Niko asks before it remembers

An AI that quietly remembers things about you is exactly the thing people are right to be wary of. Useful guidance does lean on context — an injury that flares on hills, the hours you will not train, the race you are actually pointing at — so the real question is not whether Niko remembers, but who decides what it keeps. We built its memory around one hard rule: Niko can propose something to remember — it cannot save it.

The model has no write path to your memory. When Niko thinks a fact is worth keeping — an injury, a constraint, a goal — it proposes it, in your own words, and waits. Nothing is written until you confirm on a separate turn, in plain language. If you never say yes, nothing persists.

That shape was a deliberate choice, not the default. The simpler version — let the model write straight to your record — was on the table, and we did not take it, because a guide you can trust is one that only remembers what you told it to, and because “the model literally cannot write to your memory without your yes” is a promise you can verify rather than a reassurance you take on faith.

What we do not know yet is how that confirm step feels over months of real use — whether being asked every time reads as trust or as friction. If you have a view on what Niko should have to ask before it keeps anything about you, that is the part we most want to hear.