July went deep rather than wide. The readiness score was rebuilt from the ground up, the training plan learned to actually climb toward your goal instead of flattening at today's fitness, and picking a goal race stopped being a manual chore. Here is what changed since the June update.
Your readiness score, rebuilt
The old score mixed two different questions: how recovered are you, and how does that compare to what your plan wants today. That meant a genuine rest day could still read low, and the number in your email could disagree with the number in the app.
It is now one thing only: how recovered your body is. HRV, resting heart rate, sleep, your own personal recovery curve, and your morning check-in, each compared against your own 30-day baseline rather than a population average. The training plan no longer touches the number at all. Tap the ring and you get every factor with today's value, your baseline, and which direction it pushed the score.
It also stops drifting. The score anchors once your overnight data lands and tells you when it was set. Before that you see a provisional score marked with a tilde, and a 30-second check-in firms it up on the spot. Everything downstream now reasons from that same 0 to 100 number: the briefing, the daily outlook, your coaching notes, the weekly summary, race-day advice, and Coach+.
Two things worth calling out. It works without a wearable too: if you do not wear one, the score still comes together from your recovery curve, your check-in, and accumulated training fatigue, and when there is genuinely nothing to go on we show no number rather than inventing one. And you now choose which questions the morning check-in asks: sleep quality, soreness, fatigue, stress, mood, motivation, injury, hydration, all in one card under Settings then Training.
The wording changed too. Readiness now states what was measured instead of passing judgement. "7.1h of sleep vs your recent 7.6h average" replaced verdicts like "insufficient sleep", in every language.
Training plans that build toward the goal
This was the biggest structural work of the month. Plans used to cap every future week at whatever you had been doing recently, which quietly made an ambitious goal unreachable by design and showed it as at risk from day one. Weekly targets now step up toward the fitness your goal actually needs.
Tapers got honest as well. Race and taper weeks are capped at a load you can absorb, pre-race days are capped at 75 minutes, and the day after a race always materializes as an easy recovery session. The numbers on the Training Plan page now match the plan you are training, instead of being recomputed for display and drifting.
Your calendar also holds real workouts much further out: fully structured sessions for four weeks ahead on Pro and eight on Max, synced the moment your plan exists, so they reach your head unit or Zwift well in advance. And the week rebalances itself in real time. Ride something unplanned and the remaining sessions ease back to keep the week on target. Move or delete one and the rest adjusts. Today and anything already completed are never touched, and you can turn the whole thing off.
Repetition got a lot of attention. Variant rotation was blind to workouts already sitting in your calendar, which got worse the longer your horizon; the fit-the-time step-down always landed on the same session; and an indoor or outdoor preference bonus outweighed everything that creates variety. All three are fixed. When time is tight, quality sessions now get slightly shorter, never more than 25% below planned, instead of being downgraded to an easy ride at full length.
Races you can actually choose
Setting a goal race used to mean creating the event by hand. There is now a browsable catalog of real races, gran fondos, marathons, sportives and more, across Europe, Asia, the Americas and Oceania, each with distance, course details, the full date, and a link to the official event site. Adopt one and it lands on your Intervals.icu calendar as your goal in a single step, with your own A, B or C priority.
You can also browse races other athletes have shared, and share your own. Shared races are reviewed before they appear so the catalog stays clean. Race pages got a map pin, a facts panel showing how many athletes are training for that event, similar-race suggestions, and region and country filters with a live count.
Triathlon is a first-class goal now. Set an IRONMAN, a 70.3 or another triathlon and your plan builds across swim, bike and run rather than a single sport. Races also cover walking, hiking and strength events, each with a notes field for course profile, start time or logistics.
Races are also analysed as races now, rather than as an oddly hard training session. The post-race analysis judges your pacing by phase, start, middle and finale, along with your effort signature and how you held up late on. A completed activity carrying the Intervals.icu "Race" tag counts as a race even when nothing matching sits on your calendar. And if your notes mention how it went, a podium emoji, "P4", "3rd place" or "won", the analysis opens by acknowledging the result instead of talking straight past it.
Your A, B and C races are now completely off-limits to the background plan sync as well. It can never move, shorten or remove a race.
Coach+ can act, and now lives in your AI assistant
Coach+ moved from talking about your training to changing it, and its edits now survive the background plan sync instead of being quietly overwritten.
Telling Coach+ you are sick actually protects your plan: it marks sick, injured or holiday days on your calendar with your approval, backdatable, and the next morning's plan respects it, holds training and eases you back. Asking for a strength or home workout now creates a real strength session with an exercise list, rather than quietly substituting an easy run labelled "strength". Asking to change part of a workout no longer wipes out the rest of it.
There is also a lactate threshold test you can genuinely schedule. Ask Coach+ for one, bike or run, and you get the real field-test protocol: warmup, an unbroken 30-minute time trial, cooldown. Your LTHR is read from the final stretch and written straight into your Intervals.icu heart-rate settings, so your zones recalibrate on their own.
And on Max, you can connect IntervalCoach to Claude, ChatGPT or Cursor. Your AI assistant reads your live training data, including readiness, your personal recovery curves and adaptation signals, and can use your coaching tools right inside the chat you already work in. It lives under Settings then Integrations.
Workouts you can read on the device
A workout is only as useful as what shows on your head unit, so July put real numbers next to every percentage. You now see "74% FTP · 185W", "75-79% Pace · 7:24-7:02/km", "Z4 · 144-160 bpm", on the web, iPhone and Android, in your own units.
There is a new Indoor and Outdoor preference under Settings then Training. Indoor favours shorter, more varied intervals, so a 20-minute threshold block becomes a few shorter reps. Outdoor favours longer sustained efforts and converts every warm-up and cool-down ramp into fixed steps, 50%, 60%, 70%, which any bike computer can follow. Change it and your upcoming workouts rebuild.
Ramps now read as ramps: cool-downs always descend, warm-ups always build, a descending step shows its direction instead of reading "72-67%", and the workout profile no longer paints a cool-down as a hard red effort. Workouts pulled from your library keep their lap-button warm-ups, recoveries and cool-downs, which previously came out with the hard efforts back to back and no rest at all.
Titles tell the truth too. A session named "3x7" stops claiming 7-minute blocks after being trimmed to 6, and resized library workouts drop their stale "4hr". TrainNow can now build around today's readiness instead of only your plan, and each option shows an estimated load such as "≈65 TSS · 60m".
Running gets its own numbers
Runners were being measured with cycling instruments. Run power was compared against cycling FTP, run paces were classified on power-zone boundaries, and peer percentiles were built on cycling power.
If you run with a power meter, your dashboard now tracks running power form in watts, anchored on your running eFTP, with a six-month trend and your distance from a season best. Analytics gained a Running Power history chart alongside Critical Speed.
The classification work matters just as much. Interval runs are now judged by grade-adjusted pace rather than heart rate, so a VO2max session is recognised as VO2max instead of being logged as tempo, because heart rate lags at the start of each rep and the easy jogs drag the average down. An easy run is no longer logged as a hard session because its power was read against your cycling number. Easy blocks stopped being badged "Tempo (Z3)". And if running is your primary sport, peer comparison no longer ranks you against cycling power fields, on Analytics or when you ask Coach+ how you compare.
Fitbit, Pixel Watch, and two new languages
Connect Google Health and your Fitbit, Pixel Watch or any Wear OS device feeds your daily readiness: sleep, HRV, resting heart rate, respiratory rate, blood oxygen, skin temperature and VO2 Max. Workouts recorded on the watch upload to Intervals.icu automatically. Available on Pro and Max.
IntervalCoach also now speaks Japanese and Korean. Dashboard, briefing, onboarding, emails and settings in 日本語 and 한국어, selected automatically from your browser or device language.
A few more improvements
You can set your FTP, threshold pace and threshold heart rate directly in IntervalCoach under Settings then Training, and it saves straight back to Intervals.icu so every target and analysis uses it immediately. Changes you make at Intervals.icu are picked up right away too, instead of after a delay. You can rename or delete a completed activity from its detail view without leaving the app. Analytics charts more of your history, including sleep, VO2max and your acute to chronic training load ratio, plus a Variability Index and Efficiency Factor trend. The goal wizard's "max achievable" fitness target is now measured rather than estimated: every candidate gets a trial run of your actual plan, and the ceiling is the highest target whose own plan reaches it. Runners get a threshold pace target in the wizard, derived from their own Critical Speed history, instead of a cycling FTP number. And a lot more activities look like themselves now, with their own icons for gravel, MTB, e-bike, open-water swims, rowing, gym, yoga, ski and snowboard, walks, and kayak, canoe and SUP.
Bug fixes and reliability
A heavy month underneath, most of it aimed at numbers that disagreed with each other. Post-workout analysis now credits the right session on multi-session days, where a commuter's easy spin could be scored as a failed version of the day's hard workout, and an activity tagged as a commute no longer counts as your planned session at all. The dashboard stopped marking your planned session complete on the strength of an unrelated activity. A fitness test you asked for in Settings now actually appears in your plan, even when the coming week is already full. The calendar's "+N more" button expands the day instead of opening the first hidden activity. Scheduled workouts no longer add up to more than the week's target, because interval sessions score higher at Intervals.icu than the estimate used to build them, so each session's real load is checked and trimmed to fit. Deleted sessions stay deleted and rest days stay clear, instead of being refilled by a background sync. The Analytics page stopped panning sideways on phones. On billing, a web upgrade is now blocked when you already subscribe through the App Store or Google Play, which could previously charge you twice, and downgrading keeps your current plan's features until the end of the period you already paid for. Deleting your account now clears everything, so old badges and history no longer reappear if you start over. And a pile of smaller number-truth fixes landed: completion percentages showing "83.333336%", a training streak stalled around 15 days that made the 30-day badge unearnable, and a ramp-rate warning that said "per day" for a per-week number. See all of it in the full changelog.
More coming soon. Questions or ideas? Just reply to this email.