ADRs for AI
A methodology extending the Architecture Decision Record (ADR) pattern to human-AI collaboration, using Git-versioned records of prompts, AI reasoning processes, human choices and corrections, and other deliberation content to make AI-assisted decisions traceable, reproducible, and auditable.
Timeline (last 90 days)
ADRs for AI was discussed in the Show HN section of Hacker News.
The core proposition of ADRs for AI is to produce a "deliberation record" from every collaboration with AI, containing the prompt, the AI's reasoning process (if obtainable), the human's choices and corrections, and the finally adopted solution, and to commit it to a Git repository.
ADRs for AI adds new content such as prompts, model versions, and AI reasoning summaries on top of the original ADR fields.
Some users in the Hacker News comment section mentioned that the ADRs for AI approach is feasible in personal projects, but scaling it across a team requires accompanying template standards and automation tools.
All Facts (4)
The core proposition of ADRs for AI is to produce a "deliberation record" from every collaboration with AI, containing the prompt, the AI's reasoning process (if obtainable), the human's choices and corrections, and the finally adopted solution, and to commit it to a Git repository.
50%UnverifiedADRs for AI adds new content such as prompts, model versions, and AI reasoning summaries on top of the original ADR fields.
50%UnverifiedADRs for AI was discussed in the Show HN section of Hacker News.
50%UnverifiedSome users in the Hacker News comment section mentioned that the ADRs for AI approach is feasible in personal projects, but scaling it across a team requires accompanying template standards and automation tools.
50%