After three years of limited impact, agent loops have recently transformed developer workflows completely, signaling the beginning of broader workplace disruption across every function.

All AI products are converging to horizontal platforms covering entire workflows (like Microsoft Office or Google Drive for human-agent interaction), with differentiation coming from produc…

Frontier labs cannot be model-agnostic because they're incentivized to sell their own tokens; application-layer platforms that let users switch models dynamically will win, analogous to ene…

Raising at reasonable valuations with minimal capital before product-market fit prevents down rounds and preserves optionality, contrasting with capital-intensive model training ventures th…

While the US offers easier fundraising and hiring, European founders can succeed by accepting trade-offs for sovereignty; investors ultimately follow traction regardless of geography, as sh…

Vertical AI products initially added scaffolding around weak models, but as models get better that scaffolding value disappears; defensibility shifts to network effects within the vertical,…

Flat seat-based pricing compresses margins as agent loops lengthen and token consumption grows; credit-based pricing aligns costs with usage and preserves margins at the application layer.

Frontier labs currently enjoy ~70-80% margins on model serving (evidenced by 9x price gap vs open-source equivalents like GLM 5.2 on Fireworks); open-source catch-up will compress these mar…

Open source models reaching parity with frontier models at 1/9th the inference price will force lab margin compression from 70-80% down, recalibrating the market and improving the competiti…