Today, teaching an LLM anything durable about yourself is a per-provider chore. You explain who you are to ChatGPT, then again to Claude, then again to Gemini. Each one forgets between conversations. Switching providers, switching models, even upgrading to a newer release of the same provider can dump everything you’ve ever taught it.
A handful of vendor-specific workarounds exist — ChatGPT’s “Memory,” Claude’s project files, Gemini’s gems — but none of them are portable. Your memory of yourself is locked inside whichever vendor you happened to teach. If a competitor launches a better model next year, the cost of switching includes re-teaching the new one everything.