Meet Domestique - your AI training partner.
A domestique is the support rider on a cycling team - fetches bottles, sets pace, takes pulls for the leader. Your NeverChill domestique is an AI that knows your CTL, baselines, and recent load. It is in your corner. Humans always in the loop.
The 6am "should I rest today?" Slack to your coach.
You wake up, your legs feel cooked, the calendar says intervals. The decision can wait until your coach is online - or you can ask Domestique right now and get a useful answer rooted in your actual numbers. Your coach reads the thread later. Nobody wakes up at 6am for a triage question.
The positioning is explicit: complementary to a real coach, not a substitute. Domestique handles the "what should I do for the next hour" questions so humans can focus on the "what should I do for the next six months" ones.
Four things that make Domestique different.
It sees your data.
Your CTL, ATL, TSB, FTP history, recent sleep import, weight, last threshold block - Domestique has the same dashboard your real coach uses, with the same caveats about what is measured vs. estimated.
It reads your planned workouts.
When you ask "should I do this?", it knows what "this" is. The interval structure, the target wattages tied to your current FTP, what you did yesterday, what is on the calendar tomorrow.
It chats in context per workout.
Every planned session has its own thread. Ask about pacing, fueling, terrain choice, whether to bag it and ride easy. The conversation lives where the workout lives - not in a separate app.
Your coach can override.
If you have a real coach, they are in the same thread on the same workout - not cc-d on a transcript. They read what Domestique said and post the correction underneath it, in their own name. Only you can call Domestique into that thread, so your coach is never talking over a bot.
"I just paste my numbers into ChatGPT."
You can. You will be doing it again next session. Here is what changes when the model lives where your data lives.
It has your numbers.
ChatGPT will happily give you a workout for a hypothetical 250W FTP. Domestique reads the one you actually tested - plus CP and W-prime, max and resting HR, LTHR, and your current weight. Real numbers, off your own account.
It remembers.
The thread lives on the session, not in a chat window. Every message you have ever exchanged about that workout goes back into the prompt, along with the ride you linked to it - down to the per-lap numbers. Nothing resets.
It links to the actual activity.
Replies cite the ride, not a vibe. Link the activity to the session and Domestique has the whole thing - normalized power, IF, TSS, decoupling, and every lap - so "you held 108% on lap 3 and paid for it on lap 5" is a reading, not a guess.
It says I do not know.
Tuned to flag uncertainty rather than confabulate. If your data is sparse, it asks instead of inventing a TSS curve.
Honest answers.
Why do you call it Domestique? +
A domestique is the support rider on a pro cycling team - fetches bottles, sets pace, takes pulls into the wind for the team leader. They do the unglamorous work that lets the leader save legs for the moment that matters. That is the right metaphor for what an AI training partner should do. Domestique handles the "should I rest today?" triage, the pacing math, the quick check on your CTL - so you (and your real coach) can spend energy on the decisions that matter.
Does Domestique replace my coach? +
No. It is positioned as a faster first line - the "should I rest today?" Slack message at 6am that does not need to wake your coach up. Coaches who use NeverChill see Domestique as triage that lets them focus on the high-leverage work: race strategy, periodization, the conversations that actually need a human. Domestique fetches the bottles. Your coach picks the race.
What model powers it? +
A current-generation frontier model from a major provider, with a system prompt and tool layer specific to endurance training. We swap models as better ones ship. The training-load math, FTP calculations, and TSS computations are done by our backend, not by the model.
How does Domestique use my data? +
Real-time inference goes through Anthropic with zero-retention terms - they do not train on it. Our own models (route classifiers, climb categorizers, Domestique fine-tunes) DO train on athlete data. That is how Domestique learns from how strong athletes actually train. You can opt out of training in Settings → Privacy.
Can I turn it off entirely? +
Yes. Settings → Domestique → Off. The product works without it. No degraded experience, no nags. If your coach prefers you not use it, your coach can disable it for your account from the Coach Dashboard.
Will it tell me to overtrain? +
The opposite, in practice. The default tuning errs toward recovery when TSB is deep negative and when sleep or HRV imports flag stress. Domestique is more likely to talk you out of a session than into one.
Faster decisions. Same coach.
Create a free account. Domestique reads your real numbers from the first session you schedule, on every plan including the free one. Your coach still picks the race.