a builder's codex
codex · operators · Boris Cherny · ins_latent-demand-as-search-algorithm

Latent demand — what users hack around the product to get — is the next loop

By Boris Cherny · Head of Claude Code, Anthropic · 2026-04-27 · podcast · What happens after coding is solved

Tier A · TL;DR
Latent demand — what users hack around the product to get — is the next loop

Claim

The most reliable signal for what to build next is what users are already hacking together inside or around the product to get a job done. In AI-native products, this extends to watching what the model is trying to do but cannot reach — clear that path.

Mechanism

A user willing to abuse a primitive to get a job done has revealed both the demand and the rough shape of the solution. The cost of discovery is paid by the user, not the product team. Building toward that hack converts a workaround into a primary surface and concentrates demand that was already there. The AI flavor: when the model "wants" to take an action but is missing a tool or context, the gap is itself a roadmap.

Conditions

Holds when:

Fails when:

Evidence

"In research we call this being on distribution."

Two flavors Boris named:

— Boris Cherny on Lenny's Podcast, 2026-04-27

Signals

Counter-evidence

Hacks are not always demand — sometimes they are bugs that mask the right product. And not every model refusal is a missing tool; some are correct safety boundaries that should not be cleared. The skill is reading which is which.

Cross-references

Open the interactive view → View original source → Markdown source →