AI assistant memory you can read, edit and delete

The first version of assistant memory in most products is a transcript: keep the conversation, feed it back next time, call it memory. It works for about a week. Then the transcript is longer than the context window and you are choosing what to drop, which means you were building a memory system all along and had not admitted it yet.

Three questions, and only the third is hard#

What should be remembered? Not everything anyone types. A question asked once is not a fact about a person. "Can you make me a board for the trip" tells you nothing durable; "I do not eat meat" tells you something forever.

How should it be stored? As something you can retrieve by meaning, because the useful recall is "did they ever say anything about their brother's wedding", and no keyword search gets you there.

Both of those have known answers. Extract candidate facts, score them, embed them, retrieve by similarity blended with recency and importance. It is engineering, it took a while, and none of it needed a new idea.

Who decides what is kept? That is the one that matters.

The default answer is the wrong one#

The default is that the system decides, silently, and the user finds out by inference. They notice the assistant knows something. They cannot see the list. They cannot correct an entry. They certainly cannot make it forget.

That is a bad deal even when the system is right, and when it is wrong it is worse than useless, because a false memory misleads every answer that touches it. An assistant confidently building on something you never said is harder to recover from than one that knows nothing.

So the memory is a screen. Everything the assistant durably believes about you is listed there in plain sentences, and every line can be edited or forgotten. The screen shows the actual contents, a list you can scroll, where most products offer a toggle labeled "personalization".

That page changed how we build the rest of it. Once a person can read what the machine wrote down about them, you stop being willing to write down guesses.

What we got wrong on the way#

We stored memories nobody could see for weeks. The embedding table was empty in production, and everything downstream degraded quietly into keyword search. Five separate bugs in a chain, each individually plausible, and every health check green throughout. Nothing was broken. It just was not there.

We remembered breakfast. The connection engine kept noticing that Tuesday's breakfast resembled Wednesday's breakfast, and reported it as an insight. It was correct and completely worthless. A system that cannot tell recurrence from meaning floods the useful signal out of its own output.

We collapsed people into preferences. Restoring an imported memory dropped its type, so a person became a generic fact about you. The graph still worked. It just no longer knew that your sister was a person.

The through-line in all three: memory fails politely. It does not crash. It returns something slightly worse, forever, until somebody looks at the actual rows.

Where the line sits now#

The assistant remembers facts, routines, and things it has concluded, each with an importance and each traceable to where it came from. It re-reads its own memory as content changes, so an edited note updates what is believed rather than leaving a stale copy behind.

And there is a region it cannot see at all. The Vault holds content the assistant cannot reach while it is locked: never retrieved, and never available to be paraphrased into a response about something adjacent. It hides things on screen too, though it is not a lock against the people you share the workspace with. Filtering sealed items out at the end was the obvious design, and we did not stop there, because a filter is a promise that holds until a bug. What the memory holds about a sealed item is encrypted under a key derived from your passphrase, and we never store that key in the clear. A query that slips past a filter finds ciphertext.

I would like to claim the line is in the right place. What I can actually claim is that you can see where it is, move it, and check afterwards that it moved.