Tuwa // the sports-science back room
Bring your plan.Or build it withyour own AI.
Tuwa is the sports-science back room for self-coached athletes. Import the plan you train from, or let your own Claude or ChatGPT write one for Tuwa to read. Each morning Tuwa reads your body — HRV, sleep, resting heart rate, training history — against that plan and states the day: its dose in words, and a go, modify or hold suggestion you decide on. It never writes your program.
Built for athletes who train sport skill and strength in parallel
iOS 17+ · iPhone From install to your own AI, in three steps
- 1
Get Tuwa
Install the app, connect Apple Health if you wear a watch or ring, and make your account.
- 2
Bring your plan
Paste it, add a PDF or a photo, or build it with your own AI. Tuwa never writes the program.
- 3
Add Tuwa to your AI
Turn on the connector in the app, then add one URL to your AI. It reads composite scores, never an Apple Health number.
Bring your plan, or build it with your own AI
Tuwa reads the program you already train from and adjusts each day from your body. It never writes the program.
-
I have a plan
Paste it, or bring a PDF or a photo of the whiteboard. Tuwa reads the weeks, days and sets; you check what it read, and week 1 starts next Monday or today.
-
I don't have one yet
Copy one prompt into your own Claude or ChatGPT. It reads Tuwa's plan-writing skill, asks about your training, and writes the first phase in Tuwa's import format for you to paste back.
Then the program overview: your weeks and phases on one grid, with the lifts of each week.
Today states the day
Each morning the day card names the kind of day and its dose in words — Volume, Effort and Length — against your plan. The numbers live in the brief, lift by lift, with the reason. You choose, then you start.
-
The day card
The kind of day and its dose in words: Volume, Effort and Length, against your plan.
-
The brief
Your readings, the day's suggestion and its reason, and each lift planned against today. The adjustment and Keep my plan sit side by side; you choose.
-
The load trend
One fatigue budget across sport skill, strength and conditioning, read against your own four bands.
No chat, no interpretation homework. Your readings, a suggestion with its reason, and the session you meant to do.
The whole day on one card
On Log, the whole program day sits on one card: its sections, cues and the day's rule. Start, and the session runs right in the card — tap a move, log the set, finish.
- run it on the card
- or type it in words
Training off the plan? Type the session the way you would say it; Tuwa turns it into a draft you check before it saves. A 1,328-exercise movement bank sits behind a search-first picker.
drills with no weight · one tap logs them as planned, a second tap undoes it
Trends you can read
Fatigue split into its parts, load against your own four bands, and each body signal on its own screen against your own baseline.
Ask your own AI about your training
Turn on the AI connector, and your own Claude or ChatGPT can read your Tuwa readings, sessions and plan, show the day's card in the chat, and send you a plan draft to review. Tuwa still makes the day's decision in the app.
How the connector works Add Tuwa to Claude, ChatGPT or a coding agent, step by step
Your baselines, drawn daily
Overnight HRV, sleep, and resting heart rate are scored against your own rolling baselines — not population norms — and compressed into one number with plain-language reasons. You don't get a dashboard to interpret. You get a score and the why.
every metric is scored against your own rolling baseline, not population norms
Here is what we believe,
and what would prove us wrong
Tuwa scores your sleep against a fixed target today. We are building one that adapts to the state you went to bed in. It is not finished — so here is the whole design, the weights, and the parts we are least sure of.
The target is yours, and you can see it
Your sleep target is learned from your own nights, not taken from a population average. And the weights behind the score are printed below — not a trade secret.
The score knows what the night had to do
Seven hours after an eighteen-hour day, after a match, and after a quiet Tuesday are three different nights. Same maths each time — the parts are just weighted differently.
It never leaves the phone
Your raw sleep, HRV and heart-rate readings stay in HealthKit on your device. Only the finished scores sync. This one is not a plan — it is how the app already works.
One set of weights, nudged — not a second algorithm
Every morning the app looks at what it already knows about yesterday — how long you were up, how hard you trained, how steady your sleep has been. That picks out which situations apply, and each one nudges the weights below. Every night stores what it saw and what it decided, so any score can be taken apart later.
simple arithmetic on your phone · no network · no new sensor
hours slept carries half the score — it is the best-evidenced part · deep and rem count for little, because a wrist only half-agrees with a sleep lab
we argued these numbers out — we did not fit them to data · logged as h-01
What changes, and how sure we are
| situation | when it applies | what changes | how sure we are |
|---|---|---|---|
| You were up too long | You had been awake far longer than usual before going to bed. | Hours and deep sleep matter more. Your target goes up, by 45 minutes at most. | backed · the size is our guess (h-02)Sleep pressure builds the longer you are awake — that much is settled. How much extra sleep it earns is our number, not the research's. |
| You trained hard | Yesterday was well above your normal for the last month. | Hours and deep sleep matter slightly more. Your target goes up by half an hour at most. | a guess (h-03)The only study we found that tested this directly found nothing. We kept it, wrote down what would make us delete it, and said so here. |
| Your rhythm just broke | You went to bed more than two hours off your usual time — the flight, the late game. | Being off-schedule counts for less, staying asleep counts for more, and deep and REM are trusted only half as much. | backed · the response is our guess (h-05)The first disrupted night really is the worst one. Easing off the schedule penalty is our answer to that, not a finding. |
situations only shift the weights — they never add a new ingredient to the score
Every guess, written down with the test that kills it
Anything in the score that no study backs goes on this list, next to the result that would make us change it. Nothing gets into the app without a line here — and if we cannot say what would prove a guess wrong, it does not go in at all.
| no. | what we are guessing | where it stands | what would change our mind |
|---|---|---|---|
| H-01 | The weights above are roughly right. | a guess · argued, not measured | Check each part against how people actually feel the next day, then re-fit once we have enough nights from enough athletes. |
| H-03 | A hard training day means you need up to half an hour more sleep. | a guess · the one test found nothing | If hard days and easy days show no difference, this comes out — from the situation and from the target. |
| H-07 | When you are well behind on sleep, hours matter more than sleep stages. | a guess · but well supported | Check whether deep and REM still tell us anything once you are in real deficit. |
| H-10 | Seeing your target move, and why, helps more than it worries you. | a guess · no study, just our call | Our own use, and what people tell us. If it reads as alarming, the whole layer goes quiet. |
Sleep pressure builds while you are awake. Borbély, Daan, Wirz-Justice & Deboer (2016), Journal of Sleep Research 25(2). We took the shape of it. We did not take any numbers.
Training shifts sleep, but only slightly. Kredlow et al. (2015), Journal of Behavioral Medicine 38. Small effects across 66 studies — which is why this one stays a guess.
Regularity is the middle of your night, not its start. Wittmann & Roenneberg (2006), Chronobiology International 23(1–2). That is the number we track.
Lost sleep adds up, and hours are what repay it. Van Dongen et al. (2003). The reason hours outrank sleep stages when you are behind.
The back room of a pro team — plan, physiology, and a decision — for athletes who coach themselves.
Your data stays on your phone
Tuwa reads HealthKit — it never writes to it — and raw health data never leaves the device. Only composite scores sync.
bring your plan, or build it with your own AI