June 16, 2026
Proof is in the repost
Formal Methods and the Future of Programming
After 25 years of saying no, Jane Street suddenly wants math to babysit AI code
TLDR: Jane Street says AI coding has changed its mind on using rigorous math checks to catch bugs, after years of thinking the process was too costly. But the visible community reaction was peak internet: instead of debating the big idea, one commenter instantly called the post a duplicate.
A big mood swing just hit the programming world: Jane Street, a company that spent decades basically shrugging at heavy-duty proof-based software checks, has now announced it’s all-in on them—thanks to the rise of AI coding tools. In plain English, the company says the old way of mathematically proving software was just too expensive and slow to make sense for most code. But now that AI can crank out lots of messy, bug-prone software fast, Jane Street thinks these proof tools could become the ultimate fact-checkers for machine-written code.
And then came the community reaction... which, in this case, was hilariously blunt. The loudest comment wasn’t a deep debate about the future of software at all—it was a classic internet record scratch: “Dupe”. Yes, the top visible response was basically someone saying, we’ve seen this already, complete with a link. That gives the whole discussion a very online flavor: while the article is talking about a dramatic philosophical turnaround, the crowd’s first move is to police the repost.
That mismatch is the real entertainment here. Jane Street is saying, “AI changed everything.” The community reply, at least from the visible thread, is: “Cool story, but mods?” It’s dry, petty, and weirdly funny—the kind of tiny comment-section drama that perfectly captures internet tech culture. Even when the future of programming is on the table, somebody is still here to yell that it’s a duplicate.
Key Points
- •Jane Street says it had historically viewed full formal methods as too costly for most of its software, despite valuing type systems and other reliability tools.
- •The article uses seL4 as an example of the cost of formal verification, citing 25 person-years to verify 8,700 lines of C.
- •Jane Street says agentic coding has changed the cost-benefit calculus by making formal methods easier to use and more broadly accessible.
- •The firm says it is now building a team focused on formal methods, aiming to make them as broadly useful as advanced type systems.
- •The article argues that AI-generated code increases verification needs and that formal methods, alongside testing techniques such as property-based tests and fuzzing, could help address that burden.