Builders are making their own tools, and liking it
The Ask HN thread 'What tools have you made for yourself since the advent of AI?' is one of the more energetic threads today. The answers are concrete and varied: a daemon that watches shell history and looks for automation opportunities, a CLI written in Go by someone who knew zero Go, a custom CRM with domain-specific fields, language tutors, mountain bike trail map generators, and multiple custom AI coding harnesses.
The pattern here is that AI has lowered the floor on personal tooling enough that people are building things they would never have attempted before. Importantly, several people in the thread are emphatic that they're using AI as a code generator, not an agent, and they want minimal toolchains. There's a healthy skepticism of the AI-dev-tools category embedded in people building their own versions of exactly those tools.
This matters for the developer tools market. The people who care most about their tools are increasingly opting out of commercial offerings and building custom. That's not fatal for the category, but it means the commercial tools need to serve people who don't want to spend time on this, not the power users who are already rolling their own.
So what?
If you're building developer tools, your most demanding potential customers are already solving the problem themselves. That's a signal to either go further up-market, toward teams and enterprises who won't self-build, or to make your product so fast to adopt that self-building isn't worth the comparison. The people building their own tools are also your best source of insight into what's actually missing.