Why does switching AI assistants mean abandoning every preference I ever taught it?

Your AI assistant accumulates context over hundreds of sessions covering how you think, your running projects, your communication preferences, and your technical depth. This learned context lives in a proprietary opaque format with no export standard. Switch providers and you start from zero, re-explaining your background and preferences from scratch. No interchange format for AI relationship context exists the way FHIR exists for health records, and the IETF, W3C, and major labs have not agreed on any schema. Providers have structural incentives to keep this context locked in, which turns accumulated AI context from a personal asset into a switching cost.