Built an AI coach for triathletes that reaches out to you - not the other way around

Hey, thanks for your message!

All those changes are incremental improvements or better edge case handling based on user feedback.
Every improvement you see came directly from user feedback. Paying users get a product that gets better every week/day and they can actively shape. (That’s a feature for some people, not everyone.)

I wont say we are in beta, we are just actively working on improving the product day by day and trying to ship fast to meet everyones needs - the product is ofc fully working :slight_smile:
Feel free to read the testimonials on our website.

We dont have those super long development cycles a big company has, thats why we can move quickly and even adress small bugs super fast that would probably just get buried elsewhere.

We offer a 7day (up to 30d) free trial so you can test everything on your own before committing - no pressure at all.

2 Likes

iam rather impressed with that. but I will still wait a bit to give it a try until the product is bit more mature. Long term wise I think there is a rather good potential for your product and I like the way you handle this project.

1 Like

really appreciate it!!

1 Like

This was my early impression based on reading the thread, but to be completely fair I’ve tried (and paid for) two other AI coaching platforms and they were “equally beta” in my opinion.

Seems like the market standard these days. I’ll probably try this one free for 7 days after Walchsee (if a free trial is available) and if I don’t hate it, I’ll pay to give it a fair shot.

2 Likes

The 30-day trial is there exactly for that reason :slight_smile: just shoot me a DM after Walchsee and I´ll set it up for you! no pressure to commit before you’ve actually trained with it.

Hope Walchsee goes well!!!

1 Like

Thanks - but if it does, I might decide natural intelligence has worked too well to try AI :rofl:

1 Like

What’s new

A sharper coach.
Chat now runs on our most capable model, so guidance is more precise, the math is exact, and replies come back faster.

Your workouts, exactly as intended.
Push watt-based targets (Stryd, Garmin running power) straight to your watch, with every interval’s pace and power laid out in the workout description so you can review it at a glance.
And your workout files are always one tap away, with a shareable link for anyone training without the app open.

Numbers you can trust.
Apple Health athletes now get a fully computed Training Readiness and fitness profile, and your heart-rate zones stay anchored to plausible thresholds.

Effortless nutrition.
You can refine or remove any entry directly from your dashboard.

More on the way. Happy training. :slightly_smiling_face:

1 Like

Here’s a thing I ran on anonymized data I always wanted to answer myself.

Question: how many days fully off before you actually lose fitness?

TL;DR: a missed week is basically free, two weeks is a few percent you’ll get back quickly. Full blog article: We Measured 35,735 Sessions: How Long You Can Take Off Before You Lose Fitness | athletedata.health

I didn’t want to use CTL for this, because CTL dropping during a break is just the impulse-response model decaying on its time constant whether or not your physiology changed. So I used aerobic efficiency instead: NP/HR for rides, speed/HR for runs, on sessions >=20 min, IF < 1.05 where available. Higher = more output at the same HR.

Method: 35,735 sessions across 406 athletes. Each session z-scored against that athlete’s own efficiency baseline (>=15 sessions per athlete+sport), so we’re comparing each person to themselves, not pooling a Cat 2 with a beginner.
Then sorted by gap = days since the previous training day (any sport), looking at the session that ended the gap. 1 SD of efficiency works out to ~9% within-athlete, which is how I translated the z-scores below.

Comeback efficiency vs the athlete’s own normal, by days off:

1-2 days: ~0% (n=32,758)

3-5 days: -0.7% (n=2,438)

6-9 days: -2.0% (n=396)

10-14 days: -3.2% (n=69)

15-21 days: -5.1% (n=39)

Within-athlete paired check (the part I actually trust): 26 athletes had both trained-through sessions and at least one comeback after a 10+ day gap. 85% came back below their own baseline, median ~-0.57 SD (~5%). So it’s not composition.

Lines up with the detraining literature (Bosquet review, Sousa et al.): first clear VO2max drops ~10-14 days, ~4% by two weeks, 4-7% by three. Our ~5% by 2-3 weeks is right in that band.

Caveats:

  • First session back is usually easy on purpose. Easy = lower HR = higher efficiency, so this biases toward UNDER-stating the loss. The real drop is probably a touch steeper.
  • Long gaps sometimes mean illness/injury, which is its own hit, not pure detraining.
  • Submaximal efficiency, not a VO2max or FTP test. It’s a proxy, just a continuously-available one.
2 Likes

At end of a training block HR is surpressed and when you are fresh, HR is higher. EF is a poor proxy for evaluating fitness changes within a short period of time, it is useful over seasons but not over weeks. May be better to look at aerobic decoupling rate within sessions to show lost fitness.

1 Like

What’s new (14th-17th June)

While we working on bigger updates behind the scenes (will probably do a post about our roadmap soon), here are our tweaks and things we fixed and shipped anyways.
As you know, we are very close to every user and listening to every little bit of feedback and bug reports fixing stuff the same day.

One run, counted once.
Sometimes the coach multiplied milage when several apps log the same session - thats now fixed.
Ride durations come from your real pace, weather is read for the exact window you trained in, and last night’s HRV and recovery are always current.

A coach with conviction.
It holds a prescription and explains its reasoning when you push back and better respects the constraints you’ve set.

Straight to your watch.
Since workouts on athletedata are using on dynamics HR zones based off your recovery metrics, those HR zones can now also be used when workouts get pushed to garmin/coros (instead of using garmin/coros zones) → you need to opt in for that.

  • ycling plans bound for a Wahoo unit now arrive complete.

MCP updates
Full strength detail from Hevy, second-by-second heart rate on request, and claude (and the coach) can now write notes onto your Strava activities.

Bigger things on the way. Happy training. :slightly_smiling_face:

athletedata - what’s new

Your whole training week now lives on your watch, not just the miles.

Strength, exactly as planned.
When your coach programs a lifting session, it now lands on your Garmin with every exercise, set, rep and target load. Walk up to the rack, glance at your watch, and the session is right there - the lifts, the weight, the rest between sets. Your runs, rides and strength work all sync to the same place now. Your coach builds the session; your watch carries it.

More on the way. Happy training. :slightly_smiling_face:

1 Like

This sounds all very exciting, Julian. Good luck with it. I am currently just trying to stay fit enough until September when bigger training kicks in again (new born baby says hi). Can you confirm that you don’t need really any training plan but can just define race goals somewhat and your basic training week (I tend to do the same intervals / thresholds again and it would be nice to see when things are no longer paying dividends). I have found that my limitation with trainingpeaks where I could not really take any benefit as I don’t follow a training plan. Hope your time in SF is going well!

1 Like

Thanks Shonak, and congrats on the baby!!!
To confirm: no, you don’t need to build a plan. You connect whatever you already use (Strava, intervals.icu, your watch, a recovery device if you have one, your calendar) and it learns your normal week from the actual data. You can give it a loose race goal to aim at if you want, but it’s optional. Plenty of people use it with no race on the calendar at all.

Regarding “when are things no longer paying dividends”: It’s watching your trends across blocks (pace or power at a given heart rate, how your fitness and fatigue are tracking, HRV and resting HR if you wear something), so when an interval block has plateaued or you’re digging a hole it’ll actually tell you instead of letting you grind the same session for another month.

yep, TP assumes you’ll build a plan and follow it. This is the opposite: it reacts to what you do, not what you said you’d do. Sounds like that’s the bit that was missing for you.

Happy to get you set up if you want to try it.

1 Like

Hi Julian,

ich überlege, das mal auszuprobieren. Ich wohnen in Hamburg und trainiere seit 12 Jahren ohne Coach. Zurzeit nutze ich Stamina für meine Trainingspläne.

Mein Ziel ist unter 10 Stunden beim Ironman Hamburg 2027 (mein fünfter Ironman). Ist es möglich mich anzumelden und es zu probieren? Ich arbeite zurzeit mit Limiter. Ich musste mein FTP um 20w erhöhen und mein Laufschwelle um rund 20s/km erhöhen um bei 9:42 Zielzeitzu sein. Wie geht AI damit um in der Planerstellung? Können die Pläne geändert werden, wenn zuviel ohne zu wenig geplant ist? Berücksichtigt es meine Schwellenwerte?

VG

Stuart

1 Like

Hallo Stuart, danke dir, sub-10 ist ein starkes Ziel.

Ja zu allem: Du hinterlegst FTP und Lauf-Schwelle (oder es wird aus deinen connections importiert), daraus kommen die Vorgaben, und sie laufen mit, wenn sich die Werte ändern. Einen fertigen Plan brauchst du nicht. Es nimmt dein Ziel (9:42) plus deine aktuellen Werte und sagt dir, ob die +20w und +20s/km realistisch sind und ob du auf Kurs bist. Ändern kannst du jederzeit, einfach schreiben was ansteht und der Plan passt sich an - auch proaktiv wenn die recovery nicht passt etc.

Kannst dich gerne jederzeit bei mir melden solltest du fragen haben!

Cool, danke für deine Antwort. Ich habe ein Konto erstellt und ein Call für morgen gebucht :+1:t2:

Viele Grüsse

Stuart

1 Like

Top, freue mich drauf!!

Hi Julian,

I will switch to english on here for my feedback and exchanges. So i‘m quite impressed. I linked my data & the whatsapp feedback pushed back a bit on some of the sessions I was asking for. I can see its aligning the sessions well with the goal I gave it for Ironman Hamburg. It also confirmed that my goal is realistic based on my past data. Its very intuitive to use. It also asked some questions based on my Garmin data around why HR has been higher the last days. I confirmed that we have hot weather and got a recommendation how to adjust for that.

I signed up for the annual subscription and I will definately stick with it!

In addition to the existing apps, I also use Stryd for running and Form for swimming. If you receive enough requests from others, maybe these could be future integrations to get more precise feedback on those sessions :+1:t2:

Well done on this. I think its brilliant

3 Likes

Thanks so much, Stuart, this genuinely made my day.
That’s exactly how it’s meant to work: pushing back on the sessions you asked for, sanity-checking the Hamburg goal against your actual data, and flagging the HR drift so you could tell it about the heat. Really happy it clicked for you.

And thank you for going annual. I’ll keep making sure it earns it.

Direct Stryd/Form integrations are on the list, the more people ask the more we can push them to integrate with us ^^

Appreciate you taking the time to write all this up!!!

2 Likes

We just shipped a full rebuild of the first-run experience. The old version dropped you into a dashboard and assumed you’d figure out what to connect and why.

The new flow does one thing: gets you from signup to a first real insight off your own data, one step at a time, and shows you what each connection actually unlocks before you make it.

Here’s the ask:
We learn the most by sitting with someone the very first time they use it, so we’re booking short onboarding calls. We get you set up together, you talk out loud about what’s confusing, we shut up and watch.
Nothing to prepare. Do that and your trial runs a full month instead of 7 days.

Book a slot: cal.eu/team/athletedata/onboarding-videocall

If you’d rather just poke at it cold and tell me where it’s clunky, that’s honestly just as useful.
Reply here or DM.

Thanks!!

1 Like