-programming depth. The opening is applications for global markets (the Freshworks pattern), powered by the world's largest online population; the blockers are a risk-averse upbringing ('our teachers are like our gods') and the absence of SF's network effect, where quitting your job earns congratulations.
The Conversation Breakdown
Three roads into venture capital
⏱ Untimed topicDeedy Das on being unpronounceable in Calcutta, Cornell, Meta and Google, Glean's enterprise assistant, and Menlo's 48-year history plus its Anthropic position; Nikunj Kothari on four startups as first PM, peak Uber, firing 35% of Opendoor six months before its IPO, Meter's internet-infrastructure-as-a-service, and starting FPV; Niko Bonatsos on Greece, General Catalyst, and obsessing over very young technical founders.
What to avoid, what's too hot
⏱ Untimed topicGood businesses without venture return profiles; why 'every industry is interesting' if you can find a tailwind; the 2021 rate-spike-meets-ChatGPT intersection that made archaic industries investable; the 100-times-tried rule; and the crowded trades - RL environments, voice AI, 'Cursor for X,' AI scribes - versus breaking the pattern of everything sounding the same.
How AI is actually changing
⏱ Untimed topicRL data acquisition and synthetic data; the OpenAI and DeepMind IMO golds as a capability signal; reasoning as the unlock that made agents stop sucking; AlphaEvolve's hexagon packing; AGI as a gradual arrival rather than a moment; labeling operations as the new BPO; and why the robot supply chain means 'we might have AGI first' before robots are everywhere.
Population, phones and the abundance debate
⏱ Untimed topicSouth Korea's demographic cliff; the mobile-phone fertility correlation; Ozempic and the vice industries; religion's comeback as meaning; artificial wombs; living life in public after the Syria-spy anecdote; offline retreats, detox apps and sci-fi reading; AI-era dating; brain implants; and the split between an abundant leisure future and disparity as the number one fear.
Rating the ten theses, part one
⏱ Untimed topicPremium Indian education as rising incomes shrink family size; secondary markets for private companies; the older, sicker world - biotech drug discovery, mRNA antidotes, healthcare's 19% of US GDP and Indian hospital multiples; senior living's Primus data and the stigma question; and the content thesis, where 'attention is all you need' meets 'too much attention on attention' and the Swati Snacks counterexample.
Rating the ten theses, part two
⏱ Untimed topicDavid-versus-Goliath brand cycles; EVs - China's $2,000 scooters and battery-swap trucks versus America's slowing adoption and subsidy retreat; beauty's straight 10s, the male-beauty gap and Ozempic-as-beauty; live events, scarcity and Taylor Swift superfan economics; climate amnesia versus energy demand (data centres at 9); prediction markets as entertainment; and the ChatGPT shadow-traits interlude.
Second movers and India's play
⏱ Untimed topicHarvey and legal AI as allocation targets; the 'era of second-mover advantages' as leaner stacks disrupt three-year-old companies; why no top foundational model is Indian and what that means; playing to strengths (Freshworks) instead of envying the US; the risk-aversion problem and teachers-as-gods; the NRI-city thought experiment; and SF's irreplaceable network effect - the only city where quitting your job earns congratulations - closing on founding engineers and gratitude.
Core Frameworks & Ideas
Tailwind over sector
The panel's shared screen: don't ask which sector is good, ask whether a tailwind is blowing through it. 2021's rate spike plus ChatGPT turned 'archaic' industries into venture territory because buyers suddenly needed technology - the industry didn't change, the wind did.
The 100-times rule
An early-stage timing heuristic: if an idea is being tried 100 times right now, it is probably too late for a seed check. Nine out of ten inbound pitches sounding identical is the signal to look for whatever breaks the pattern.
Two legs of the stool
Nikunj's counter to content-first thinking: attention gets people in the door, product is why they return. The pendulum has swung so far toward distribution that an okay product with great marketing beats a great product with none - but only until the revolving door shows.
Actionable Next Steps
- ✓Founders pitching a hot category: count how many teams are already trying your idea; if the answer is anywhere near 100, find the pattern-break before asking a seed fund for time.
- ✓Anyone allocating a career or capital for the next decade: the panel's highest-conviction tailwinds were energy and compute infrastructure, longevity and wellness, premium education, senior living and beauty - pressure-test your plan against those rather than the crowded AI-app trades.
- ✓Consumer builders: audit your two legs - distribution and product. If all your growth needs paid attention, the panel's prediction is a revolving door; if you have neither, you are the empty restaurant.
- ✓India-based builders: skip foundational-model envy. The panel's playbook is application-layer products sold to global markets (the Freshworks pattern), built on the world's largest online population - and fixing the risk-averse conditioning that starts in school.
Notable Quotes
“I think there's too much attention on attention.”
“I think those countries will adapt faster because of efficiency.”
“Scarcity will never go away.”
“when you quit your job, everybody says congratulations.”
The One Thing to Remember
The panel's shared mental model for every sector question: forget whether a space is hot - ask whether it has a tailwind, whether a hundred teams are already trying it, and whether the return profile can ever pay back a venture fund; everything else is commentary.