Find the bottleneck. Own the toll.

This is the research framework we use to find the durable layer in a technology buildout — the part every buyer has to pass through no matter which product, model, or company ends up winning. It was built on the AI datacenter buildout and then carried into AI-assisted drug discovery. It is free, and it is a method rather than a portfolio.

↓ Download CLAUDE.md Read it in the browser

No positions, weights, or price targets in it. Deliberately.

How to use it

Save it into the root of a project folder as CLAUDE.md. Claude Code, or any agent that reads that convention, picks it up automatically. Then point it at filings, transcripts, earnings decks, or your own notes.

~/my-research/
  ├── CLAUDE.md        ← this file
  ├── filings/
  └── transcripts/

Then ask it things like: “Map the supply chain for this buildout. Score each layer on replicability and recurrence, and tell me which one everyone has to pass through.”

The five principles

The bottleneck always migrates. Markets fixate on the glamorous layer and misprice the constraint. Ask where it is now, and where it moves next when this one is relieved.
Own the toll, not the coin flip. Prefer businesses paid on the attempt over businesses paid on the outcome.
Separate capital expenditure from consumables. Two businesses in the same “tools” bucket can run on completely different clocks.
What diffuses cannot be a moat. If it can be replicated by reading a paper and renting compute, it will not hold pricing power.
Scarcity is the input nobody can synthesize. Find the input the model cannot generate for itself.

Also free

The AI Biolab primer — the full-stack primer on AI-assisted drug discovery, layer by layer.
The interactive supply-chain map — every layer and company, public and private.
The AI-Science Basket tracker — the thesis marked to market against the S&P 500 and biotech.