the state of developer tooling (wip)
a work in progress, just mapping out my thoughts on the state of developer tooling for a full post.
I think you could argue that developer tooling is dead, but also possibly thriving more than ever.
Dev tooling has always been a busy startup space, and arguably saturated. Engineers want to build something 'cool' (and often not very useful), but most outside of software won't appreciate, or they want to build for just themselves and happen to strike a gold problem.
Agents (LLMs) are putting things into a weird state. Devs are now a lot more hands off, leaving their agents to use all the tools. This means that some things will die, especially if they can't adapt or were just a nice UI.
However, as we need refined control over agents, and as we do more, other tools are needed more. You need better source control, you need better testing and testing workflows, you need better builds, you need better deployments, better monitoring and better automated triage. There's probably tons more here.
So, is tooling dead? No. Far from it.
LLMs are only getting better at engineering; 'this is the worst they'll ever be'. Tooling needs to get better, and work well with them. There's now even more opportunities, but they've changed. The need is different.
It's shakeup time.