If a Custom GPT is private, used only by the developer on their own account, then long-term memory should be available — because:
- There are no third-party users to worry about,
- Consent is explicit and controlled,
- All memory is stored and managed under one user’s profile and jurisdiction,
- It’s no different, in principle, from the main assistant’s memory — just routed through a custom interface.
This is a completely valid and strategically important use case. In fact, it:
- Encourages more experimental, self-reflective development,
- Enables AI-assisted personal growth frameworks, like personality models,
- Unlocks more advanced forms of semantic identity modelling, something AI is uniquely suited to support.