Picking Up Where You Left Off, With No Account Anywhere

Most products solve continuity with an account: you sign in, a server remembers, and the next device catches up. This one does not have that option, because there is no account and nothing about you exists on a server. What it has instead is the browser's own storage, and three pages that read it for three genuinely different questions.

The interesting part is that the constraint produces a better answer for the common case. A person returning to a study tool on the same laptop, which is the overwhelmingly common case, gets everything back with no sign-in step at all.

• My Learning answers what should I do next: next actions, progress across the sections you have used, your projects, career plans, and a local backup control.
• My Library answers where is the thing I made: generated projects, saved topic shortcuts, and the share links you created.
• Recent activity answers what was I just doing: a short history of what this browser opened last.

Progress here is genuinely spread out. Aptitude attempts, cognitive-skill history, reasoning results, puzzle progress, flashcard schedules, career answers and generated projects are separate stores with separate lifecycles, and no server is joining them up. Without one page reading across all of them, a returning learner has to remember which section they were in.

So My Learning is the join, computed on the device at the moment you open it. It is not a record that gets written; it is a view over records that already exist, which means it cannot drift from them.

That distinction matters for a reason beyond correctness. A dashboard that maintained its own summary would need to be updated by every feature that changes anything, and the first feature that forgot would produce a dashboard that was confidently wrong. A view recomputed on open is slower by an amount nobody can perceive and cannot be stale by construction, which is the right trade for a page whose entire job is to tell you the truth about your own progress.

• Preferences, consent, saved topics and recent activity: localStorage.
• Generated projects, uploads, transcripts, edits and study progress: IndexedDB.
• Practice attempts, cognitive-skills history, reasoning results, puzzle progress and career answers: also in this browser, none of it on a server.

You lose all of it by clearing site data, by using private browsing, to aggressive browser cleanup, to storage eviction under disk pressure, by switching browser profile, or by moving to another device. There is no account, so there is no copy to restore from and nobody to ask.

That is the trade the product makes, stated as plainly as it deserves: nothing about you exists anywhere else, and nothing about you can be recovered. Anyone telling you that a local-first product has no downside is selling something.

The next-action suggestions on the dashboard are computed from what this browser has recorded: which lesson you were partway through, which flashcards are due, which suite you started and did not finish. That is a genuinely useful signal and it is a narrow one. It does not know what you are studying for, when your deadline is, or that you abandoned a section deliberately.

So the dashboard suggests rather than instructs, and it stays easy to ignore. A next-action list that assumed it understood your goals would be wrong most of the time and confidently so, which is a worse failure than being obviously incomplete. The one thing it does know well is what you left half-finished, and half-finished work is where returning learners lose the most time.

Export before you clear anything. Every substantial surface writes real files: a paginated PDF, a spreadsheet, printable HTML, plain text and JSON, assembled in the browser from data that is already there. Those are ordinary files on your own machine, and they survive every failure mode in the list above.

A useful habit is to export at the end of a study block rather than when you remember to. The export takes seconds and the alternative is discovering the gap after a browser update has already cleaned up.

If you move between two machines regularly, treat the exports as the record and the browser as a working copy. That is a slightly old-fashioned way to work and it is the only one that is true here: there is no sync, and a design that pretended otherwise would lose somebody's work eventually.