Frequent questions

The plan is a living thing: it reacts to what actually happened. A missed session, an unexpectedly hard week or a red day’s light is reason enough to reshuffle the days ahead.

How that happens matters. Ordinary changes arrive as a suggestion with ‘accept’ and ‘decline’ buttons — the plan is never rewritten silently. Only hard safety rules apply themselves, and the platform always says so plainly: what fired, and why.

Nothing. Making up a missed session is unnecessary and harmful: the week was built as a whole, and a session squeezed in on top of it wrecks your recovery.

Tap “Not today” on the session card — no reason asked. A receipt comes back: how many planned load points were not done and that there is no need to add them to the next sessions, with the projected fitness by the next key day. For a key session the platform looks for a slot within the next three days and offers a move in one tap; an easy one stays missed, and the right answer is simply to carry on with the plan.

There is a ceiling on how fast load can grow in a week, and it depends on your experience. It is a guardrail, not a prediction of injury: little here is proven, and the ceiling was deliberately set below the one risk boundary that has actually been observed — an increase in running volume of more than 30 % in a week.

After a break the same rule applies — you will not ‘catch up’ what you missed in a single week, and that is a deliberate choice in your favour.

Because endurance is built by the volume of easy work, and hard work is needed in measured doses. The guide is around 80–90 % easy time in the base phase.

If the easy sessions feel far too easy to you, the likely reason is an understated threshold: the zone boundaries are measured out from it, and too low a number makes your target pace slower than it should be. Check your threshold in your settings.

Gains happen during recovery, not during the work itself. A deload week is not a pause in your preparation but part of it: volume comes down, while the number of sessions mostly stays.

Feeling good at the end of a loading block is perfectly ordinary: fatigue builds up faster than it is felt.

Yes, but a different one. Without a main race the platform does not build a taper and keeps you in maintenance and general preparation.

As soon as you add a race and its date, the plan rebuilds around it: a preparation phase and a taper appear.

Bricks only go where you genuinely have the time on that day and where they will not wreck your recovery. If there is no suitable window in your schedule, or the week is a recovery one, no brick appears.

One tap marks today's session as missed — no reason asked, nothing moved. What comes back is a receipt: how many planned load points were not done, that there is no need to add them to the next sessions, and the projected fitness by the next key day assuming the rest goes to plan.

If a recording of that session arrives later, the mark is lifted by itself. The written-off load is shown as a caption next to the week totals; the plan does not change.

Form is the difference between the fitness you have banked and your fresh fatigue. A minus means the load has not let go yet.

In the middle of a build block a minus is normal, not a worry: you are working, and fatigue is meant to accumulate. The plus arrives in time for the race, after the taper. Form is a planning tool, not a verdict on your health, and not a goal in itself.

Because the load score is computed from your threshold, and every service has a threshold of its own. The methods differ too: here a fallback ladder is at work — power, pace, heart rate, feel — and for each sport the most accurate available signal is chosen.

Comparing absolute numbers between services is pointless. What does make sense is the trend within one system.

From your own sessions: the platform looks for near-maximal efforts of different lengths — roughly 2–5, 8–12 and 15–20 minutes — and estimates your threshold from them. A test is not compulsory.

Every estimate carries a confidence. While data is thin, the threshold is marked as an estimate, load is not built up, and the platform may suggest a short field test. Until a threshold appears, the plan runs quite happily by feel.

Usually it is one of three things: your sessions did not arrive from an external service; some of them had no load computed (no power, no pace, no heart rate and no effort rating); or the week really was easier than the ones before it.

Check the cards for your recent sessions: if the load on one of them is empty, add an effort rating — and the score will appear.

No, and that is deliberate. The new value applies going forwards, and past records stay as they were.

Otherwise every refinement of your threshold would redraw the whole history of your form, and the chart would tell you nothing about what actually happened.

Internally the platform works with five zones and shows three — easy, moderate, hard. For deciding how to run today, nothing more is needed, and extra boundaries create false precision: they still rest on an estimate of your threshold, not on a laboratory measurement.

Because we have no honest way of computing one from amateur data yet. A result over a long distance depends on fuelling, the weather, the transitions, sleep and a dozen other things the platform does not know about.

Instead of a number you could not trust, we will show what we can stand behind: durability as a trend and goal readiness — that section is specified and approved, but it is not in the app yet. A numerical forecast will appear when we can answer for it.

Swimming is the hardest sport to compute automatically: heart rate measures poorly in water, and pace depends on technique as much as on effort. If there is little data about a swim, the safest thing is to add an effort rating — the score will come out fairer than one based on a doubtful pace.

The stroke count comes from the watch file, length by length, and is built only from the control block: freestyle without equipment, one pool, pace within a narrow band. With fewer than eight usable lengths, or a swim that arrived without a file (Strava and intervals.icu do not pass lengths on), there is no metric — honestly, rather than approximately.

You can type in one number after the pool — it is stored apart from the watch and no goal is set from it. Stroke count has no effect on load or readiness.

Grey means ‘not enough data yet’, not ‘all is well’. Your personal baseline is built from your own figures over several weeks, and until then there is nothing to compare against.

We make a point of never substituting green for grey: saying ‘there is no data’ is more honest than handing you a reassuring colour with nothing behind it. The fastest way to get a colour is to answer the morning check-in.

They are different things. The branded readiness scores from Garmin, Whoop and Oura are computed by closed formulas, and they cannot be compared with one another.

From devices we take only the raw measurements — heart rate variability, resting heart rate, sleep — and compute our own status from them, with its workings described here in full. None of these statuses is an assessment of health.

Mark the illness — the plan will stop on its own, and there will be nothing to make up afterwards.

With a temperature, chest pain, breathlessness at rest or a racing heart you must not train, and here the platform leaves no choice: it blocks the plan and asks you to see a doctor. The return to training follows a gentle, stepped route rather than picking up where you left off: at least five steps, each lasting a day at minimum, and the whole return never shorter than the symptoms lasted.

No, it is entirely voluntary. Without it the platform relies on the data from your devices, and if there is none of that either, it leaves the status grey rather than pretending to know how you are.

The check-in takes under ten seconds and noticeably improves the accuracy of the light, especially if you have no devices.

Soreness is usually symmetrical, turns up a day after unaccustomed work and eases once you warm up. Pain that is in one spot, sharp, worse when you move, or still there after several days does not look like ordinary soreness.

The platform cannot tell these apart for you. Mark an injury and the plan will rebuild and stop offering impact work; the pain itself is a matter for a doctor.

If you want to, you can keep a cycle diary and adjust your load gently around how you feel. We do not impose rigid periodisation by phase: the evidence for it is still weak, and we would rather not promise precision that does not exist.

These entries are health data — a special category: we process them only with your separate consent, store them in Russia and never pass them to a language model.

No. Silence from a sync is not a missed session. The platform will mark the day as ‘no data’ and will not go reshuffling your plan for nothing.

If the session did happen, you can always add it by hand — a sport, a duration and an effort rating are enough.

No. The platform does all the maths itself: the connections are a convenient way of bringing sessions in, not a condition of it working.

Without them you can log sessions by hand or upload files. If an external service ever becomes unavailable, your plan will not suffer for it.

Yes. The fallback ladder does the job: no power, and we compute from pace; no pace, and we use heart rate; none of that either, and we use your effort rating.

Accuracy is lower that way, and the platform does not hide it: estimates get a confidence label, and load will be built up more cautiously.

Yes, your history can be brought in from a file or from a connected service — the platform then sees your background straight away instead of starting from nothing.

It is not compulsory: without a history, the first couple of weeks run as a calibration by feel.

By hand: the sport, the duration and an effort rating on a scale from 0 to 10. That is enough for the session to get an honest load score and count towards your form.

By default, nothing. If you switch the caption on in your profile yourself, a short note is added to the activity description: what the plan was, how it went (on plan, part of plan or beyond plan), load as a percentage of planned, reps on target, fitness and days to your next race, plus a separate line with the load score in points. Before switching on you see samples for every outcome, undershoot and overshoot included. Nothing about health, readiness or the reasons behind plan changes goes into the caption; missed and unmatched sessions are never captioned.

On a day of a medical stop the caption is not sent. To others, the public page of a Strava activity shows only the sport, date, load estimate and a link to Strava — Strava's terms require it.

No. Medical and personal data are never passed to a language model — neither the raw data nor anything derived from it: not readiness, not illness, not your cycle, not pregnancy, not signs of energy deficiency.

Explanations on medical topics are built by a deterministic template, not by a model. Only anonymised training context goes to the assistant.

No. The platform plans training; it does not diagnose, does not prescribe treatment and does not assess your health. Fitness, fatigue and form are planning tools.

Where warning signs appear, the platform does one thing: it stops the load and asks you to see a doctor.

The rules of the methodology are described openly, and each one has an evidence tier and a source. Robust findings are wired in hard; contested ones are made adjustable and marked as such; weak and individual ones stay an option you can choose.

In doubt the platform leans towards doing less — because in a model with no coach beside you, the cost of erring towards ‘more’ is out of all proportion.

In three ways. First, the plan is built and adapted for you — you do not have to be your own coach. Second, there is a safety circuit: illness, warning signs and growth that is too fast all stop the load. Third, every decision is explainable — you can always ask ‘why this?’ and get an answer in numbers rather than general advice.

The familiar metrics stay where they are: the load scale means the same thing as the one you are used to.

Because thin data is no reason to push. When confidence is low, the platform does not build load, does not show a forecast and leaves the light grey.

Showing a pretty number with nothing behind it would be easier, but then it could not be used for anything. We would rather give you an honest ‘calibrating’ status.

Every uploaded workout and everything the coach puts in the thread while you are away: a plan suggestion, a request to rate an effort, a receipt, the week review, the morning check-in, an updated threshold, team and follower events. The title of a notification names the reason — “There’s a plan suggestion”, for a workout a short encouraging line — and the text gives the gist. It never carries medical states, readiness or the reasons behind a suggestion: it sits on the lock screen where others can see it. The text is encrypted all the way to your device. Arrives as soon as the reason appears (at night, before 07:30, in the morning), and never on a day when the medical layer has stopped the plan.

You can switch it off in your profile; on iPhone, notifications work only for the app added to the Home Screen.

Yes: while the field in the coach thread is empty, the microphone takes the place of "Send". Hold it, say what you need, and release — the text appears in the field and the button turns into "Send". You send it yourself, like any other message: transcription makes mistakes, and it is worth seeing the line before it goes out. One recording is up to 30 seconds.

The recording is turned into text by the Yandex Cloud speech recognition service — the same provider the coach uses, with data staying in Russia. TriCoo does not store the recording and does not log the transcript; the request is sent with data retention switched off at the provider. Nobody is identified by voice. The recording goes out whole, before any reading of its content, so it is better not to speak about how you feel into the microphone — the coach does not discuss that anyway. The keyboard remains an equal way in.