Boris Cherny names latent demand as “probably the single biggest principle in product” and traces every major Claude Code feature to it. The principle: “People will only do a thing that they already do. You can’t get people to do a new thing. If people are trying to do a thing and you make it easier, that’s a good idea.”
Three concrete examples from Claude Code’s history:
CLAUDE.md — Engineers were already writing markdown files for themselves and having the model read them. Boris didn’t invent the concept; he observed the behavior and formalized it. “This is where CLAUDE.md came from.”
Plan mode — Users on GitHub issues were already saying “hey Claude, plan this out but don’t write code yet.” Boris saw this on a Sunday night at 10 p.m., wrote plan mode in 30 minutes, and shipped it Monday morning. “Literally all plan mode does is add one sentence to the prompt: please don’t code.”
Co-work — Non-technical employees at Anthropic were “jumping over hoops to install a thing in the terminal.” Every designer, the entire finance team, and the data science team were using Claude Code — not for coding. The latent demand for a GUI wrapper was obvious.
This connects to Revealed preferences trump stated preferences — track what users do, not what they say: Boris doesn’t do user surveys — he walks around the office standing behind people watching how they use Claude Code, and reads GitHub issues to see what people are already trying to do. The insight also extends Compound engineering makes each unit of work improve all future work: when each feature comes from observed demand rather than speculative planning, it compounds because real users have already validated the need.