A user suggests the need for a standardized identity file format that can be used across different AI tools to maintain user preferences and context, preventing the need to re-explain oneself in each tool.
Ok this might be dumb. Spent a lot of time loking at llms.txt and thinking about content and ai AUTHORSHIP. So I made identity.txt, does the same thing as llms.txt for people. The problem: every AI tool has "custom instructions" but they're siloed. Switch tools and you lose everything. Your tone, your expertise, your preferences. You end up re-explaining yourself constantly. identity.txt is just a markdown file. Same idea as llms.txt, humans.txt, robots.txt. You write it once and it works everywhere. Paste it into ChatGPT, Claude, Gemini, wherever. Or host it at [yourdomain.com/identity.txt](http://yourdomain.com/identity.txt) and link to it. What's in it: \- Your name (H1 heading) \- Sections like ## Voice (how you write), ## Expertise (what you know), ## Preferences (hard rules) \- A ## Terms section - basically robots.txt for your identity. We're also experimenting with hosting at [identitytxt.org](http://identitytxt.org) where you sign in with Google and get a permanent URL. But honestly the spec is the point, not the service. Self-hosting works fine. This is very early and experimental. We're trying to start a conversation about portable identity for AI, not ship a finished product. The spec is CC-BY 4.0 and completely open: [https://github.com/Fifty-Five-and-Five/identitytxt](https://github.com/Fifty-Five-and-Five/identitytxt) Would love to know: do you find yourself re-explaining who you are to AI tools? Is a file convention the right answer or is there a better approach? [https://identitytxt.org](https://identitytxt.org)