Enscrive Canvas is the thought processor that remembers. As you write, it will return your own drafts, fragments, and half-finished discoveries at the moment they matter — not generic AI suggestions, but your thinking, made available again.
The experience will feel ambient. The recall is measured — because it's built on an engine that already is.
There is a particular quiet that families pass down without ever naming it.
My grandfather kept his war in a drawer. We knew it was there the way you know a room is cold.
I keep wondering whether silence is a kind of inheritance — something handed forward
A word processor remembers the document.
A thought processor remembers the mind behind it.
As you write on the left, an intelligent margin will open on the right — surfacing fragments from this document, from notes you made months ago, from drafts you'd half-forgotten. Not search results you went looking for, but your own thoughts, returning on their own, because your earlier language is speaking to the sentence you're writing now. That's the experience we're building.
Canvas won't generate, summarize, or tell you what to think. It will return your actual past words, in their original form, and let you make the connection. It will feel like magic when the right thought comes back — and it will work because the recall is measured.
Fragments will surface in the rhythm of thought — gently, when you stop to think, never flickering as you type. Less like a search box, more like the quiet moment when something you'd forgotten drifts back to you.
Every document you write will live in the same memory. A note today could return a thought from a notebook you started a year ago. The longer you write, the more the canvas will have to give back.
Canvas will run on the same Enscrive engine you can use directly today — the same embedding models, the same neural search, the same evaluation loop. The fragments it returns won't be chosen by vibes. They'll be shaped, scored, and improved through evals — because that engine is already real, and already measured.
Some memories will come back brightly because they answer the moment directly. Others will arrive faintly — because they're older, adjacent, strange, or unfinished. The faded ones aren't wrong. They're farther away. And distance is exactly what lets a forgotten thought drift beside the one you're writing now, and quietly change it. That's the kind of memory we're building toward.
Bright memories keep you accurate. Faded memories keep you original. Hold a thought close and it will return more readily; let one grow quiet and Canvas will keep it faintly within reach — the way an old idea can still color today's page. Both kinds of memory, working beside you.
Canvas will be a customer of Enscrive — using the same embedding models, the same neural search, the same /v1 surface, and the same eval loop available to you today. No private backdoor. No special treatment. The instrument we'll trust with our own thinking is the one already in your hands.
And because Enscrive is built to be measured, recall quality is never a matter of faith. Every embedding model that shapes a memory is scored against retrieval tasks — better recall isn't merely admired, it's proven, compared, and improved. This is recall you can measure, not just trust. That is the platform creed, true today: semantics evolve through evaluation, not assertion.
Your chosen embedding model decides how a thought is represented, searched, and returned — tuned for the way a writer's mind actually moves.
Enscrive evals measure whether the right fragments came back, using the same retrieval workflow offered to every developer.
NDCG@10 · Recall@10 · MRR
Improvements are judged by evidence, not assertion. Recall gets better over time — and the gains are demonstrable.
Recall without trust can feel uncanny. So the foundations aren't promises we hope to keep — they're how Enscrive is built, today. Your data is never used for training. It is yours. You can export all of it, anytime. And every fragment Canvas returns to your margin will remember where it came from — the draft, the date, the trail back — so you'll always know a thought is yours, and where it lives. The guarantees are real now; the canvas that rests on them is what we're building.
Your writing is never used to train a model — not ours, not anyone's. Canvas will remember your thoughts so they can return to you. They are never raw material for someone else's system.
Every tenant on Enscrive is strictly isolated — zero leakage between workspaces, platform-wide, enforced in the engine itself. No shared pool, no platform-tenant bypass. Your canvas is yours alone.
Data written in the US stays in the US; the EU in the EU; AP in AP. Jurisdiction is pinned, with no opt-out and no crossing — so where your thinking lives is never in question.
Export everything, anytime — full portability, no lock-in, built into the platform today. Follow a fragment back to its source, and carry your whole canvas out whenever you choose. The memory will be yours to keep and to move.
A canvas will be an open page. Click and a document appears; start writing. Bring in work you already have, or start from a blank line. Each canvas will hold a single, consistent memory — chosen once, when you create it.
As you finish thoughts, Canvas will quietly turn them into recallable memory — keeping only what's new or changed, leaving the rest untouched. You won't manage this. You'll just write.
The moment you pause, your chosen embedding model will retrieve what's relevant from across the canvas — bright fragments that answer the line, and faded ones that arrive from farther away. You read, recognize, and write on.
Hold a thought close or let one grow faint. Follow a fragment back to its source. Gather documents into portfolios. And beneath it all, the recall loop will keep getting better — because the engine it's built on is evaluated continuously, today.
For the essayist circling an idea across a dozen sittings. The researcher who can never find the note she knows she wrote. The founder whose best thinking is scattered across a year of fragments. Canvas is being built for anyone whose mind moves faster than any single document can hold.
Not a notebook you have to organize — a memory that will organize itself around what you're thinking about right now.
Canvas will be a writing surface for you — no AI looking over your shoulder. But the memory beneath it will speak the same language your agents do. Point your own reasoning model at your canvas through the Enscrive platform, and it will recall, reason over, and build on everything you've written.
We build the instrument. You bring the intelligence. Your memory stays yours to steer. And for organizations that need it, Canvas will run on a dedicated Enscrive instance — isolated entirely to you, pinned to the single jurisdiction you choose.
This is the Canvas we're building — on an engine that's already real, already measured, already yours to use. When it opens, the longer you stay, the more it will have to return to you.
Canvas will be the first place most people feel Enscrive.