Fable 5, feature by feature — the deep dive for reading your own health.

Most people meet Fable 5 through the one feature it shares with every model before it: a box you type into. Ask a question, get an answer, close the tab. Judged on that box alone, it is a slightly better search engine. But the box is the least interesting part. The upgrade that matters lives in the features nobody reads the release notes for — memory, uploads, projects, scheduled actions — and those are the ones that turn it from a thing you consult into a thing that reads your own body on your behalf.

This is a tour of those switches. Not the settings menu — the small number of capabilities that actually change what the tool is for. Think of it as the difference between owning a car and knowing it has a gearbox. You can drive in first the whole time. You will just wonder why everyone else is getting further.

feature one: the context window is a memory, not a bigger box

The specification everyone quotes — a huge context window — sounds like a technical footnote. In practice it is the whole game. It means Fable 5 can hold, in a single piece of work, more than you could brief a human in an afternoon: a year of sleep notes, your last three sets of labs, the two things your GP said, the supplement you actually take versus the one you meant to. A model that can only hold a paragraph makes you re-explain yourself every time. A model that can hold your whole context reasons across all of it at once.

The practical move is to stop treating each chat as a fresh start. Give it the full picture — who you are, what you are working on, what has already been tried — and let it answer questions in the context of everything else it now knows. The answer to ‘should I take magnesium’ is different when the model can see your sleep, your caffeine, and the fact that you already take three other things. That difference is the feature.

feature two: uploads — let it read your real data, not your memory of it

The single biggest jump in usefulness comes from feeding it your actual exports instead of your description of them. Fable 5 can read a PDF of your blood panel, a CSV of your wearable data, a photo of a supplement label, a screenshot of a confusing app dashboard. You are no longer asking it to reason about the health you remember — you are handing it the health you measured.

This is where the tool stops being generic. Anyone can ask a model about sleep. Only you can hand it your sleep. Upload the export, tell it what you are trying to understand, and ask it to find the patterns you would never spot by scrolling — the nights that followed late meals, the weeks your resting rate crept up, the correlation the app’s own summary quietly buried. The data was always yours. The reading is the part you were missing.

feature three: projects — give the job a room to live in

A project (some tools call it a space, or a custom workspace) is a persistent container: a set of instructions and files the model keeps in view every time you open it. Instead of re-briefing from scratch, you set it up once — here is my context, here is how I want you to talk to me, here are my labs — and every conversation inside that room starts already knowing the job.

For your health this is the difference between a hundred scattered chats and one continuous piece of work. Make a project called something plain — ‘my health, read honestly’ — drop your real context and data in, and write it a standing instruction: be specific, show me the trade-offs, tell me when the evidence is thin. Now every question you bring lands in a room that already has the file open. That is what a colleague feels like, and it is a setting, not a prompt.

feature four: scheduled actions — make it show up without you

The most underused feature is the one that runs on a clock. Fable 5 can be told to do a piece of work on a schedule — a weekly review, a monthly check against your own numbers, a nudge that reads your latest export and tells you what changed. You are not remembering to ask. It is remembering to look.

The move here is quiet and powerful: set one recurring job. ‘Every Sunday, look at this week against last week and tell me the one thing worth my attention.’ Most health tracking dies at week two because the checking is a chore. A scheduled action hands the chore to the worker and leaves you with the only part that needed a human — deciding what to do about what it found.

“A chatbot waits for you to ask. A worker with a memory, your data, a room to work in, and a clock — it just gets on with the reading, and hands you the decision.”

the line that keeps this safe

None of these features make the model your doctor. Memory can hold a mistake as faithfully as a fact. An upload can be misread. A scheduled review can flag the wrong thing with total confidence. Everything Fable 5 produces about your body is context to bring to a real clinician, not a verdict to act on alone. The features automate the reading, the sorting, and the remembering — the drudge work that was never the point. The judgement about what any of it means for you is the part that stays yours, and it is the part worth keeping.

That is the honest shape of the tour. The model did not just get cleverer. It got a memory, hands to hold your data, a room to work in, and a clock. Learn those four switches and you are not using a chatbot anymore — you are running a small, tireless research assistant on your own health, and you are still the one who decides.

what to do this week

Pick one feature you have never touched. Most people should start with uploads: take your most recent health export — labs, sleep, steps, anything you already have — and hand it over with four sentences of context about who you are and what you are trying to understand. Read the result the way you would read a sharp colleague’s draft: gratefully, and skeptically. Then, if it earns its keep, set one scheduled review so it keeps reading after you have closed the tab.