Formal verification gets a second look as AI writes more code
A thread on formal methods and the future of programming drew a pointed observation: as generative AI writes more code, the scarce human skill shifts from writing to verification. If AI can produce plausible code at volume, the bottleneck becomes proving that the code does what you think it does. Several commenters said they have been actively testing formal verification tools on real products and want to understand what works at scale.
The question the thread could not fully answer is whether formal methods and AI are complements or competitors. One reading is that AI helps write the formal specs faster, making the verification step accessible to more teams. Another reading is that AI-generated code is harder to formally verify because it lacks the explicit structure that formal tools expect.
This is a live debate, not a settled one. But the signal is clear: developers who spent the last two years asking 'how do I generate more code with AI' are starting to ask 'how do I trust the code AI generated'. That is a meaningful shift in what the tooling market needs to build.
So what?
If you are building developer tools, the next underserved category is not code generation but code verification. Formal methods have historically had steep learning curves and niche audiences, but the combination of AI-generated code at volume and rising software reliability stakes is creating a new customer who was not there before. The founders who figure out how to make formal verification accessible to mid-level engineers will own a real market.