gist.design is a web standard and toolset for making product decisions readable to AI tools. It produces a /gist.design markdown file that sits at a project root — like llms.txt but for product intent, interaction flows, rationale, and boundaries instead of documentation.
AI agents misrepresent products — blending with competitors, inventing features, describing things wrong. Coding assistants build the wrong thing — they see structure but not intent, features but not rationale.
Every product has implicit decisions that never get written down. When AI tools encounter these gaps, they fill them with guesses from training data. The result is code that looks right but implements the wrong thing.
A guided conversation asks about your product decisions, positioning, and boundaries. The tool captures your thinking and produces a structured file that AI tools read automatically.
There are two ways to generate a file:
Both produce identical files. The file format is the product. The tools are creation channels.
gist.design is powered by aiuxdesign.guide, which documents 28+ AI UX patterns from 50+ shipped products. During conversation, the AI identifies patterns you are implementing and includes them in your file with links back to the pattern library.
gist.design is built by Imran. The project is open for community input. If you have feedback or want to contribute to the spec, reach out on GitHub.