We’re recording three podcasts in the next week, and we’d love your questions!
We’re interviewing Siren Seiler, who has written articles about her research into Hyrox training. What are your questions about Hyrox and training for it?
We’re interviewing Dr. Peter Raeburn about longevity in endurance sports. We’re going to discuss the key elements in what enables masters athletes to continue performing their best into their older age.
Finally, we’re discussing the AI Coach in Athletica. I’m going to ask Paul and Marjaana questions about what the AI coach can do for us, what kinds of questions we should be asking it. What have you been asking the AI Coach and what would you like to know?
I’d love to hear a conversation on “what can we do with the AI Coach vs. what should we do with the AI coach.”
Longer Context: I’ve gone through the plan builder, have my training roadmap, and can see my workouts planned through my next event. I’ve been using the AI Coach to essentially override/overwrite the planned calendar workouts by giving it a prompt like “I am in a base phase for an (event type) event in X weeks. Look at my Performance Potential data over the past 42 days and over the last 7 days. Take note of any high intensity or high load workouts over the last 7 days. With this data, and knowing that I do 4 strength sessions, 2 upper body and 2 lower body, during base weeks suggest an outline for next week and then give me specific workouts that fit that outline. One day must provide a choice between an indoor bike ride and a run.”
I do this because I tend to throw in some unplanned outlier intense workouts in the winter months (mostly related to team sports) and want to prevent over training.
I feel like this undercuts, or at least circumvents, the hard work that folks have put into the various training plan templates and the compute work that happens during the plan building to get workouts onto the calendar. Realistically, I don’t even need workouts on the calendar if I am diligent enough to prompt the Coach AI every Sunday afternoon and add them myself. On the other hand, I feel like this is the direction the platform should be going, especially if the AI Coach has been trained well on HIIT Science. Customizing an athletes week ahead based on acute load and other recent or upcoming events is what a lot of folks are looking for from an AI based training platform.
No rider of the age 80 or over has successfully completed Paris Brest Paris. A 1200km event that takes place over 4 days every 4 years. Riders over the age of 80 have tried but failed to finish. For an athlete currently in their 50/60/70s what would you say is a good approach roadmap to being that successful 80 year old?
AI Conversational coach
Ive been asking it about changes to my cycling all rounder plan to tailor it better to the demands of ultra endurance. I’ve asked it to produce a sample week based on my current fitness. I have asked it to change some days round based on my preferences. Ive asked it to apply the principles of progressive overload and give an outline of the next 8 weeks.
I have manually adjusted my plan for the next 4 weeks (including this week) based on the suggestions. I will then review how i feel it has gone with the changes.
You can see how the performance potential looks a bit different with the twins peaks now being there. This week doesn’t fully as i am away hill walking over the weekend.
Really interesting question, @jhbuchholz. You’re essentially using AI to create a training plan based on your parameters and schedule as well as based on your training history.
Exactly. And I’m curious to know whether this approach should be used sparingly or if the AI Coach can create a full training plan (one week at a time) of similar quality to the Plan Generator.
I discussed this with my Athletica ai coach and her answer was Yes. I am not fully convinced though she could do it. That’ll be an interesting point to discuss about:
Question to AI coach as asked: “Every morning just after waking-up I do the stairs test to assess my readiness. It consists of climbing the stairs of my house quickly and feel the pain in my legs upstairs. Today there was some pain (rpe 6 of 10), but not excessive, from yesterday’s HIIT session. If I give you in every debrief my RPE in this stairs test can you use to assess my fatigue and adapt accordingly future workouts? For me personally it is far more reliable than hrv, which I have already tried in the past and doesn’t make any sense for me”
I was thinking more in technical terms, compared to other LLMs, would be helpful to explain some of the basics of LLMs and what technically is a bottleneck for the platform? How you try and solve it? How much space is need for it, and how the membership money helps developing it further? Some feedback I’m seeing on the forum might stem from lack of understanding LLMs rather than it be a sports aspect on the top level. It might give you as the owner less headache in the long run I’d argue
Speaking of which it’d be nice if Athletica gave us the possibility of adding our own freshness test to the app where we could give feedback every day to ai coach
