Will Wilson: Antithesis uses deterministic simulation to find rare bugs — Jane Street went from customer to strategic investor
Dec 8, 2025 with Will Wilson
Key Points
- Antithesis raises $105 million Series A led by Jane Street, which converted from customer to investor after stress-testing the platform in production.
- The startup uses deterministic simulation to reproduce rare bugs reliably, eliminating the 'works on my machine' problem that plagues conventional testing.
- Wilson argues traditional testing cannot find the bugs that cause real disasters because they require anticipating scenarios developers never thought to test.
Summary
Will Wilson is the founder of Antithesis, which raised a $105 million Series A led by Jane Street on December 3rd. The investor list includes Dorri Patel, Shelto Douglas, and Patrick Collison — notable partly because Collison and at least one other investor are former roommates of Wilson's.
The company's premise is blunt: software has been built and tested wrong for 70 years, and developers have developed a kind of Stockholm syndrome about it. Users tolerate random failures from software that they would never accept from a physical structure.
The problem with conventional testing
Traditional testing asks developers to anticipate every scenario, write manual test cases, and run them through a CI system. Wilson argues this fundamentally can't work — the bugs that cause real disasters are the ones nobody thought to test for.
How Antithesis works
Antithesis takes a customer's actual software and runs it inside a simulation of the real world, stress-testing it with adversarial users, crashing infrastructure, and hacking attempts. Customers specify what their software should never do, and Antithesis tries to make it do exactly that.
The "deterministic" part is the core technical differentiator. Antithesis has built a way to run any computer program fully deterministically — meaning if the system finds a bug that only manifests one in a billion executions, it can reproduce that exact failure perfectly, every time, until the engineer fixes it. This eliminates the "works on my machine" problem entirely.
Jane Street went from customer to lead investor, which is a reasonable signal about how the product performs in high-stakes production environments.
The AI tailwind
Wilson's longer thesis is that an AI-driven future means dramatically more software, and a significant portion of it will be written by agents rather than humans. Testing code you didn't write, and don't fully understand, is where deterministic simulation becomes critical rather than just useful.