Coding Products Trigger Second-Order Effects

量子位 · wechat · 2026-07-13

This post discusses the assertion that Cognition and Cursor have already "cracked" the post-training recipe, and other teams will quickly catch up.

The author outlines several second-order impacts:

Demand for training infrastructure, model serving, and inference access will rise
Decreasing token costs will weaken the stickiness of closed-source models
Frontier labs will fiercely use AI products to build moats
User distribution, data, and environments will become key differentiators

Overall, it uses the progress of coding products to extrapolate changes across the AI supply chain and the broader product competitive landscape.

More from Embodied