FaultWright — Verifiable AI Coding Tasks

Verifiable AI coding tasks built from real software failures.

by hello

Visit

Oct 6, 2026 AI & Machine Learning

ai coding auditable evidence autonomous tasks software failures verification

About

FaultWright is being built to autonomously execute permitted software-engineering AI-training tasks, independently verify the result, and produce auditable evidence.FaultWright turns working software into realistic AI coding tasks by introducing controlled, reproducible failures. AI coding systems attempt to repair them, and FaultWright independently checks whether the software actually works again. Each task produces clear evidence of the failure, the attempted repair, and the verified result. This makes the tasks useful for training, evaluating, and testing coding AI systems.