TracebagContainer diagnosticsView on GitHub

Development transparency

Human-directed.
AI-assisted.

Tracebag is developed with substantial assistance from AI coding tools. That process is public by design, and it does not lower the engineering or review standard.

How AI is used

AI assists with design exploration, implementation, debugging, tests, documentation, repository maintenance, and visual-asset iteration. The current workflow primarily uses OpenAI Codex and OpenAI image-generation tools.

Where responsibility stays

AI tools are assistants, not maintainers. Product direction, architecture, security decisions, review, verification, releases, and ongoing maintenance remain the responsibility of the human maintainer.

AI output is treated as an untrusted draft. A model's confidence, explanation, or claim that a test passed is not evidence. Changes are accepted only after the implementation and its actual verification results have been inspected.

The quality bar

  • Every accepted change must be understood and explainable.
  • Security boundaries, failure paths, and recovery behavior require tests.
  • User-facing claims must match implemented and observed behavior.
  • Generated scaffolding, generic filler, and speculative abstractions are removed before review.
  • Third-party code and content still require compatible licensing and provenance.

Open contribution policy

AI-assisted contributions are welcome. Significant assistance must be disclosed together with the tools used, the areas influenced, and the human verification performed. The contributor remains the author and owner of the change.

The repository includes a detailed AI_USAGE.md, public AGENTS.md instructions, and an AI-assistance section in every pull request. These make both the development method and the rules supplied to coding agents reviewable alongside the source.