Your AI health coach needs a record, not a memory
Chatbots “remember” by rereading conversations. That breaks quickly for health, where what matters is structured, correctable, longitudinal data — and proof of every write.
Tell a general-purpose chatbot what you ate today and it will respond warmly, maybe even usefully. Do it for a month, then ask “how has my protein trended since I started lifting?” You’ll get something fluent and confident, but you won’t get your actual protein trend, because nothing in that system ever stored one.
Memory is not a record
When a chatbot “remembers”, it’s rereading text: conversation history, retrieved snippets, a notes file. That works for preferences and context. But health questions are aggregation questions — averages over weeks, trends against targets, this month versus last — and text doesn’t aggregate. You can’t do arithmetic over “I had a big lunch, maybe 700 calories?” scattered across ninety conversations. All the model can do is improvise over fragments, and models are much better at filling gaps confidently than at admitting them.
A record is a different kind of object, and three properties make the difference:
- Structured. A logged meal isn’t a sentence in a transcript. It’s foods, portions, calories and macros attached to a date, which is what makes “average protein, Tuesdays excluded” a computation instead of a hallucination.
- Longitudinal. Its unit of meaning is the trend, not the message. Month-old data isn’t stale context; it’s the baseline your progress is measured against.
- Correctable. Records have edits, removals and ground truth. A transcript where you misspoke just permanently contains the mistake.
The trust loop: prove every write
A record the AI both reads and writes raises an obvious question: what exactly is it writing? Most assistants answer with “Got it, logged!” and leave you to check. Moraine’s answer is a visible write-trail. Every write appears as a chip under the reply (2 foods added, 1 exercise updated), and hovering shows what was written, old value to new on corrections. Say “actually it was three eggs” and you can watch the record change:
The same principle applies to numbers flowing the other way. In the weekly recap, every figure is computed from the record by ordinary code, and the AI narrates without being able to touch the arithmetic. Where numbers enter the record as estimates, like photo logging, each one is labeled as verified against a database or marked as an estimate. That chain is what lets a coach say “protein has been under target on training days” and have it be true.
A record is also something you can own
Structure has a second benefit: a record can be handed to you in a way a scattered chat history never can. In Moraine, one tap exports everything (journal, goals, targets, all of it) as JSON you can keep, inspect or move elsewhere, and one typed confirmation deletes all of it, including uploaded files. The chat itself clears after seven days by design. It’s the scratchpad; the record is the asset, and it’s yours.
The test to run on any “AI health coach”
Three checks tell you whether there’s a real record underneath:
- Ask for a specific aggregate — “average sleep over the last two weeks?” — and check it against reality.
- Correct something — “that run was 5k, not 10k” — then ask again tomorrow and see which version survived.
- Ask to see everything it has on you, in a file, today.
A system built on a real record passes all three without ceremony, and a system built on chat memory fails at least one. That gap is the product we’re building (here’s what exists today), but the test is worth running on anything that offers to coach your health, ours included.