You tell ChatGPT how you like answers. Then Claude. Then the next agent, from zero, forever. There are three realistic ways out: per-app settings, a paste-block you maintain yourself, or one portable file that every AI reads.
ChatGPT has custom instructions and memory. Claude has projects and preferences. Gemini has saved info. Each works inside its own walls, none of them see each other, and every new tool means starting over. This is fine if you live in exactly one assistant and never plan to switch. Most people no longer do.
Write a short profile of yourself once, keep it in your notes, and paste it at the start of any AI conversation. This genuinely works and costs nothing. Its weaknesses are honesty and upkeep: self-written profiles say who you wish you were, and the file quietly rots as your work and life move.
A pupil is built from a short behavioral read, so it captures how you actually move rather than how you describe yourself. It becomes one signed file you add to any AI in about two minutes, you curate every line before it travels, and one tap revokes every AI at once, verifiable offline. The read is free, and the file stays yours.
Take the free four-minute read, get your word and your one-of-a-kind eye, and add your pupil to ChatGPT or Claude in about two minutes. Revoke it all with one tap, anytime.
Take your read, free Watch the revoke demoHow you like answers, how you work through a day, what you are focused on, and how to meet you when things are hard. What it should never contain is anything you would not want a third party holding, which is why a curatable, revocable file beats a raw braindump.
Yes, and it matters more for agents. An agent acting on your behalf needs consented context and a kill switch even more than a chat window does.