The developer tooling market has been reshaped twice in the past three years. The first reshaping was the arrival of AI code completion — GitHub Copilot, then a wave of competitors — which changed the micro-economics of individual coding: keystroke-level tasks got faster, boilerplate essentially disappeared as a developer time cost, and the average developer's output rate increased measurably. That reshaping is now complete. Code completion is infrastructure. It's bundled into IDEs, it's table stakes in any developer productivity pitch, and the marginal value of incremental improvement in code completion is declining as the capability matures.
The second reshaping is happening now and is more interesting from an investment perspective. AI has changed what gets written faster — but it hasn't yet solved the problem of understanding what was written. The ratio of code generation to code comprehension has shifted dramatically in the wrong direction for large codebases. Teams are generating more code than they can review, more changes than they can fully understand, and more dependencies than they can audit. The next wave of developer tooling is about giving development teams confidence in AI-generated code at scale: verification, testing, change analysis, and migration tooling that can operate at the velocity AI code generation demands.
The categories we're watching
AI-generated code testing infrastructure. Test generation has been a backwater of developer tooling for years — the tools existed, but developers wrote tests by hand because it was the only way to encode the intent the tests were supposed to capture. AI code generation has broken this equilibrium: there's too much new code being written for hand-written tests to keep pace. The tooling that automates test generation in a way that captures the correct invariants — not just the happy path — is a real gap and a real infrastructure investment opportunity.
Change analysis and impact tracing. In AI-assisted development workflows, the developer often doesn't have the same mental model of the change they just made that they'd have if they'd written every line by hand. Change analysis tooling that can explain what a proposed change does, what other systems it touches, what invariants it might violate, and what the risk profile of the change is at the codebase level is becoming a critical safeguard. This is the category that sits above static analysis and below full code review — and it's largely unbuilt.