Right now, every AI tool that remembers you builds its own private portrait. Claude keeps one version, ChatGPT keeps another, and none of them talk to each other. Duncan Davidson, an independent consultant and longtime software engineer, has been trying to connect those fragments, and his experience makes a strong case that the industry should stop treating memory as a feature of each product and start treating it as something a person owns.
Davidson's own workaround shows the problem. He first had his coding agents save useful information into Markdown files, then found that once commercial agents added memory, "Claude knew me one way and ChatGPT knows me another way." He could push context back to each tool with instructions and occasional reminders, but he acknowledged that this takes more tinkering than most people want to do. A system that only a determined engineer can maintain is not a solution for everyone else.
The more important point in the discussion is that portable memory needs boundaries. Davidson does not want a single file that exposes everything. He keeps general notes apart from information about his family and finances, and his test for a useful system is whether he can say, "Give Claude access to these scopes of my knowledge, but not everything." That is the right standard. Any personal memory layer should default to narrow access, with each request scoped and each scope visible to the person it describes.
The most interesting idea he offered is a phone-based memory that answers requests only with permission. A car dealership's agent, in his example, could ask for specific details, and the phone would seek approval before sharing anything. I think that model should be the baseline, not an optional extra. Companies that want information about customers should have to ask for it in a way the customer can see and refuse.
Davidson also left a harder question open: where an employee's own expertise ends and an employer's information begins. That gap will matter as personal agents encounter knowledge gained at work. Until the companies building these tools answer it clearly, people should keep the most sensitive parts of their lives out of any agent they do not yet trust.