Most note apps are a place you put things. With an embedding model and a chat model installed (see Local AI), MojoPad also reads back over what you've written — on your Mac — and surfaces the things you'd never think to go looking for: the note you forgot, the place two of your own notes disagree, the moment your mind changed. Everything here is a suggestion you curate, never an automatic edit, and each lives as a palette tab (see Palettes) you can drag into whatever order you like. The model can misjudge — treat these as prompts for your thinking, not verdicts.
The Rediscover tab brings back notes you haven't touched in a while that relate to what you've been writing lately. It ranks them by two signals at once: how forgotten a note is (how long since you last touched it) and how relevant it is to your recent writing — so what surfaces is an old note that bears on what's on your mind now, not a random one. Each card explains itself in a line ("you've been writing about onboarding; you worked this out last year"), and ✦ Why now asks the model for a sharper, specific reason. A small count on the tab shows how many are waiting. For each one you can open it, Connect it to the page you're on (drawing a real link), or Dismiss it so it never resurfaces again.
Rediscover needs an embedding model and a little history — notes return once they've gone a couple of weeks untouched and something you've written since relates to them.
Backlinks and Related surface similarity. Insight more often comes from tension — the thing you believe that an older note of yours disputes. The Provocations tab reads the notes that relate by meaning and asks your local model where two of them genuinely contradict each other, then states the disagreement in a sentence. Click ✦ Find a provocation and it checks the pairs touching your most recent writing first. This is deliberately rare — most notes simply agree — so it tells you when it checked and found none, and Find more digs deeper. Each provocation links to both notes, so you can reconcile them, supersede one, or keep both.
MojoPad timestamps every page and keeps a version history, so a local model can tell you not just what you wrote but how your thinking has moved.
The Bisect tab answers "when did I change my mind about this?" Type a belief —
"cold showers are worth it" — and MojoPad gathers the notes about it, lines them up oldest
to newest, and binary-searches the model's read of each one to pinpoint the note where your
view turned. Like git bisect for your own thinking: it reads only a handful of
notes to find the pivot, and shows you the turn and the note just before it.
Open a page's history (its menu ▸ Page Versions…; see Recently Deleted and Backups), pick an earlier snapshot, and click ✦ How my thinking changed. Instead of a line-by-line text diff, the model reports what shifted in meaning — a claim added, dropped, reversed, softened, or strengthened — ignoring pure rewording.
The Shifts tab is a deliberate record of changing your mind. Press ✎ Record a belief change, jot what you used to think and what you think now (and optionally why), and it joins an "I used to think X, now Y" timeline — each entry stamped with the date and the page you were on. Select the old wording in a page first and it's filled in for you.
The Lint tab is a "problems panel" for meaning, not spelling — like the warnings panel in a code editor, but for your ideas. Press 🔍 Check this page and your local model flags:
Every finding is a suggestion you act on or ignore — MojoPad never edits the page for you.
Like the rest of MojoPad's AI, everything here runs against Ollama on 127.0.0.1;
encrypted pages are excluded; and if no model is running, each tab explains what to install while
the rest of the app carries on. See Local AI for setup.