Here is my approach to being better at cycling — one that could help you in any sport.
Most AI content focuses on productivity or business. That's not surprising, given that even households with high internet usage and Amazon Prime rarely use AI tools. If you're mostly scrolling LinkedIn, you might not notice how few people know that prompt quality does matter.
After a few posts in the same direction, I decided to write about things I use — not daily, but hourly.
Short disclaimer: I'm a big fan of road/gravel cycling. I celebrated my 10th anniversary in the sport this year, watching it and having fun on the road. However, the story is not about cycling; the story is definitely about the LLMs.
Improving as a cyclist is straightforward: it takes focused training, strength work, and good nutrition.
Back then, it used to be expensive and difficult to manage everything. One coach would handle cycling, and another would handle your gym work. You might use apps to track nutrition and activities, and spend hours searching Reddit or Quora for insights. Or you could just ride for fun — which, honestly, might be the best option if that's enough for you. :)
Things have changed dramatically. It's still not a revolution, but it's a big leap compared to the reality we had two years ago.
Let's Set It Up
Here is my approach to being better at cycling — one that could help you in any sport. I have three coaches: one for cycling plans, one for nutrition, and another one for strength and conditioning.
The first coach uses my data from Garmin and Whoop to analyze my current conditioning, recent activities, recovery, and prepares a weekly cycling workout plan.
The second coach uses photos and simple notes to calculate my daily protein and carb intake, syncing them with the first coach.
The third coach reviews the outputs from the first two coaches, prepares a general conditioning program for the gym, and works on my injuries and physical problems.
Guess — how many of them are real people?
The answer is simple — just one: the third coach.
I hope it's only a matter of time before all your favorite sports apps stop delivering simple activity aggregation with absolutely useless insights like 'You've done a great job.' ChatGPT can already read your data — even from screenshots — and give you meaningful, structured feedback.
A typical answer from a nutrition coach:
🧠 – Friday, July 26 Interpretation
You completed an interval session at 1:44 with an IF of 0.82 and a TSS of 118 — solid work in Z3–Z4, with some inclusion of Z5. Heart rate response was appropriate; NP was 179W with an FTP of 217W — all within target ranges.
Major Message
To build power and endurance for long gravel races (like Tenerife or Traka), you need to consistently stress Z4 and adapt your body to working in tempo and threshold zones.
Question
How should the weekend workload be structured to preserve Friday's gains, avoid burnout during long rides, and still improve endurance?
Minor Message
You're ready for a 3.5-hour Saturday ride, as long as Z4 is kept in check and structured Z3 efforts are included — with proper fueling and heart rate control.
Action Plan
Structure Saturday's session as:
• 3.5 hours with Z2 as the base
• 3×12 minutes @ Z3 (tempo) efforts with cadence at 65–75
• Consistent fueling (60g carbs/hour), monitor heart rate → to develop sustainable gravel performance without burnout
A typical answer from a cycling coach:
✅ Weekly Recap (July 22–28) & Strategy Ahead
We're wrapping up a well-balanced and productive week with solid progression and recovery integration. Let's dive into the data and map out the next phase.
📅 1. Training Overview
Date | Session | Focus | Platform | Completed | Δ (%)
Mon | Zone 2 Ride | Recovery | Garmin | ✅ 1:06, 138W | —
Tue | Strength (Lower Body) | Hypertrophy | Strava | ✅ 1:14 | —
Wed | OFF | Recovery | — | ✅ | —
Thu | Strength (Upper/Core) | Core + Endurance | Strava | ✅ 1:19 | —
Fri | Intervals Z4/Z5 | Threshold + VO2max | Garmin | ✅ 1:44, TSS 118 | —
Sat | Zone 2 Volume Ride | Base + Social Ride | Garmin | ✅ 2:42, TSS 62 | ↓
Sun | Tempo Intervals | TSS Top-Up, Z3–Z4 | Garmin | ✅ 1:32, TSS 93 | ↑
📈 2. Weekly Load Summary
Period: Jul 22–28
Sessions: 2× Strength + 4× Rides
Duration: 9:23
TSS (est.): ~390
IF: 0.68–0.82
Key Zones: Z2, Z3, Z4
Recovery: Stable
📌 3. Weekly Insights
• ✅ Two quality Z4–Z5 interval sessions (Fri & Sun)
• ✅ Consistent strength work: 2× ~75 min = solid foundation
• ✅ Strategic rest: midweek OFF aligned with recovery feedback
• 📉 Saturday was lighter—but effectively offset by Sunday
• 🧠 Progressive training load without recovery compromise
🧭 4. Next Week Plan (July 29 – August 4)
Mon | OFF or Easy Z1 Spin | Recovery | 30–45 min
Tue | Strength | Lower Body | 1:15
Wed | Bike: Z5 + Z4 Intervals | Power | 1:20 | 3×2 min @ Z5 + 3×5 min @ Z4
Thu | Strength | Upper/Core | 1:00
Fri | Bike: Cadence + Z3 | Technique + Tempo | 1:15 | 3×12 min @ Z3 @ 65rpm
Sat | Long Zone 2 Ride | Endurance | 3:30 | Z2 steady, preferably with tailwind
Sun | Recovery Ride | Recovery | 1:00 | Easy Z1–Z2
🎯 Weekly Goal: TSS 400–420. Focus on building Z5 power, refining cadence, and stabilizing Z3 endurance. Maintain 2× strength sessions.
Here is my heads up to Strava with their AI observations :)
The process
How I'm doing this — For the first two, I wrote two big prompts for separate projects in ChatGPT. DM me if you would like some prompt examples. Each started as a single chat that I can use for over a month without restarting.
Then, I manually align them and allow them to read each other's recommendations and my data. Here is another call to action for all sports app developers.
Then the real coach adjusts my workload and prepares injury-specific training. Here we go – one more use-case for sport apps.
It works relatively well with limited interruptions. ChatGPT can easily replace your favorite (and boring) calorie tracker — you don't have to do anything, just chat, speak, or send photos of your meals to calculate your daily intake and receive a human-like message such as "it would be better to add some protein by the end of the day for better recovery." It works with very high accuracy – just try it.
ChatGPT can easily analyze your data from Garmin or Strava and provide much better insights on structuring your weekly/monthly activities, rather than paying for Strava Premium, etc.
The real coach can take a look with a certain amount of skepticism, slightly adjust everything, and delete irrelevant things.
While it may sound complex, this setup is far more straightforward — and cheaper — than hiring three human experts or configuring clunky fitness apps. Anyway, you're probably already paying for the ChatGPT.
Send me your daily use cases that are not related to productivity or your business activity, and let's organize a second part of this article with your insights!
And if you are planning to race at Traka/Unbound next year – let's stay in touch 😉