Open source · MIT
Gopnik
Your coding agent says it works. Gopnik tries to prove it doesn't.
An adversarial verification skill that attacks the claim that a change is done, checks the delivered revision, and reports what was and was not proven.
Independent technology workshop
Mad Duck builds open tools and shares practical field notes for people creating products with AI.
Explore projects
Projects
Our projects solve practical problems for teams that already build software with coding agents.
Open source · MIT
Your coding agent says it works. Gopnik tries to prove it doesn't.
An adversarial verification skill that attacks the claim that a change is done, checks the delivered revision, and reports what was and was not proven.
Field notes
Building with AI without magical thinking: experiments, tools, and lessons from real work.
Why coding agents struggle to verify their own work and how an adversarial approach exposes claims that remain unproven.
Read article