I Hired an AI Cycling Coach…
Apparently, This Is What Midlife Optimisation Looks Like
There are various recognised responses to turning 55.
Some people take up yoga.
Some buy expensive watches.
Some carry on regardless, staring quietly into the abyss.
Sounds dramatic
I, on the other hand, have decided to outsource my cycling fitness to artificial intelligence.
Not because I was broken. Not because I’d hit rock bottom. But because, somewhere between Zwift races, Mallorca climbs, and the creeping suspicion that gravity was winning, I began to wonder whether my long-standing training philosophy of ride hard, hope for the best was perhaps no longer scientifically sound.
Also, if I’m being honest, there was a subliminal ambition humming away in the background: to feel again like the version of myself who, in 2013, rode from Le Havre to Paris to watch the Tour de France, rolling around the Champs-Élysées a few hours before the pros. Fit, motivated, and utterly convinced I’d cracked cycling forever.
Spoiler: I hadn’t.
The Problem With “Just Riding More”
I’ve loved my cycling for the best part of twenty years. Really, it began much earlier, specifically the moment my parents bought me a Raleigh Arena for Christmas. That bike carried me through childhood adventures and into my first job as a paper boy. Freedom on two wheels, paid for in newsprint, dirty hands and early mornings.
Somewhere between sixteen and thirty-five, I lost my way a little when it came to physical fitness. Life happened. Then, thanks to a really cheap and really heavy mountain bike, cycling found me again.
What followed were more adventures, this time shared with friends who, like me, were becoming increasingly aware of the passing of time and the quiet expansion of waistlines.
More recently, the bike has meant Mallorca trips, iconic climbs, long days with friends, and solo rides where it becomes a therapist with wheels.
But training-wise, I’d fallen into a familiar middle-aged groove.
Zwift races. Hard sessions. Chasing watts. Treating rest days like a moral failing.
I was still reasonably fit, but everything felt slightly… brittle. I could go hard, but not often. I could survive tough rides, but recovery took longer. And beneath it all lurked the uncomfortable thought that maybe this - slightly fading, slightly fatigued - was simply how it was going to be now.
Enter AI (Via a Petrol Station)
The idea of using AI crept in quietly rather than arriving as a grand plan.
I was in Mallorca in October and discovered, halfway up Sa Batalla, that enthusiasm had once again outpaced preparation. I ran out of legs, then energy, and finally excuses; a reminder that fitness, like optimism, needs supporting properly.
I was ultimately rescued by the last bite of Anne’s protein bar and a couple of Sue’s jelly babies, while Richard, who had been racing ahead of everyone in his unofficial role as trip photographer, dropped back and gently nursed me up to the infamous Repsol petrol station at the top of the climb.
No, I’m not sharing those photographs.
That evening, slightly humbled, I started asking ChatGPT some straightforward questions about nutrition. When I shared the next day’s route, it suggested a fuelling and hydration approach that was practical, specific, and, crucially, doable.
I followed it. The difference was noticeable.
Back home, the thought lingered. If it could help me think more clearly about what to eat and drink, could it also help me train a little more intelligently? I decided to find out.
Human, Assisted
I started feeding it ride data from Zwift and Strava. I explained what I’m working toward. In an almost daily dialogue I talk to it about time constraints, work, travel, sleep. Real life.
What comes back isn’t a rigid plan. It’s something far more unsettling.
Restraint.
Zone 2 rides. Tempo sessions that feel suspiciously manageable. Sweet Spot workouts that end with me standing up normally rather than crawling toward the fridge like a wounded animal.
It actively encourages rest days. Celebrates them, even. Calmly and persistently explains that adaptation happens when you recover, not when you bury yourself heroically for the fourth time in a week.
This is deeply inconvenient information because it runs completely counter to everything I’ve spent years believing about fitness. That progress comes from suffering. That if you’re not empty at the end of a ride, you haven’t done it properly. That backing off is weakness, and rest days are something you earn only after near-collapse.
AI, it turns out, has no interest in those narratives.
Conversation, Not Commands
The biggest change isn’t the sessions, it’s the interaction.
I’m not handed a plan and told to comply. I talk through how rides feel. I question stats. I admit when I’m tired, busy, or mildly resentful of Lycra. I explain late nights filming gigs, long drives, disrupted sleep.
And instead of the plan collapsing, it adapts.
Miss a day? Fine.
Feel flat? Back off.
Feeling strong? Hold steady — don’t get greedy.
For perhaps the first time, training feels adult.
Flexible, but not vague. Structured, but not punitive.
The Results
Let’s manage expectations. This is a work in progress.
I have not woken up looking like a mid-50s Geraint Thomas. I still possess a perfectly respectable layer of winter insulation.
And yet…
Zone 2 feels easier.
Tempo feels comfortable.
180 watts now feels like somewhere I arrive, rather than wrestle into submission.
Sweet Spot sessions leave me energised instead of wrecked. I’m finishing rides feeling revived — which, at 55, feels suspiciously like witchcraft.
Even my belt has noticed. Tentatively. We’re not making any promises yet.
Zwift, Reconsidered
Don’t get me wrong, I think it’s fantastic but I’ve always suspected Zwift was part of the problem. Too virtual. Too gamified. Too easy to turn every ride into a race against strangers called things like PainCave69.
But it turns out Zwift wasn’t the issue - I was.
By riding free rides properly - coasting downhill, easing off, riding terrain as I would outdoors - it’s become a genuinely effective training tool. Not a simulation of real cycling, but a controlled environment for practising restraint.
Which, frankly, has never been my strongest suit.
Why Coach AI Actually Feels Like It’s Working
AI isn’t magic. And it’s not replacing good human coaches any time soon. But what it does brilliantly is remove ego from the equation.
It doesn’t care that I once rode the Champs-Élysées.
It doesn’t flatter me.
It doesn’t panic when I miss a session.
It doesn’t get bored when I ask the same question three times in slightly different ways.
It just quietly nudges me back toward consistency.
And consistency, it turns out, is devastatingly effective.
The Long Game
This isn’t about chasing KOMs or reliving youthful glory. But there is, undeniably, a ghost riding alongside me … the 2013 version of myself who felt light, strong, and utterly in love with cycling.
I don’t need to be him again. But if I can meet him halfway - fitter, wiser, less obsessed with suffering - I’ll happily take that deal.
The immediate goal is simple enough: a couple of spring sportives with my mate Alan, threading our way through the Yorkshire Dales. Long miles. Honest climbs. Steady effort. Good company. The kind of rides where you’re not racing anyone, but you still want to feel capable … strong enough to enjoy the day, and sensible enough to finish it with something left in the tank.
Which brings us neatly to Christmas.
Late nights. Big meals. Social obligations. The festive period’s well-worn promise to loosen routines and soften resolve.
Will discipline survive mince pies and red wine?
Will AI intervene?
Will January be a calm continuation or the traditional, slightly sheepish “reset”?
That, I suspect, is where this experiment starts to get properly interesting.
Merry Christmas — and welcome to The Drop Bar Diaries.
Next time: festive excess vs artificial intelligence. Place your bets.
If you’re on Zwift, you’ll find me riding as Tim Sidwell — usually somewhere in Zone 2, practising restraint.
