Show HN: My triathlon coach (not a human) now texts me on iMessage

daniel_s1 pts0 comments

When I started triathlon in 2018, my “training plan” was writing down what I did in Apple Notes and watching YouTube for advice. Later, training for my first full-distance IRONMAN, I paid a coach 130eur/month with the goal to learn early so I could self-coach later. When I switched to a static plan (bought a 40-week one I’d reuse), I started spending 20-30min every Sunday reshuffling sessions around work, travel and private commitments.In early 2025, training for the 70.3 IRONMAN World Championship, I figured AI would finally fix my biggest pain point: automatically adapting my zones and plan from my activity data. But nothing out there did, and I was reluctant to build it myself (I was determined to build in B2B). A year later, once LLMs had improved a lot, I handed my Sunday planning to Claude, and it worked far better than I expected. Friends wanted to try, so I started shipping it to them.Hundreds of commits later, I believe the thing that matters most is the data layer. Most “AI coaches” smell like AI slop from a mile away (more often than not, just a ChatGPT wrapper designed in lovable). Not saying our landing page isn’t detectable as Claude-generated, but the core has always been proper integration of wearable and app data (Garmin, Whoop, Hevy, Zwift…) - our core product is essentially an MCP server.Also, we don’t think the future is apps built for millions (n=∞), but custom (n=1) or conversational products. So we now solve for both: direct MCP access so you can build your own dashboards in your preferred tools, and a coach on iMessage that proactively writes and adapts your plan from your activity and recovery data (it reads 20+ wearables and apps) plus your preferences. Founder tip: connect your calendar, so it picks up travel/commitments automatically.As a bootstrapped consumer startup, our biggest challenge is pricing (we charge yearly/monthly subscription): consumsers still expect a flat subscription, but our marginal cost is per-message inference. Curious to hear how others handle variable costs in an environment where old (subscription) habits prevail. And what your thoughts are on the future of apps!!

coach plan later data started training

Related Articles