AI-Assisted Software Development Research

This site is an evolving research corpus on AI-assisted software development.

It studies what breaks in one response or across change, what support artifacts help preserve truth, and where authority does and does not live when software work is mediated by AI.

The corpus includes papers, experiments, protocols and contracts, and archived lineage from earlier phases of the work.

Earlier phases tested stronger contract-centered claims. Some of those claims did not hold. The work continued by narrowing, revising, and following what survived.

Current threads include VDG, omitted scope, blended inference, default fill-in, review burden, preserved truth, and refusal surfaces.