Every AI tool you use today builds its own private idea of who you are — your goals, your history, your preferences — and locks it inside that one tool. Switch products, lose the context, start over. There's no shared, portable way for a person's own context to move safely between the AI tools they use, the way a resume or a contact card can move between systems. As more of daily and working life runs through AI, that gap gets more expensive to ignore, not less.
How we're doing it
This is early, honest research-and-standards work, not a shipped product — we're building the structure first and earning the standard, not claiming one. We've completed a first research phase across the core questions (how context stays portable between tools, how it stays provably intact, how a tool proves it's actually following the standard), settled the first round of architectural decisions, and cut a first version of the schema itself with a passing test suite behind it. The next phase is a single, specific engineering gate we're working through before the reference implementation moves further.
Where it stands today
The Loadstar Project is a real, in-progress research effort, not yet a product you can use. The landing page is open for anyone who wants to follow the build and be notified when there's something to see — one signal when it matters, nothing before that.