Investing & Models

Investing in AGI: Key Ideas

Premise

Intelligence too cheap
to meter.

The obvious plays — compute, models — get crowded. The real question: where does scarcity migrate when thinking becomes free?

The Premise

Scarcity moves from thinking to permissioned action.

When analysis, code, and synthesis cost only inference, value pools in whatever stays scarce.

  • Cognitive labor becomes a metered utility.
  • Scarcity moves to energy, atoms, distribution, verification.
  • Ownership shifts to the checkers — whoever certifies agentic output.
Drop inference cost. Watch scarcity relocate.
Cognition $10.00 / hr Bottlenecks Standard SCARCITY FLOW

Cognition deflates; validation and energy take the pricing power.

The Predictions

Moats migrate from generation to verification.

Seat-based SaaS decays. The winner knows when agent work is safe to ship.

Pick a sector. See where value goes.
Seat-tax SaaS Margins Decay ↓ Workflow Owners Data + Embed ↑ Generation (Cheap) Verification (Moat) Pure Model wrapper Commoditizes Distribution Hub Durable Early Diagnostics Targeted Prevention

Different paths, one destination: verification.

Macro Path

Software deflation colliding with physical capex inflation.

Not a clean disinflation story. Labor shocks pull short rates down; grid, land, and energy push long rates up.

Toggle the phase. Watch the pressure curves shift.
Now Future Pressure Short Rates Long Rates Power & Atoms

Capex steepens the curve; labor shock flattens it.

App Winners

The durable company owns the customer and the verification loop.

Wrappers collapse to zero margin. Compounders fuse owned distribution with the checking layer.

  • Cool Demo: High model leverage, zero distribution. Competed away.
  • Legacy Channel: High distribution, zero model. Decays unless upgraded.
  • Compounder: Owns the customer, the model, and the checking layer.
Hover a quadrant. See how it wins or decays.
distribution owned → model leverage ↑ Cool Demo Compounder Outsourced Legacy

Own both verification and access. That's the moat.

The Counter

Timing and evaluation speed are the hidden hazards.

A thesis can be right and uninvestable — priced too early, or undone once models check their own work and the moat dissolves.

Raise evaluation velocity. Watch the moat collapse.
Market Expectation (Hype) Realized Cash Flows

Fast evaluation prices the gap away instantly.

Conclusion

Underwrite the bottleneck.
Not the label.

Don't pay for the label "AI." Pay for the bottleneck. What stays scarce when thinking is free? Who owns it? How much is already priced in? The returns live in those three answers.

01 / 07