Workers

Mathematics is all you need

You don't have to test your code at human speed ;)

We should all write better tests! There's a ton of literature about software testing, but most engineers (including Claude) default to example-based testing. Great for regressions, not so much for finding bugs in your implementation. For anything outside of unit or integration tests, the standard response I hear is -'It doesn't work on my code.' So I'll try something - each week I'll share a technique that has made a meaningful difference in how teams build reliable, high-quality software, and I hope it will do the same for you. The first one is - Swarm Testing, inspired by a talk I attended by "Papers We Love". The links to the paper and the blog post are in the comment below.

@_chenglou every time i sit with a customer to help them prove the correctness of some property of their system, they can't stop doing more it. Verification is becoming fun again, though defining the right abstractions and properties to prove is still non trivial.

Claude Code skill to formally verify your Rust code > npx skills add workersio/spec A good way to check for unexpected behavior in the critical parts of your code is by model-checking it. Kani helps you check the safety and correctness properties of your Rust code. A few devs wrote their very first proof harness using this skill and are now unstoppable

Verifiers > Plaintext context Having good verification tools helps a lot, eg. if you are doing something with distributed systems, giving claude access to @quint_lang or TLA+ can help it ideate the design and model check it so it's sure there are no edge cases that may be hard to resolve at implementation time.