Memory
What already happened to you comes back when it matters. "This looks familiar" is not a nice phrase: it is an earlier analysis of yours, of the same asset, with its date.
“This looks familiar” has to be provable
It is not a feeling APEX simulates. It is an earlier analysis of yours, of the same asset, with its date, its thesis and how it ended.
APEX's memory is dated facts with their evidence. Each episode stores which asset, on which timeframe, what you said you were looking for, what APEX answered and —if the trade existed— how it went and what the result was attributed to.
When that memory appears in a new analysis, it travels with its denominator: “similar to 3 earlier ones on BTC”. Not with a confidence synthesized out of nothing — deriving a 0.72 from the sample would have created a second confidence scale different from the engine's, and two scales that do not match are worse than a single imperfect one.
The memory echo lives inside “This is yours, not the market's”. It does not open its own section: “similar to 3 earlier ones on BTC” is exactly that.
Chosen by context, not by date
What you did with this same asset is more relevant than what you did with another one on the same day.
An analysis of yours on the same timescale talks about the same kind of decision.
If your current thesis and an earlier one talk about the same thing, that overlap counts. With a cap of two concepts, so a long thesis does not drag the whole memory along through loose coincidences.
All three break ties between comparable candidates; they do not discard. Someone with little history cannot be left without memory, and an analysis from two months ago does not jump ahead of one from ten days ago just for sharing the timeframe.
Memory is a snapshot; the truth is your data
The model of you that APEX keeps is a cache, not the source. The truth is still your analyses, your trades and their events: the whole model can be rebuilt from there at any time.
And that cache has no write policies from the browser, on purpose. Your browser cannot claim hypotheses about yourself: what APEX holds about you is derived by the server, with the data the server can verify.
The signals you see next to an analysis are saved with it. They are not recalculated when you reopen it from your history: showing today's context next to last week's opinion would rewrite the why of that opinion.
APEX cannot say “this pattern got stronger since your last visit”: it does not store visits. What it can say is what changed against a snapshot of the model from at least a week earlier, with that date. It is a data gap, not an interface gap, and it is declared instead of approximated.