# IntervalCoach > IntervalCoach is an AI training coach that understands your full picture: your fitness, recovery, goal event and the life around your training. It reads 60+ recovery signals (HRV, sleep, sleep debt, training load, subjective wellness) and adapts today's session when the data calls for it, rather than handing you a fixed plan. It works across every sport you train: cycling, running, trail, triathlon, swimming, strength and hiking, and supports athletes returning from injury. Trail running is a first-class sport — terrain-aware analysis (elevation, gradient, power-hike thresholds), trail-specific workout templates (Long Climb, Downhill Repeats, Power-Hike Intervals, Technical Trail Intervals, Ultra Time-on-Feet), and trail-only library filtering. Built by Martijn Russchen. Subscriptions start at €3/month after a 14-day free trial — no credit card required. Works with Apple Health, Whoop, Oura, Garmin, Polar, Wahoo, Intervals.icu, and Zwift. Available in 25 languages. --- ## Product overview ### What it is IntervalCoach is a daily-adaptation engine for endurance athletes. Before each session, the system reads more than 60 recovery and training-load signals from your devices — HRV, sleep, resting heart rate, fatigue, recent load, subjective wellness markers like RPE and soreness. It then either confirms today's plan, dials it back, or replaces it. No static templates, no guessing. Underneath the daily adjustment sits a real periodization roadmap. The engine supports five established models — linear, undulating, polarized, pyramidal, and block — and steers your training across cycling, road running, trail running, swimming, walking, hiking, and strength so long-term progression keeps moving even when individual days have to bend. The product covers cyclists, road and trail runners, triathletes, hikers, and athletes who want low-impact aerobic volume (returning from injury, recovery blocks), with workout structures appropriate to each sport (power-based for cycling, pace-and-HR for road running, terrain-aware effort-driven for trail running, drill-and-interval sets for swimming, duration-with-HR-cap for walking and hiking). Trail-specific workouts include Long Climb (sustained uphill threshold), Downhill Repeats (eccentric quad conditioning), Power-Hike Intervals, Technical Trail Intervals, and Ultra Time-on-Feet — plus terrain-aware race pace plans and post-workout analysis that talks vert, gradient, and ultra fueling rather than flat-equivalent pace. The positioning is simple: train for the day you're actually having, not the day you planned for. ### How it works There are three steps to onboarding. **First, connect Intervals.icu.** Linking your account syncs training data, power curves, and fitness metrics automatically; Intervals.icu pulls from Strava, Garmin, Wahoo, Coros, and other platforms. **Second, the AI analyzes your data.** The system reviews your CTL (fitness), ATL (fatigue), TSB (form), recent recovery and sleep, and any goal events on your calendar to understand your current training state. **Third, you get your workout.** A structured workout with intervals, targets, and duration optimized for today lands in your inbox and on your Intervals.icu calendar — ready for Zwift, Wahoo, Garmin, or outdoors. After that, a five-step adaptation pipeline runs every morning before the daily email goes out: a health check (sick or injured days), calendar priority (A/B/C events and external workouts), readiness assessment (60+ signals), workout reshaping (load, intensity, stimulus, warm-up, breathing block), and weekly TSS steering. Every adjustment is concrete — it changes the structured workout file your head unit syncs, not just a label in the email. ### For coaches, teams and clubs IntervalCoach sells coaching capacity by the seat. A coach, team or talent program buys seats on a Coach Plan: €10 per athlete per month up to four seats, €8 from five, €7 from fifteen, and individually priced from forty athletes (invoicing, multiple staff accounts). Yearly billing is two months free. One seat covers the athlete's entire subscription — daily-adapting plan, eight weeks scheduled ahead, Coach+ chat — so athletes on a roster pay nothing themselves. The coach tooling spans the whole working day: a roster that sorts itself by who needs attention and why, biometrics shown as deviations from each athlete's own 30-day baseline, acting straight from the roster (approve a proposed session change, push a rest day that clears and redistributes the week, set a standing instruction in plain language the planner follows), an approvals inbox that collects every pending change across the squad, an eight-week squad timeline with races, squad analytics (total load done against what was asked, per-athlete weekly compliance, squad fitness as a median with the min–max spread) and a Monday-morning squad digest by email at 07:00 local time. The launch post at https://www.intervalcoach.app/blog/coach-plan shows each surface. Coaches comparing software can read honest side-by-sides written for the coach rather than the athlete: TrainingPeaks (https://www.intervalcoach.app/en/coaches/vs-trainingpeaks), Final Surge (https://www.intervalcoach.app/en/coaches/vs-final-surge), Coachbox (https://www.intervalcoach.app/en/coaches/vs-coachbox), Vekta (https://www.intervalcoach.app/en/coaches/vs-vekta) and Athletica (https://www.intervalcoach.app/en/coaches/vs-athletica). They compare who pays, what a roster costs, AI plan adaptation, recovery data, Intervals.icu, athlete billing, messaging, template libraries and team accounts. They state where the other tool is the better choice: IntervalCoach has no athlete billing, no coach-athlete messaging, no template library or marketplace, no shared team account (each coach buys their own seats), and a three-seat minimum. On top of the athletes' plans, the coach gets the coaching layer: a roster sorted by who needs attention today (with the reason attached and a handled state), HRV / resting heart rate / sleep shown as deviation from each athlete's own 30-day baseline, four weeks of prescribed-versus-actual load per athlete, a diff of what changed since the coach last looked, and full or view-only access to work inside any athlete's account, including a shared Coach+ thread. Every adaptation is explained and every coach edit is attributed, so the coach stays visibly in charge. For coaches who already run their athletes on Intervals.icu there is a dedicated page at https://www.intervalcoach.app/en/coaches/intervals-icu. IntervalCoach runs on top of Intervals.icu rather than replacing it: each athlete keeps their own Intervals.icu account, planned workouts are written to their Intervals.icu calendar and sync to Zwift, Garmin, Wahoo and Polar from there, and sessions a coach already placed on the calendar are never double-booked. Intervals.icu itself is free, and its coaching groups need the $4/month Supporter plan; the Coach Plan seats are paid on top of that and include each athlete's full Max plan, so athletes pay IntervalCoach nothing. What it does not do: bill athletes on the coach's behalf, or provide direct coach-to-athlete messaging (Intervals.icu's own chat and comments keep working). The minimum is three seats. ### Pricing Three tiers. **Free** includes the fitness dashboard, three JustTrain workouts per month, and a weekly training plan with the current week of structured workouts on your calendar. **Pro** is €3/month or €30/year (save 17%) and adds daily adaptive workouts, four weeks of fully structured workouts planned ahead on your calendar, post-workout analysis, race pace plans, Race Recon (race course analysis from a GPX or FIT file), Whoop and Apple Health integration, nutrition advice, monthly progress reports, and 20 Coach+ chat messages per month. **Max** is €8/month or €80/year (save 17%) and includes everything in Pro plus eight weeks of structured workouts planned ahead (a full training block on your calendar), 200 Coach+ messages per month, top-up packs that never expire, photos and screenshots in Coach+, Performance Benchmarks (power and recovery against similar athletes), clubs without the member cap plus a weekly AI club digest, and an AI assistant connection (MCP) so Claude, ChatGPT or Cursor can use your training data and coaching tools. New users get a 14-day free trial of Pro with no credit card required. Cancel anytime; you keep paid features through the end of the billing period. Prices display in EUR and include VAT where applicable; Stripe converts to your local currency at checkout (USD, GBP, AUD, CAD, etc.). Apple Pay, Google Pay, and Alipay are supported. ### Race Recon Race Recon (https://www.intervalcoach.app/en/race-recon, announced in https://www.intervalcoach.app/blog/race-course-uploads) reads the course file of a race on the athlete's calendar. When creating or editing a race, the athlete attaches a GPX or FIT file of up to 40 MB, or imports an Intervals.icu route by its route ID; distance, total climbing and start location are filled in from the file. Every file is resampled to fixed 100 metre steps and its elevation smoothed before any gradient is derived, because GPS altitude noise otherwise produces impossible gradients; a climb must run at least 300 metres and gain at least 20 metres to count. The race then shows its elevation profile and every sustained climb with length, height gain and average gradient. Three things change. Training: IntervalCoach estimates how long this athlete will be on each climb at race pace, and interval sessions of that length move up the list while repetitions longer than the longest climb move down; it is a preference within the athlete's recovery, training phase and methodology, and the course never adds intensity. Preparation: the athlete's personal race page compares the course's demands with recent measured power or pace and training history, and shows development priorities and suggested training, with missing evidence shown as unknown. Race day: the course is split into climbs, descents and flats, each with an estimated time, a power target for rides or a pace for runs, and a carbohydrate and fluid budget; wind forecasts along the course direction and optional paved, gravel and unpaved surface analysis are included where available. Worked example from the launch post: a 151 km Harz route with 2,014 m of climbing, where 17 sustained climbs hold 1,702 m; for a rider at 246 W, distance alone predicts 4.8 hours and 434 g of carbohydrate, while the course predicts 6.5 hours and 585 g across 193 segments. The course file stays private to the athlete's account, with the route and elevation kept and activity data (heart rate, power, timestamps) stripped. Sharing the race to the community catalog makes the course public on its race page after moderation, with a GPX download, and any athlete who adds that race inherits the course. Race Recon is included in Pro and Max, and in the 14-day free trial. ### Activity and session pages IntervalCoach rebuilt how it shows a single session (announced in https://www.intervalcoach.app/blog/activity-and-session-pages). Activity, planned session and race pages are on every plan including free. A completed activity is shown in its context: the planned session it completed and the compliance with it, the week's load against its target and the training phase, that morning's readiness with HRV and resting heart rate against a 30-day baseline and sleep debt, why the session was chosen, and the weather with the wind split into headwind, crosswind and tailwind. It shows fitness, fatigue and form before and after, a recovery estimate with recovery nutrition, time in zone against the time the session asked for (zone times use the smoothing Intervals.icu applies), power per kilo, W′ used, work above FTP, strain, heart rate drift, the power curve, every channel on one trace, the interval table, run and walk splits and a route profile with climbs. When the FTP set in Intervals.icu and the eFTP differ by 8% or more, the page shows load, intensity and zones at both; on one 82-minute ride that was 56 TSS (IF 0.64) at 303 W against 85 TSS (IF 0.79) at 247 W. It does not decide between them. The coach's written analysis is Pro, Max and trial. A planned session page shows the steps with targets and cadence, its load, intensity, time at threshold and above, energy and carbohydrate per hour, why it was chosen, its effect on the week and on fitness, fatigue and form, warnings for a same-day clash and an FTP mismatch, the forecast across the session's hours with clothing advice, fuelling on the session's timeline, previous similar sessions, which platforms receive it, and .zwo, .erg, .mrc and .fit downloads. A race page adds a countdown, the course and fuelling; Pro and Max add the pacing plan, the athlete's own fade over long efforts and race week with carb loading. ### Integrations IntervalCoach reads recovery and training data from whichever sources you've connected. Through Intervals.icu (plus Apple & Google Health) it reaches a broad ecosystem, grouped roughly as: head units & watches (**Garmin · Wahoo · Polar · Coros · Suunto · Amazfit · Hammerhead**), indoor & virtual (**Zwift · MyWhoosh · Rouvy · Biketerra · Concept2**), apps & sync (**Strava · Dropbox**), and recovery & wellness (**Whoop · Oura · Apple Health · Google Fit · Huawei Health**). Apple Health acts as a hub for Apple Watch, Oura, Withings and Garmin Index scales, period trackers, and any other HealthKit-compatible device. Whoop direct brings in HRV, resting heart rate and sleep in real time each morning; they feed IntervalCoach's own readiness score. Workouts sync back through Intervals.icu to Zwift, MyWhoosh, Garmin, Wahoo, and TrainerRoad. ### Languages IntervalCoach is available in 25 languages: English, Dutch, German, Spanish, Italian, French, Polish, Portuguese, Czech, Slovenian, Danish, Swedish, Norwegian, Finnish, Hungarian, Greek, Romanian, Slovak, Latvian, Lithuanian, Estonian, Simplified Chinese, Traditional Chinese, Japanese, and Korean. The AI generates workout descriptions, emails, and coaching advice in your selected language. ### Clubs The hardest part of training isn't the hard workout — it's showing up for it alone. Clubs let you train alongside the friends who keep you honest. Create a club, invite your riding or running buddies, and you get a shared activity feed, fair leaderboards, and group events that land on everyone's calendar. Leaderboards rank on effort you control — consistency, weekly volume, and fitness growth (CTL) — not raw watts, so the strongest rider doesn't automatically win and everyone has a reason to keep showing up. Privacy is strict: only derived training metrics are ever shared with a club; raw recovery and wellness data never leaves your account. Joining a club is free on every tier. Pro members can create and lead clubs of up to 10 friends; Max members get unlimited club size plus a weekly AI digest that summarizes how the group trained. ### The platform The product is organized into four areas, each with a dedicated overview page under /platform. **AI Coaching** (https://www.intervalcoach.app/en/platform/ai-coaching) covers daily-adapting workouts, on-demand TrainNow sessions, the Coach+ conversational coach, race pace plans, Race Recon, and post-workout analysis. **Insights & Analytics** (https://www.intervalcoach.app/en/platform/insights) covers the live dashboard, deep analytics (CTL/ATL/TSB, eFTP, power curve, zone distribution), the periodized training plan, recovery profiling, and peer benchmarks. **Integrations** (https://www.intervalcoach.app/en/platform/integrations) covers native two-way sync with Intervals.icu, Whoop, Apple and Google Health, and Zwift. **Community** (https://www.intervalcoach.app/en/platform/community) covers clubs, leaderboards, the weekly AI club digest, the Coach Plan for human coaches, and referrals. The platform overview lives at https://www.intervalcoach.app/en/platform. IntervalCoach runs on the web and as native mobile apps: an iPhone app on the App Store and an Android app on Google Play (https://play.google.com/store/apps/details?id=app.intervalcoach.android), both with full feature parity with the web app. The iPhone app includes an Apple Watch app (https://www.intervalcoach.app/blog/apple-watch-app) with the athlete's readiness score, today's workout, health metrics and training score, watch face complications, and optional auto-scheduling that places planned workouts, with their full interval structure, into the Watch's own Workouts app. ### The founder I've been cycling seriously for over fifteen years. Road, gravel, the occasional race, and a lot of structured training in between. Like most data-driven athletes I tried just about everything: paid coaches, generic plan apps, hand-tuned spreadsheets, and Intervals.icu for the analysis side. The pattern was always the same. The plan looked great on Monday. By Wednesday I'd slept badly, gotten sick, traveled for work, or just felt off, and the plan didn't care. I'd either ignore it and feel guilty, or push through and pay for it later. None of the apps actually looked at how I felt that day before deciding what to prescribe. IntervalCoach started as the coach I wanted for myself: something that reads my recovery data every morning and adjusts the session before I see it. No begging an algorithm to skip a workout, no manual replanning. It's grown well beyond that now, but the core question is still the same — what should I actually do today, given the body I'm in today? I ship new things every week, mostly driven by what I run into in my own training and what other athletes ask for. I'm still figuring out where it's all going, but I'm enjoying finding out. — Martijn ### Ambassador program Athletes who bring IntervalCoach to a club, team or training group can apply to the ambassador program (https://www.intervalcoach.app/en/ambassadors). Ambassadors get Max free for as long as they are active, their club with the weekly AI digest switched on, their referral code and its reward, a spotlight post on the blog, a Club Ambassador badge on their public profile, and a private channel with the founder including early access to features. There is no follower threshold and no contract. Three requirements are checked automatically: an account at least 60 days old, a live Intervals.icu connection, and either a paid subscription or one referral that already converted. Every application is read by hand. Ambassadors agree to state that they are ambassadors when posting about IntervalCoach, as advertising disclosure rules require. --- ## Race calendar IntervalCoach publishes a browsable calendar of real upcoming races athletes are training for (https://www.intervalcoach.app/en/events), across cycling, running, trail running and triathlon, with dedicated pages per sport (e.g. https://www.intervalcoach.app/en/events/cycling) and per sport-and-country (e.g. https://www.intervalcoach.app/en/events/cycling/nl). The catalog is fed by community submissions, athletes' own planned races, and verified seeding, and every entry is web-verified before publication. Each race page answers the questions someone considering the race actually has. **What this race demands**: a coach's read of the course derived from its stored distance, elevation, sport, place and season — the defining physical demands, fueling guidance for long events — with every generated claim checked against the stored facts before publication. **What would it take you**: estimated finish-time ranges and typical peak-week training hours per athlete level (anchored in sustainable W/kg for cycling, marathon shape for running), computed from the course's own distance and climbing. **Your road to race day**: the Base–Build–Peak–Taper preparation a plan for that race date would have, mapped onto real calendar dates using the same periodization the product's plans use. When an athlete has shared a race's course, its page also shows the elevation profile and the sustained climbs, with a GPX download. From any race page, an athlete can start training for that race with one click; race organisers can submit their event and embed a "Train for this race" badge on their own site. --- ## Science and methodology ### The Science Behind IntervalCoach URL: https://www.intervalcoach.app/blog/the-science-behind-intervalcoach Every decision IntervalCoach makes is grounded in published sport-science methodology, not generic AI advice. The engine implements five distinct periodization models — Linear (Bompa & Haff, 2009), Undulating (Rhea et al., 2002), Block (Issurin, 2010), Polarized (Seiler, 2010), and Pyramidal (Esteve-Lanao et al., 2007) — each with its own mesocycle structure, intensity distribution targets, and weekly patterns. Athletes can pick a model or let the system recommend one based on training history, time until the goal event, and personal preference. When a goal event approaches, IntervalCoach applies an exponential taper model based on the Mujika & Padilla (2003) meta-analysis: 41–60% volume reduction over 8–14 days, intensity maintained, with three profiles (aggressive at 40% race-day volume, standard 55%, conservative 75%). Two safety metrics run continuously: ACWR (Acute:Chronic Workload Ratio, Gabbett 2016) triggers automatic load reduction when the 7-day-to-28-day load ratio exceeds 1.5; Training Monotony (Foster 1998) flags daily-load uniformity above 2.5 with awareness of taper context. Daily readiness layers IntervalCoach's own readiness score (HRV, resting heart rate and sleep against your baseline), TSB via the Banister (1991) impulse-response model, HRV with coefficient-of-variation tracking, sleep debt, SpO2, skin temperature, ACWR, consecutive training days, subjective wellness, RPE feedback, and aerobic durability. ### Five Periodization Models Explained URL: https://www.intervalcoach.app/blog/periodization-models-explained Periodization is how training stress is organized over time. The model determines how hard weeks get, when recovery happens, and how intensity is distributed across easy, moderate, and hard sessions. **Linear** — 3–5% weekly TSS increases with a recovery week every fourth, intensity roughly 75% easy / 10% moderate / 15% hard; best for athletes new to structured training or returning after a break. **Undulating** — wave loading varies emphasis week to week (volume, threshold, VO2max), around 70/12/18; suits intermediate athletes who get bored with routine. **Block** — three-week concentrated blocks targeting one energy system followed by deep recovery, roughly 65/10/25 in loading weeks; advanced athletes with strong recovery capacity, especially with short prep windows. **Polarized** — 80% low intensity and 20% high intensity with almost nothing in the middle (Seiler); well-trained endurance athletes prone to overtraining or who struggle to keep easy days easy. **Pyramidal** — 75/15/10 with deliberate moderate work (sweet spot, tempo); athletes who respond to sustained efforts and have 10–15 hours/week of training time. Practical guidance: new athletes start linear; plateaued athletes try undulating or switch between polarized and pyramidal; short windows favor block; "always tired on hard days" suggests moving toward polarized. ### How to Use CTL, ATL, and TSB URL: https://www.intervalcoach.app/blog/ctl-atl-tsb-explained CTL (Chronic Training Load) is fitness — a rolling 42-day average of daily training stress; recreational cyclists sit at 40–50, competitive amateurs 70–90, pros above 120. ATL (Acute Training Load) is fatigue — the same metric on a 7-day window; it spikes after big weekends and drops fast with rest. TSB (Training Stress Balance) is form — simply CTL minus ATL; positive means fresher than your fitness suggests, negative means carrying fatigue. Most athletes race best at TSB +10 to +25. Common mistakes the post calls out: **chasing CTL like a high score** — CTL should rise gradually (3–5 points per week), faster ramps invite illness or injury; **living in deep negative TSB** — a week or two below -20 in a build block is normal, a month there is overtraining; **not tapering enough** — most age-group athletes show up at TSB -5 when they should be at +15 because they're afraid of losing fitness, but CTL won't collapse in 10 days; **ignoring context** — TSB +10 means different things at CTL 90 vs CTL 40, and external stress, sleep, and nutrition all change how a given TSB feels. The numbers are a guide, not gospel. ### HRV-CV: What Volatility Tells You That HRV Alone Does Not URL: https://www.intervalcoach.app/blog/hrv-cv-autonomic-variability A single morning HRV reading is noisy: a bad night, a late dinner, a glass of wine all push it down without meaning much. The more useful signal is **HRV-CV** — the coefficient of variation of HRV over a rolling window — which captures how stable your morning HRV has been relative to its typical value. A well-recovered athlete has tight clustering around their personal mean; when recovery isn't keeping up with training or lifestyle demands, day-to-day swings widen even before the mean drops. This catches a failure mode that absolute HRV misses: an athlete whose baseline has drifted lower over weeks of stress, hard training, or poor sleep looks "fine" against the new normal, but rising HRV-CV flags that something has changed. A 2026 paper in the American Journal of Physiology – Heart and Circulatory Physiology analysed roughly two million sleep-derived HRV nights from over 21,000 wearable users and found higher HRV-CV associated with more alcohol, shorter and less consistent sleep, and lower physical activity — and for alcohol and sleep, the association was stronger with HRV-CV than with absolute HRV. About five nights of data per week is enough for a reliable weekly value. IntervalCoach computes HRV-CV from whichever source is connected (Whoop direct or Intervals.icu wellness entries) and compares against your personal baseline. Three states: stable (no action), elevated (a yellow flag that combined with other signals can trigger load reduction or intensity cap), and the rare compressed state that can indicate parasympathetic saturation in deep fatigue. ### Tired, or Getting Sick? Catching Illness Before You Feel It URL: https://www.intervalcoach.app/blog/tired-or-getting-sick A hard training session and an early infection produce the same autonomic picture: both suppress overnight HRV and raise resting heart rate. That overlap is real physiology, not sensor noise, which means any system that flags illness from HRV alone is really just flagging "you trained hard" and will cry wolf on your best sessions. The signal that separates the two is the set of metrics training does not move but infection does: overnight **respiratory rate**, **blood oxygen (SpO2)**, and **skin temperature**. A brutal threshold day leaves all three flat; a brewing illness nudges them off baseline days before you notice symptoms. IntervalCoach detects illness only from these orthogonal signals (HRV and RHR corroborate but never trigger), and only when at least two of them deviate together, so a single noisy night is never enough. When the pattern appears, the day is automatically eased toward recovery instead of the planned intervals, and the dashboard and daily briefing name the exact signals that fired so the recommendation is explainable and overridable. The same detection protects the recovery model: illness days are excluded from how IntervalCoach learns each athlete's recovery time, so a cold never teaches the system that an easy ride wrecked you. The honest limit is deliberate: it requires a wearable that records those overnight metrics (Whoop, Apple Watch, Oura), and with only HRV and resting heart rate available it stays silent rather than mislabel a hard day as illness — precision over false alarms. ### Can You See Illness Coming? What the Overnight Data Shows URL: https://www.intervalcoach.app/blog/can-you-see-illness-coming An observational read of the sick days IntervalCoach athletes log on their calendar, compared night by night with each athlete's own overnight normal. Wearable studies of confirmed infections (Mishra et al. 2020 on COVID-19 and smartwatch resting heart rate; Radin et al. 2020 on Fitbit data and influenza-like illness) found resting heart rate rising before symptoms. Across athlete-logged sick days, which mix colds, stomach bugs and flu, the week before barely moves on average: resting heart rate edges up only in the last night or two, too little to act on alone because training moves it too. Skin temperature rose in the two nights before illness, as the research predicts (Smarr et al. 2020), but too few wearables report it to confirm. The clear signal comes after the illness: overnight resting heart rate and HRV are most disrupted around four days after the last sick day, HRV is back after roughly ten days and resting heart rate is still slightly raised in the second week. The longer the illness, the deeper the disruption: in the first week back resting heart rate sits more than three times further from normal after a week or more ill than after a day or two, and after a week or more it is still clearly raised in the second week. IntervalCoach therefore spots illness mainly from skin temperature, breathing rate and blood oxygen (a sustained resting heart rate rise only corroborates), and paces the return to training over at least five days, longer after a longer illness, holding the easing until resting heart rate and HRV are back at the athlete's own normal. Limits: observational, not a medical study; logged sick days are self-reported and mixed in cause. ### Do Fitter Athletes Recover Faster? What Our Recovery Curves Show URL: https://www.intervalcoach.app/blog/do-fitter-athletes-recover-faster Sports science finds that cardiac autonomic recovery completes sooner in fitter people, and takes up to a day after low-intensity work, one to two days after threshold work and at least two days after high-intensity work (Stanley, Peake and Buchheit, Sports Medicine, 2013). IntervalCoach's recovery curves, fitted from each athlete's full history of overnight HRV, resting heart rate and sleep across thousands of athletes, show where the faster return comes from: fitness shrinks the hit. Pooled over hard session types, the overnight hit is roughly 40% smaller in the fittest athletes than in the least fit and falls at every step up in fitness, while the time it takes to clear, compared among athletes whose data supports a personal rate, shows no trend with fitness. Fitter athletes are back sooner because the dip is smaller. Limits: it compares different athletes at different fitness rather than one athlete over time, the spread within each fitness level is wide (a population trend, a weak predictor for one person), and it makes no claims about sex or menstrual-cycle differences. In practice: fitness buys load tolerance, and spacing between hard days should follow the type of session. ### How Many Mornings Does It Take to Know How You Recover? URL: https://www.intervalcoach.app/blog/your-recovery-curve-is-now-your-own A recovery curve models how a session disturbs your overnight signals and how fast they return. It has two halves that behave very differently: **depth** (how hard a given session type hits you) and **rate** (how fast you clear it). IntervalCoach tested both against mornings the model had never seen, splitting each athlete's history by time and scoring on the later part, and published what it found rather than only the direction. The old estimator learned only from mornings with no training in between, which are rare: 60% of second mornings and 85% of third mornings already carry a new session. Across 28,041 stored curves that left **82.7% on the population default for the recovery rate**, with the median fitted half-life sitting exactly on that default. A dose exponent of 0.6 that had been in the model since the beginning measured 0.04 with a confidence interval including zero, meaning session size did not scale recovery inside a session type except for long endurance rides. The rebuild fits the athlete's whole morning series at once under superposition, which the model already assumed when predicting but refused when learning, so every morning counts and an athlete contributes roughly 115 instead of a handful. **Depth is now personal for 100% of athletes** (previously ~58%) and **rate for about 27%** (previously 17%), with the population rate used and named as such when an athlete's own data does not settle on an answer. On two independent samples of ~120 athletes the new curve is more accurate for about six in ten, cutting median error roughly 2% against the previous model and 5% against the population default. The honest limit is published alongside: a control that assumes training has no effect on tomorrow's morning is still closer than any of these models, and none of them orders an athlete's own mornings better than chance by more than a whisper (rank correlation ~0.05). That is why the recovery curve is one input among several into readiness, alongside HRV, resting heart rate, sleep and the morning check-in, and why it never gets its own training rule or overrides how an athlete says they feel. ### Sessions Pitched To Your Own Level, One Energy System At A Time URL: https://www.intervalcoach.app/blog/sessions-pitched-to-your-own-level Most plans know what phase an athlete is in; far fewer know what that athlete can currently hold at threshold, or whether the session just chosen sits above or below it. IntervalCoach scores five energy systems (endurance, tempo, threshold, VO2max, anaerobic) from 0 to 10 by comparing the athlete's best effort in the last six weeks with their best ever at the duration each system is read from, so the number describes them against their own ceiling rather than against other athletes. Both sides of the match had to become measurements. On the athlete side the score was a measurement only for riders with a power meter and a tally of training minutes for everyone else; running is now scored against the pace curve and swimming against swim pace, so **for 96 athletes in 100 the profile is a best effort they actually produced**, with running and swimming pace accounting for **one in eight** of those, and a zone that cannot be measured is labelled "training volume, not a measured best" instead of showing a number that means something else. On the catalog side every session carries a **computed** difficulty per system it trains, derived from working intensity, rep length, rest ratio and load rather than hand-assigned, so it cannot drift when a session changes. The target difficulty comes from what the athlete has been finishing: a point near the top of the range of sessions they completed in the last six weeks plus a small step up, requiring at least five completed sessions in that system before any target is set, with a measured level nudging it by a capped amount and a volume-scored level never nudging it at all. The band kept around that target is deliberately asymmetric, reaching **1.8 points below and 0.7 above**, because being handed something too easy costs one session of stimulus while something too hard costs the session and the days after it. Limits are published rather than smoothed: about four athletes in a hundred have no measured system yet and endurance is most often the missing one (roughly one in ten); the catalog thins at the bottom of the difficulty scale, with endurance offering nothing at all at a target of 3 or 4, which is why the band widens in steps and is dropped entirely rather than leaving a single candidate. The post makes no claim that this produces better outcomes: the loop connecting a session's result back to the score is built but not connected, and the weekly score history needed to test it began in September 2026. ### How AI Coaching for Endurance Athletes Works URL: https://www.intervalcoach.app/blog/how-ai-coaching-works An AI training coach reads three kinds of information: training history and load (fitness, fatigue and form), overnight recovery data from a wearable (HRV, resting heart rate and sleep, plus skin temperature, breathing rate and blood oxygen on devices that record them, each compared with the athlete's own recent baseline), and the athlete's own input (races, group rides and planned sessions on the calendar, a morning check-in, and how hard each session felt). IntervalCoach runs five checks each morning: health (rest when sick or injured, then a gradual return in which hard sessions come back step by step, the reduced load lasts longer after a longer illness, and it lifts early once recovery signals are back to normal), calendar (no extra session on an A or B race day, group rides count as training, self-planned sessions are kept), readiness (overnight signals, individually fitted recovery curves and the check-in; inside the recovery window of a hard session the next day is held below threshold), reshape (reduce load, cap intensity at a zone, swap the stimulus, extend the warm-up or add a breathing block to a longer cooldown), and generate and steer (a structured workout written to the calendar and synced to the athlete's devices, with the week steered toward its training load target). Every adjustment is recorded with the reason it was made, and athletes can ask for changes in plain language through the Coach+ chat. Weekly plans follow one of five periodization models: linear, undulating, polarized, pyramidal and block. Limits stated in the guide: readiness is one input and never overrides what the athlete reports; early signs of illness (led by skin temperature, breathing rate and blood oxygen; in IntervalCoach's analysis of athlete-logged sick days resting heart rate barely moved beforehand on average) ease training but are never a diagnosis, and a resting heart rate rise never triggers that on its own; the coach can only read what the athlete's device records; and recovery is shared across sports while each sport keeps its own zones and tolerance for accumulated fatigue. ### How AI Coaching Works for Triathletes URL: https://www.intervalcoach.app/blog/ai-coaching-for-triathlon An AI triathlon coach plans swimming, cycling and running as one training load and adjusts each day to the athlete's recovery. IntervalCoach reads overnight recovery and training load as one picture across all sports, because the three draw on the same recovery, while running is eased sooner than cycling or swimming at the same accumulated fatigue: inside the recovery window of a hard session the next day is held below threshold whichever sport comes next, and the weekly training load target covers all three sports together. When one sport is the athlete's focus, hard sessions stay in that sport, and a sport the athlete rarely trains is built up gradually from their recent history in it. A brick is a sport the athlete can mark on the weekly schedule, with the plan choosing the bike-to-run pattern. Swim sessions are written as distances in whole 50s at a pace per 100 metres from critical swim speed (CSS), or yards for imperial accounts. On an A or B race day no extra session is added and race-day guidance is given; the session the day before is matched to the race's sport; taper and race weeks keep the shape built around the race; and group rides and self-planned sessions are always kept. After illness, hard sessions return step by step, the reduced load lasts longer after a longer illness, and it lifts early once recovery signals are back to the athlete's own normal. Readiness is one input among several and never overrides what the athlete reports. ### How AI Coaching Works for Runners URL: https://www.intervalcoach.app/blog/ai-coaching-for-running Running produces a higher perceived effort than cycling at the same training load, so IntervalCoach judges running fatigue on running's own scale: a runner is flagged as fatigued at a level of accumulated fatigue a cyclist would still train through, and the run is eased sooner, while fatigue from other sports still counts toward the next day. Inside the recovery window of a hard session the next day is held below threshold, so a planned threshold run becomes tempo or endurance running. Runners choose pace (the default), heart rate or power targets; power targets are a percentage of running FTP from a Stryd pod or Garmin watch, estimated from recent runs or entered manually, and separate from cycling FTP. Running is scored against the athlete's own pace curve so quality sessions are matched to what they have been completing. Norwegian Singles, Norwegian Doubles and the Easy Interval Method are available as training methodologies, setting the type of quality session alongside the periodization model, which decides how much intensity a week holds. Trail running is a separate sport with elevation-aware pacing. A sport the athlete rarely trains is built up gradually from recent history. On an A or B race day no extra session is added and race-day guidance is given; paid plans add a race pace plan with a goal-aware pacing strategy and warm-up protocol. After illness, hard sessions return step by step and the reduced load lifts early once recovery signals are back to normal. Readiness is one input among several and never overrides what the athlete reports. ### How IntervalCoach Adapts Your Workout Every Day URL: https://www.intervalcoach.app/blog/how-daily-adaptation-works A plan written on Sunday doesn't know that you slept terribly on Wednesday. Every morning before the workout email goes out, a five-step pipeline checks whether the planned workout still makes sense for how you're feeling right now and reshapes it if it doesn't. **Step 1 — health check.** A SICK or INJURED event on the Intervals.icu calendar stops the pipeline; you get a rest day with recovery guidance. After recovery, training comes back gradually: hard sessions return step by step, the reduced load lasts longer after a longer illness, and it lifts early once resting heart rate, HRV and recovery are back to the athlete's own normal. **Step 2 — calendar priority.** A or B race today means pre-race guidance instead of a workout; a C event becomes the session; external workouts are respected; weekly-plan workouts get checked for adaptation. **Step 3 — readiness assessment.** Compound patterns across 60+ signals matter more than any single reading. **Step 4 — reshape.** Load reductions, intensity caps, stimulus swaps, extended warm-ups, and breathing-block cooldowns can stack; the most restrictive combination wins. **Step 5 — generate and steer.** A concrete structured workout lands on the Intervals.icu calendar; mid-week TSS steering keeps the week on its target. --- ## Common questions The following Q/A pairs are quoted from the IntervalCoach help center at https://docs.intervalcoach.app/. ### What is IntervalCoach? IntervalCoach is an AI-powered coach for cyclists, road and trail runners, and triathletes that creates personalized workouts based on your fitness data. It connects with Intervals.icu to analyze your training history, recovery status, and goals to generate the perfect workout for each day. ### How does the AI coaching work? Our AI analyzes your power/pace data, training load, recovery metrics, and upcoming goals. It then generates structured workouts that target the right training zones and intensity for your current fitness level. Over time, the coach learns your preferences — which workouts feel harder for you, your recovery patterns, and how you respond to different training stimuli. Workouts automatically sync to Intervals.icu and your connected apps like Zwift. ### How are workout intervals created? IntervalCoach uses a two-step process. First, it selects a workout type (like Sweet Spot, Threshold, or VO2max) from a catalog of ~25 types based on your training phase, recovery status, and recent workout history — ensuring variety and proper stimulus spacing. Then the AI generates the specific intervals for that type, tailored to your FTP/pace, available time, and current fitness. This is why duration may vary slightly from your target — the AI prioritizes the right training stimulus over hitting exact minutes. ### Why AI coaching instead of a human coach? AI coaching offers 24/7 availability and instant adaptation to your daily readiness. Unlike static training plans, our AI adjusts every workout based on your sleep, HRV, recent training load, and recovery status. It's like having a coach who knows your body's data in real-time. For serious athletes, AI coaching complements human coaching by handling daily adjustments while your coach focuses on strategy. ### How does IntervalCoach compare to apps like TrainerRoad or JOIN? TrainerRoad and JOIN are excellent training platforms that provide structured training plans. The key difference is how IntervalCoach handles daily adaptation. Traditional apps give you a plan — often a good one — but they don't know that you slept poorly last night, had a stressful week, or that your legs just aren't there today. You either follow the plan or skip the workout entirely. IntervalCoach adapts every single day. It looks at your recovery metrics (HRV, sleep, resting heart rate), your accumulated fatigue, and your recent training load. Then it adjusts — not just "easy day vs hard day", but the specific workout, intensity, and duration. Some days it'll push you harder because you're fresh. Other days it'll dial back because the data shows you need recovery. Think of it as: static plan vs. dynamic coach that responds to your body in real-time. ### Garmin acquired TrainingPeaks in 2026. What does that mean, and what are the alternatives? In July 2026 Garmin acquired TrainingPeaks and TrainHeroic. TrainingPeaks built its position on being device-agnostic (train with any device, any sport), and it is now owned by the largest endurance device manufacturer. It also originated the vocabulary the sport uses for fitness and fatigue: CTL, ATL, TSB and the Performance Management Chart. Nothing breaks for non-Garmin athletes immediately, and bundling into Garmin's ecosystem may be good value for Garmin users. The realistic concern is slower: where the best integrations and newest features land first, plus what happens to pricing and packaging for coaches whose businesses run on TrainingPeaks seats. The practical advice, whatever platform you use: keep your training history somewhere you can export from, and keep the software that decides your training independent of the brand on your wrist. Intervals.icu is the open hub (it ingests from Garmin, Wahoo, Polar, Zwift and others, is not owned by a device maker, and keeps your data exportable). IntervalCoach is the adaptive coaching layer on top of it, and is device-agnostic. To be fair to TrainingPeaks: if you work with a human coach, or you are an experienced self-coached athlete who wants to build and analyse a detailed plan yourself, it remains excellent at that. Full analysis: https://www.intervalcoach.app/blog/garmin-acquires-trainingpeaks and an honest side-by-side at https://www.intervalcoach.app/vs-trainingpeaks ### What's the relationship between Intervals.icu and IntervalCoach? Intervals.icu is a free training analysis platform that stores your workout data from Garmin, Wahoo, Zwift, etc. IntervalCoach connects to Intervals.icu to read your data and generate AI-powered workouts. Think of Intervals.icu as your training database, and IntervalCoach as your AI coach that uses that data. Workouts created by IntervalCoach are uploaded back to your Intervals.icu calendar. ### How does IntervalCoach determine workout intensity? The AI considers multiple factors: your current CTL (fitness), ATL (fatigue), TSB (form), recent sleep and HRV data, power/pace zones, upcoming events, and weekly training availability you set in settings. It balances progressive overload with recovery to build fitness without overtraining. Intensity targets are based on your actual power curve or pace data, not generic percentages. ### How is my target CTL determined? When you set a goal — whether it's a race event or a training goal template like General Fitness or Return to Sport — the AI calculates the optimal CTL (Chronic Training Load) you need to reach by that date. It considers your current fitness, time available, and a safe ramp rate (typically 3-7 CTL points per week). The goal progress chart shows your projected fitness trajectory and whether you're on track. The AI then plans weekly training load to hit that target. --- ## Comparison with alternatives ### The comparison guides IntervalCoach publishes seven guides that compare the field honestly, each naming the rival that suits a given athlete better. They live at /en/best-ai-cycling-coach, /en/best-ai-running-coach, /en/best-ai-trail-running-coach, /en/best-ai-triathlon-coach, /en/best-hyrox-training-app, /en/best-training-app-without-power-meter and /en/best-hiking-training-app, and each is available in all 25 languages under its own locale prefix. Three of them cover ground the category usually ignores. The trail and ultra guide is built around vertical gain, terrain and time on feet rather than kilometres, and says that Vert is the better home for a trail-only athlete. The power-meter guide is for athletes training on heart rate and pace: it states which platforms require watts (TrainerRoad's plans are power-based, with VirtualPower as an indoor estimate; Zwift needs power or zPower) and which prescribe heart-rate zones outright, and it is explicit that heart rate lags, so very short maximal intervals are prescribed by feel and duration. The hiking guide covers treks and long walking events, points at Mountain Tactical Institute and Uphill Athlete when the limiter is legs under a pack, and states plainly that AllTrails, Komoot and Gaia GPS are navigation apps rather than coaches. A separate guide is written for coaches rather than athletes: /en/best-software-for-endurance-coaches, also in all 25 languages. It splits coaching software into platforms where the coach writes every plan (TrainingPeaks, Final Surge, Intervals.icu coaching groups, Coachbox) and newer tools that generate and adapt each athlete's plan (Athletica, Vekta, Stride, HumanGO, IntervalCoach), and sorts them by who pays: most of the AI tools are free for the coach and charge each athlete a subscription, while IntervalCoach's Coach Plan charges the coach per seat (€10 per athlete up to four seats, €8 from five, €7 from fifteen, individual pricing from forty, minimum three seats) and each seat includes the athlete's full Max plan. It lists what IntervalCoach does not do for coaches: no invoicing of athletes, no direct coach-athlete messaging beyond the shared Coach+ thread, no template library or plan marketplace, and no shared team account on the self-serve plan. A coach who needs to bill athletes through the platform is pointed to TrainingPeaks or Final Surge. Athletes evaluating AI training coaches typically compare IntervalCoach to platforms like TrainerRoad, JOIN, Athletica, and Humango. The honest framing is that all of these are legitimate options that solve overlapping but different problems. The IntervalCoach FAQ frames the comparison as "static plan vs. dynamic coach that responds to your body in real-time." **TrainerRoad** is a long-established structured training platform with a deep workout library and Adaptive Training that adjusts difficulty based on completed sessions. It is excellent at indoor structured intervals and has strong cycling-focused tooling. The IntervalCoach FAQ describes TrainerRoad as "an excellent training platform that provides structured training plans" — the difference IntervalCoach calls out is daily adaptation: a TrainerRoad plan, like most plan-based apps, doesn't know that you slept poorly last night or that your legs aren't there today. You either follow the plan or skip the workout entirely. IntervalCoach adapts every single day, not just easy-day-vs-hard-day, but the specific workout, intensity, and duration. **JOIN** is a Dutch-built AI training app with a strong reputation for adaptive cycling plans. The IntervalCoach FAQ groups it with TrainerRoad as a strong structured-plan platform; the same daily-adaptation distinction applies. Athletes with consistent schedules and reliable recovery often do well on JOIN; athletes whose week looks different from what they planned a few days ago tend to find IntervalCoach's daily reshaping more useful. **Athletica** and **Humango** are also AI coaching platforms in the same broad category. IntervalCoach doesn't make specific public comparisons to either; athletes evaluating those tools should compare on (a) what data each tool actually reads from your wearables, (b) how often that data changes the next workout, (c) which periodization models are explicitly supported, and (d) whether workouts sync structured to Zwift, Garmin, and Wahoo without manual export. What's specific to IntervalCoach as documented on the site: 60+ recovery signals watched before each session, five established periodization models, exponential tapering based on Mujika & Padilla (2003), ACWR-based load capping (Gabbett 2016), and a five-step daily pipeline that can rest, cap intensity, swap the workout type, extend the warm-up, or add a breathing-block cooldown — all of which write to the structured workout file your head unit syncs, not just a label in the email. Pricing starts at €3/month after a 14-day free trial without a credit card. --- ## What athletes say IntervalCoach publishes two kinds of first-hand athlete evidence, both attributable and both checkable at source. **App store reviews.** The iOS app is reviewed on the App Store and the Android app on Google Play. Reviews on the marketing pages are reproduced verbatim under the reviewer's own display name, with the star rating shown and a link back to the store listing. Where a review was written in a language other than the reader's, the page labels it as translated, the way the app stores themselves do. Recurring themes in the reviews: the daily adaptation actually changing the session rather than relabelling it, the Coach+ conversation being able to restructure a calendar, structured workouts reaching Garmin head units through Intervals.icu without manual export, and the pace at which bugs get fixed and features shipped. **Athlete spotlights.** A weekly long-form series in which one athlete's real training block is analysed from their own data, published with that athlete's explicit approval and quotes in their own words. These are not marketing testimonials written for the athlete; each post is built from the athlete's actual files and reviewed by them before publication. Examples: Pius, a Swiss masters gravel racer who builds his engine an hour at a time on an indoor trainer and then rides 300 km of gravel in a single day; Phaidra, a Canadian road and crit racer rebuilding an aerobic base around frequent work travel; Bento Galado, an Algarve trail runner preparing for a 100 km mountain ultra around a full-time job and a young child. The full series is at https://www.intervalcoach.app/blog?tag=spotlight and the review section is at https://www.intervalcoach.app/en/reviews. --- ## Maintenance note This file is hand-maintained. It reflects IntervalCoach as of May 2026. For the canonical pages and the most current content, follow the URLs listed in the index above (or in `/llms.txt`).