A mirror has a strange job. To work, it has to know what you look like — and yet you should always be able to say that's not me and have it listen.
For most of Jubilancy's life, the second half was missing. Felt would read your reflections, distill what it noticed, and quietly hold a memory card about you: your themes, your values, your current season, the people who keep showing up. The card got smarter every week. It became how Felt grounded its responses — not a chatbot starting fresh each time, but a presence who remembered.
But here's what wasn't true: the memory was the AI's. You could see hints of it in the way Felt responded. You couldn't change it.
That's the wrong shape for a mirror.
What changed: a page for the holding
There's a new page at /memory. It shows you what Felt has been holding about you — your current season, your recurring themes, the people in your story, the self-description Felt has been refining as it reads. For now you can tap your current season and rewrite it. The themes, values, and people will become editable too in the next pass.
When you edit a field, your version wins. Until you release it back to Felt, no amount of new writing on your part will let the AI overwrite what you've claimed about yourself. You are not the AI's data. You are the author of your own season.
A second principle: every claim, traceable
Memory in your own hand is one half of the mirror promise. The other half is: every claim Felt makes should be traceable back to your own words.
So your Becoming Essay — the periodic prose synthesis Pro unlocks — now opens to a quiet toggle: show me what shaped this. Tap it, and you see the three to five reflections that most shaped what Felt wrote. Each one gives the date and a single sentence naming what that entry contributed.
Not a chart. Not a feature. Just receipts. The mirror should always be able to point to the moments behind what it sees.
A third: confidence, named
Some themes Felt notices have been with you for months. Others showed up twice in the last week. The old way: every theme landed with equal weight, which meant the AI either over-claimed on the new ones or under-emphasized the deep ones.
Now each theme on your Memory page carries a quiet label: strong (recurs across months), emerging (multiple recent entries), or tentative (thin evidence). Felt itself reads those labels when it reflects on a new entry — naming strong patterns directly, speaking about tentative ones only as observation.
The mirror moderates its own conviction. That's mirror voice at the data layer.
What the three changes share
The first is: you can edit me. The second is: here's why I said that. The third is: I know how sure I am.
Together they make Felt feel less like an AI and more like what we always meant for it to be — a witness that remembers, that can be corrected, and that names what it sees with the right amount of certainty for the evidence underneath it.
The product got quieter. Not less alive. More accountable.
Open Memory and see what Felt has been holding.