Tuin isn't a workout generator with a friendly voice bolted on. Underneath the chat sits a real coaching method — the same physiology and the same training schools a good human coach draws on — turned into numbers, checked against your actual runs, and explained back to you in plain words. Here is exactly how it works, whether this is your first plan or your fifteenth season.
You don't need to understand any of this to benefit from it — you just run and talk to Tuin. But you can understand it, and that's the point: every number comes with a plain-language reason, so you finish the season understanding your own training instead of blindly following a spreadsheet. Nothing here assumes you've read a training book.
You already know a good coach is worth it and most apps aren't — because they can't reason, can't adapt, and can't tell you why. This page is where you check our work: the schools we synthesize, the load model we use, and the places the science is genuinely unsettled where we say so out loud. If the method doesn't hold up, you'll find out here, not after three weeks of bad workouts.
Tuin works in two steps, and keeping them separate is what makes it trustworthy.
Every run you upload is turned into hard numbers by fixed, transparent formulas — no AI, no guessing. Your pace is adjusted for the hills you actually climbed, your effort is scored against your personal threshold, and those scores roll up into fitness, fatigue and form. This all happens on the server, the same way every time, so the numbers can't be invented, rounded up, or talked into sounding better than they are.
Then the finished numbers, your goal, your recent runs and your own messages go to the reasoning layer — Anthropic's Claude — which decides what to actually do: what this week should look like, what to change when you're tired, how to answer your question. It reasons from a written body of coaching doctrine (below), and it is forbidden from inventing or recomputing any of the numbers the engine produced. Judgment on top; facts underneath, load-bearing and untouchable.
The load model is the same Performance Management Chart that has anchored serious endurance coaching for decades — an impulse-response model made everyday by TrainingPeaks, and traceable to Banister's work in the 1970s. In plain terms: stress from each run flows in, your body responds, and a few honest numbers fall out.
6:00/km up a steep wall can cost more than 4:30/km on the flat. Before scoring anything, Tuin converts every metre of your run to its flat-ground equivalent — so a slow-looking mountain day counts for what it really took out of you.
One number for how much a session took out of you, with intensity weighted double — an easy half-hour might be 30, a hard hour 90, a big mountain day over 300. Cross-training counts too: a long hike or ride adds real stress to the same ledger.
Fitness is your training load averaged over about six weeks — the size of your engine, slow to build and slow to lose. Fatigue is the same idea over the last week — how heavy your legs are right now. Form is simply fitness minus fatigue. Racing well means arriving with a big engine and fresh legs, which is exactly what a taper buys.
Tuin watches how quickly your load is rising against what your body is used to, and flags the single sharpest injury signal there is: a run much longer than anything you've done lately. It's a soft guardrail, not a hard rule — more on why in the honest-limits section below.
Every one of these has a plain-language page inside the app, and Tuin links you to it the first time it shows you the number — so nothing on your dashboard stays a mystery.
Tuin doesn't follow one guru. It synthesizes the schools that have actually produced results — on road and on trail — and holds each one where it's strongest. In plain terms, this is the house style.
Roughly 80% of your running should be genuinely easy, clearly below the effort where holding a conversation gets hard. The most common mistake self-coached runners make is letting easy runs drift into the moderate no-man's-land — fast enough to tire you, too slow to build speed. Tuin polices this gently but firmly.
Quality sessions do their job at threshold and sub-threshold efforts you could repeat — not at the ragged edge. And there's a ceiling on how much fast work a week can hold (threshold work stays a small share of weekly volume, intervals smaller still), because past the dose that drives adaptation, more isn't better — it's just more fatigue.
Base first — patient aerobic volume — then progressively more race-specific work, so the last six-to-eight weeks of a plan start to look like the race itself. A marathon block funnels from general fitness toward marathon pace; a 5K sharpens raw speed last. The shape of the plan changes as the date gets closer, on purpose.
This is where most self-coached runners lose the race they trained for: they cut everything in the final week and arrive on the start line flat and heavy-legged. The evidence says cut the volume — by half or more — and keep the intensity, which is the one thing a taper must not touch. So every Tuin race week carries a short sharpening session at goal pace four to five days out, with the work slashed and the pace left exactly where it was, plus a handful of 20-second strides the day before. It builds nothing; it's far too late for that. It keeps your legs awake, rehearses your pace one last time, and you recover from it in a day.
Fitness is built by stressing, then recovering. Tuin raises load gradually and drops a lighter down week roughly every fourth week — deliberately, before you're wrecked — because the adaptation happens during the recovery, not during the effort. Two monster weekends back to back don't make you fit; they make you tired.
The long run is capped by a share of your week and by time on feet — roughly two-and-a-half to three hours — not by a magic distance. The mandatory 20-miler is right for some runners and an injury waiting to happen for others; Tuin sizes it to you, not to a number you read in a magazine.
For trail and mountain running, vertical is its own stream. Climbing strength and the durability to descend are trained on purpose, effort leads over pace on steep ground, and a long slow mountain day is scored for the cost it really carried — not for how modest it looks on paper. Tuin was built by a trail runner, and it shows here.
The best plan is the one you actually absorb, week after week — not the most impressive one on paper. Tuin optimizes for the training you'll finish, which is exactly why it would rather reshape a bad week than hand you a wall of red missed sessions. Showing up for months is the workout that matters most.
A plan that only works if life behaves isn't a plan. This is where an adapting coach beats a fixed schedule — and where the rules come from physiology, not from vibes.
One to three days missed: skip them and resume — never cram the lost volume into the rest of the week, because that spike is the actual injury driver, not the gap. Tuin reshapes the week forward instead of asking you to "make it up", and it never shames you for the days you missed.
Tuin follows the coach's neck rule: symptoms above the neck (a runny nose, a sore throat) usually mean easy running is fine if it feels fine; anything below the neck, or any fever at all, means a full stop — no exceptions, because training through a fever carries a genuine heart risk. The return afterwards is graded, never a jump straight back to where you left off.
Tell Tuin about the holiday, the work trip, the wedding, the newborn week — and it becomes part of the plan instead of the reason the plan broke. A week with a bike or some hikes instead of runs holds on to most of your fitness, and Tuin knows that and will say so rather than write the week off.
After time off, Tuin rebuilds from your current fitness, not your remembered PBs — and treats the first weeks as tissue-tolerance weeks, deliberately gentle, because your legs need longer to come back than your lungs do. That's the boring part most runners skip, and skipping it is exactly how comebacks turn into injuries.
The fastest way to spot a coach worth trusting is whether they'll admit what isn't known. Tuin is built to hold contested ideas loosely and to tell you when it's making a judgment call rather than stating a fact.
The popular 0.8–1.3 acute-to-chronic load ratio has had its evidence as an injury predictor substantially weakened since 2019. Tuin uses it as a descriptive lens and a soft guardrail — never as a hard gate that blocks a session — and leans instead on a sharper, better-supported signal: a single run more than about 10% longer than your longest of the past month.
Daniels programs weekly VO2max intervals; the Norwegian approach barely touches them until a short sharpening block near the race. Both have produced personal bests for thousands of runners. Tuin chooses based on your distance, your history and where you are in the plan — not because one camp beat the other in an internet argument.
Your VO2max figure, your threshold pace, the race-time equivalents next to them — these are calibrated from your runs, not measured in a lab. A bad heart-rate strap day or a heatwave will wobble them. Tuin shows you the trend rather than any single point, tells you plainly it's an estimate, and lets you correct it in a sentence: "my threshold is about 4:30/km" is enough.
Persistent heavy legs, poor sleep, a workout that felt hard at your normal easy pace — these often predict trouble better than any green readiness score. Tuin is built to weight what you tell it heavily, and it will never overrule an honest "I feel terrible" with a cheerful number. You are the best sensor you own.
The reasoning — reading your week, deciding what to change, writing the explanation — is done by Anthropic's Claude, a large language model. We're telling you that plainly because you deserve to know what you're talking to. It's the same technology behind the best AI assistants, pointed at one job: coaching your running.
What that buys you is a coach that genuinely reads your specific runs and your specific words and responds to them — in English, Spanish or Catalan, at any hour — instead of a decision tree wearing a chat skin. What it does not do is make up your numbers: the training figures come from the fixed engine and are handed to Claude as facts it is not allowed to change. That guardrail is what keeps an eloquent model honest.
And no human is reading your training. Your run summaries and your plan go to the Claude API to generate the coaching and nowhere else — nobody at Tuin is browsing your runs. Tuin also never stores your raw GPS tracks or location history; the privacy note spells out exactly what's kept and what never is.
Tuin is in closed beta while it's young. If the method sounds like the coach you'd actually listen to, leave your email and you'll hear from us as places open up.
Read more and join the list →