https://www.repth.com, an AI cycling coach that I vibe-coded back when gpt-3.5 was hot and vibe coding wasn't yet a thing. We've come a long way!

Free for athletes, but I license the underlying "coach" logic to actual, human coaches.

▲

spuzvabob 4 days ago | parent | next [-]

Neat! We made a similar thing years ago in the pre-AI era, with strong focus on detection of how the planned workouts were executed rather than making plan adjustments just on a single scalar metric (e.g. TSS). Didn't really go anywhere unfortunately.

https://www.kickstarter.com/projects/1725424103/summit-train...

	▲	mvkel 4 days ago \| parent [-]
		Nice! Curious, what metrics did you land on? IF, VI, duration, etc., or an interval detector? That is the unfortunate thing about the serious cyclist industry; it's tiny! Not worth pursuing as a business unless the TAM is expanded broadly

▲

muzani 5 days ago | parent | prev | next [-]

Congrats. This sounds like a great way to build a "wrapper" product. Gather clean data on a particular niche and keep it in one place. Have something that's good at processing that data. No matter what "startup killers" the big boys release, it's still resilient and valuable.

▲

Brystephor 5 days ago | parent | prev [-]

How does this compare to Trainer Road?

▲

mvkel 4 days ago | parent [-]

TR's "AI" claims are a bit of a stretch. It basically just monitors fatigue and will ramp down a prescribed workout to ensure you don't overtrain.

The workouts themselves are templates chosen from a list and do not adapt to the rider as an individual. It plugs in a standard periodization schedule with flexible dates.

Repth uses AI for everything with a few guardrails. You pick a peak date, describe your peak event, weekly availability, and ftp.

Repth then generates a macro plan, and the next week of workouts. As you perform, it will monitor for compliance and adjust the prescription depending on your compliance and feedback. All of it is unique per user, optimized for the demands of the specific event.

Overall, TR's approach is fine. Any plan can work as long as you stick to it. I built Repth simply to replace my (human) coach in 2022 and in that regard it has been a huge success

▲

ferrufino 4 days ago | parent [-]

How did you promote it? I find myself in the same market but for planning based on route weather and fitness conditions w/ https://brezza.cc - It is very hard.

▲

mvkel 3 days ago | parent [-]

Haven't promoted it at all, beyond having a marketing site live on the web with competent SEO. I should say, it's not hugely popular or anything. But considering that I only built it for my own needs, I look at it more as art than craft. In that way, it can only be a success.

	▲	ferrufino 3 days ago \| parent [-]
		love that response, congrats!