The Self-Coached

AI Made the Training Plan Worthless

The plan was never the point. Here's what's actually scarce now — and why I built Tempo.

By Dylan, founder of Tempo · 6 min read

You can get a marathon plan in ten seconds. For free.

Open ChatGPT, type "build me a 16-week sub-3 marathon plan, I run five days a week, here's my recent race times," and it will hand you a periodized block with tempo runs, long runs, a taper, and a confident little note about trusting the process. It will look good. It will be good — better than most of the generic PDFs that coaches were quietly selling for €150 a few years ago.

This should terrify anyone whose business is selling training plans. It mostly hasn't, because they haven't noticed yet. But it changes everything about what's worth building for runners — and it's the reason Tempo exists.

The whole industry was selling the wrong thing

For decades, the implicit deal in running coaching was: I have the plan, you don't, pay me for the plan. The human coach sold a plan with some accountability attached. The app sold a plan with some automation attached. Different price points, same core product: someone tells you what to do, and you do it.

That product is now free. Not cheap — free, bundled inside a tool you already pay €20 a month for, that also helps you with work emails and dinner recipes. The marginal cost of generating one more "personalised" plan has gone to zero.

When the price of something goes to zero, you find out whether anyone actually wanted it — or whether they wanted something adjacent that came stapled to it. And I'm convinced, after years of coaching myself and watching every runner I know do the same: almost nobody wanted the plan. They wanted the thing the plan was standing in for.

What people actually wanted

Think about the moments that actually mattered in your training. They were never "I received a plan." They were:

It's 10pm. Your HRV is down 8%, but you feel fine. Tomorrow is a tempo run. Do you do it? You've got a niggle behind your knee — is this the kind you run through, or the kind that becomes six weeks off? Everyone on the podcast is talking about double thresholds and zone 2 religion — does any of it apply to you? You're three weeks out and you feel flat — did you peak too early, or is this just taper anxiety?

None of those are plan problems. They're judgment problems. They need a second opinion from something that knows your training, respects that you know your body, and is willing to think it through with you instead of barking an instruction.

That's what the good coaches actually sold, underneath the plan. Not the periodization — the conversation. The "here's what I'd weigh, but it's your call." AI didn't make that part cheaper. AI made it more scarce — because the free tool that now writes flawless plans is uniquely bad at exactly this.

Why the free tool can't do the part that matters

Ask a general-purpose AI about your training and three things go wrong, every time.

It has no memory of you. You paste in your last few runs. It reasons over that paragraph and forgets you the moment the chat closes. A thinking partner that resets every conversation isn't a partner. It's a very smart stranger you re-introduce yourself to every day.

It agrees with you. Tell it "I want to do intervals today even though I'm wrecked" and it will find a way to say yes, with caveats. The entire value of a second opinion is that it's willing to be a second opinion. Agreement isn't coaching. It's a mirror that nods.

It can't see your numbers. It doesn't know your HRV trend, your acute-to-chronic ratio, last night's sleep, the fact that this is your third hard day in a row. It reacts to what you type; it never surfaces what you didn't think to ask.

So the runner ends up doing the work the AI can't: holding the whole context in their own head, deciding alone, second-guessing alone, at 10pm, again.

The runner who never wanted a coach

There's a specific person this matters to most, and it's the person I built Tempo for — because it's me, and probably you.

You coach yourself, and you're good at it. You've read Daniels and Pfitzinger. You wear the watch, the chest strap, maybe the ring. You adjust your plan based on how you feel, and then you lie awake wondering if you should have.

You've considered hiring a real coach. You didn't. Not because of the money, though that's real — but because you don't want to give up control. You enjoy the decisions. For years, the industry read that person as a budget problem. That's backwards. Self-coaching isn't a compromise you make until you can afford a coach. For a lot of us it's the preferred mode — the part of running we actually like. The problem was never that we lacked a plan. The problem is that doing it entirely alone is heavy.

What that runner needs has a precise shape, and it is not a coach. It's a witness with expertise. Something that knows their training as well as they do, holds the context so they don't have to carry all of it, has a real opinion when the evidence is clear, and then gets out of the way and lets them make the call.

What I built instead

Tempo is that. Deliberately, it is not a coach. It doesn't hand you a plan and tell you to follow it. It doesn't grade your week or manufacture guilt when you miss a run. It doesn't take the wheel you insist on holding.

What it does: it syncs your actual data — your runs, your HRV, your sleep, your load — from your Garmin or Whoop or Oura, so it already knows your last twelve weeks. It notices when you've had three hard days and says so, before you ask. And when you bring it a decision, it does what a good thinking partner does:

"HRV is down but I feel good. Tempo run today?"

"Your HRV drop is modest — about 5%. Combined with good subjective feel, you could do the tempo. I'd start conservatively: if you hit 3K feeling strong, proceed as planned; if the effort feels high, drop to steady state. What's your read?"

Notice what that isn't. It isn't "yes." It isn't "no, rest." It's the reasoning laid out, the trade-off named, the numbers specific — and the decision left exactly where it belongs. With you.

The measure of whether it's working isn't whether you become dependent on it. It's the opposite. In six months you're a better coach of yourself than when you started — and the ceiling keeps rising.

Why now, and why me

Two things had to be true for this to be buildable. The wearables got good and ubiquitous, so the longitudinal context — the thing the free chatbot lacks — can be there automatically. And the models got good enough to reason over that context with real judgment, to hold an opinion, to push back without being told to. Five years ago you could have one or the other. Now you can have both, in your pocket, at 10pm.

As for why me: I'm not a coach selling you my plan. I'm a self-coached runner who got tired of carrying it alone and was unwilling to give up control to fix that. I built the thing I wanted to exist. Every design decision in Tempo comes from one conviction, which is also the bet of the whole company:

AI can write the plan. It can't think alongside you. That part just got scarce — and it's the only part that was ever worth paying for.