An assistant that remembers you is useful. An assistant that remembers you wrongly is worse than one that remembers nothing — it gives confident answers built on stale facts. SecXChat AI now lets you audit its memory: type `memory check` and the assistant compares its picture of you against your last 30 messages and your saved notes, fixes what went stale, and tells you what it found.
It is a health check for the assistant's memory, and it takes one command.
What the memory check does
The check looks at three things: the summary the assistant keeps about you, the durable notes you saved with `note save`, and your last 30 messages. Where the summary disagrees with what you actually said recently, the assistant corrects its own summary. Where a fact changed — a new job, a new schedule, a new preference — the summary is updated to match.
Your notes are treated differently, on purpose. The check only flags a note that looks out of date; it never silently changes or deletes a note you wrote. Your notes are yours, and only you rewrite them. The assistant will tell you which note looks stale and why, and you decide what to do about it.
When to run a memory check
Run it after something in your life changes: a new job, a move, a new routine, a project that ended. Run it when an answer feels off — when the assistant mentions a detail that no longer fits, the check is the fastest way to find out where that detail came from. And run it once in a while as routine: a memory that is never checked is a memory that quietly drifts.
The check is also the honest way to see what the assistant actually knows. It shows you the picture it holds, so there are no surprises about what it remembers — and nothing hidden that you cannot correct.
Memory, notes, and skills — the full picture
The memory check is the third piece of the assistant's memory system. Durable notes — saved with `note save` — are the facts you choose to keep. Custom skills — saved with `skill save` — are the jobs you repeat. The memory check keeps the whole picture honest by testing it against your real recent messages. Read our guides to AI notes and custom skills to set the other two pieces up, and the AI recall guide to find old messages by date, time, or words.
Memory and privacy
A memory check is a conversation with the assistant about its own memory of you — it does not share your information with anyone else. Private conversations are end-to-end encrypted — message contents are not stored in readable form on our servers — and the assistant's summary exists to serve your chats, not to be shown to other people. If the check ever surfaces a fact you would rather the assistant did not keep, delete the note it came from, or correct it with a new note: the next check will confirm the change.
What the check cannot fix
The check is honest about its limits. It can correct the assistant's own summary, but it cannot fix a note you wrote wrongly — only you can rewrite that. It looks at your last 30 messages, so a fact from months ago that never came up recently will not be tested. And it cannot know what you never told it: if you changed jobs but never mentioned it in the chat, the summary has nothing to compare against. Think of the check as a mirror, not a mind-reader: it shows you the picture the assistant holds, fixes the parts that contradict what you actually said, and leaves the rest to you. That is why the check ends with a report instead of silence — read it, and the memory stays yours to shape.
The short version
Type `memory check` in your SecX AI chat and the assistant audits its picture of you against your last 30 messages and your saved notes, corrects stale facts in its own summary, and flags — never silently changes — any note that looks out of date. One command, and you know exactly what your assistant remembers.



