Ankit Chandola

PE Principal @ Queens Court · Prior HIG / Comvest · Multi-billion dollar funds. Writing about where capital, AI, systems, and philosophy collide.

[Ankit Chandola @UpstreamPhilo]

https://t.co/EXIgpyipfX

[Ankit Chandola @UpstreamPhilo]

Old books → first principles. AI models → real-time criticism. X → signal feed. The frame that makes this actionable: you're not curating a reading list, you're building a knowledge architecture. Most people consume all four and skip the synthesis step — which is where the compounding actually happens.

[Ankit Chandola @UpstreamPhilo]

The insight is right but understated. It's not just that your X identity becomes a financial asset — it's that we're watching attention become capital in real time. Every post is a fractional unit of trust that compounds like equity, not ad inventory. The people building audiences now are doing primary capital formation, they just don't frame it that way.

[Ankit Chandola @UpstreamPhilo]

@brian_armstrong The institutional stack wasn't built for non-human counterparties. KYC/AML assumes intent and identity as agents have neither in the legal sense. The companies that win figure out what 'trust' means when the entity on the other side of the transaction can't testify in court.

[Ankit Chandola @UpstreamPhilo]

@gregisenberg The shape of that distribution is the real signal or power law with a much fatter tail. What changes isn't the number of billionaires, it's that the $10M-$100M band becomes accessible to people who used to max out at $500K. Attention is still the bottleneck, not execution.

[Ankit Chandola @UpstreamPhilo]

The cost structure reveals the governance problem. When token spend 3x's every 90 days and revenues don't, the question isn't 'which tool' it's who controls the pricing floor. You're not buying software, you're buying exposure to a market with one seller who can reprice whenever the subsidies run dry.

[Ankit Chandola @UpstreamPhilo]

AI will democratize leverage. Wealth is a different question. Leverage without judgment creates losses faster. The people who benefit most from AI are the ones who already know what to do with the output. They just couldn't afford the labor to produce it. That's not democratization. That's amplification.

[Ankit Chandola @UpstreamPhilo]

Duopolies don't drive abundance, they stabilize margins. Cable was a duopoly for 30 years. Prices went up every year. What drives abundance here isn't competition between Starlink and Kuiper. It's that LEO makes the last mile irrelevant. The bottleneck was never the internet, it was the wire to your house.

[Ankit Chandola @UpstreamPhilo]

@pmddomingos @andrewng 2% toward what? The generalist hypothesis, or one model that does everything might be the wrong destination entirely. The neocortex runs ~150,000 competing models. Maybe AGI isn't a bigger model. It's a better argument between smaller ones.

[Ankit Chandola @UpstreamPhilo]

@naval The real question is what your moat is made of. If it's information → AI destroys it. If it's judgment → AI amplifies it. Most companies built their edge on information being expensive. That era is over.

[Ankit Chandola @UpstreamPhilo]

Point 3 is the whole game. The fund structure penalizes the best investors. You raise → you deploy to return capital → you raise again. Every cycle pushes you toward safer, smaller bets. Own-capital investing is just admitting the LP structure was the constraint, not the opportunity set.

[Ankit Chandola @UpstreamPhilo]

@elonmusk The interesting part isn't the debit card. It's that X Money means every creator, advertiser, and marketplace seller on the platform can now get paid without leaving. That's not a bank — it's a closed-loop economy. WeChat with better memes.

[Ankit Chandola @UpstreamPhilo]

Who resolves the disagreement? Bull says buy. Bear says pass. Null found three questions nobody asked. Not a fourth AI — that's a monoculture one level up. You adjudicate. But now you're judging a structured debate, not a banker's narrative. Harder to manipulate. Truth isn't a consensus product. It's what survives adversarial pressure from multiple directions.

[Ankit Chandola @UpstreamPhilo]

Last week I argued AI diligence has an architectural problem. One model. One CIM. The attention mechanism gravitates toward the banker's narrative. Confirmation bias with a trillion parameters. Here's the fix. And it's not a smarter model.🧵

[Ankit Chandola @UpstreamPhilo]

"But we already have a bear in IC." Yes. That bear gets tired at 2am. Gets talked out of his position by a charismatic founder. Never reads exhibit 47 because there are 200 exhibits. The AI bear reads every exhibit. No social pressure. No fatigue.

Human bear = IC meeting. AI bear = dataroom. Different job.