Sure, it’s probably optimal to bias towards intensity the less time you have. However that was not what was said.
It’s was pretty explicitly “you don’t make gains on endurance” and “your progress is only 100% on hard intervals”
Yeah, agree that the last statement is silly. I was speaking more to the broader topic of why tr doesn’t generally prescribe long endurance rides. Much of their approach is about progressing the watts, not the duration. And if you have a fixed and limited amount of weekly training time (tr’s primary target audience) it makes a lot of sense.
If you have a schedule that allows you 5 hours on a Saturday and only 2 hours the other 6 days, than the two one hour workouts (intensity) and a fun, epic, 5 hour ride outside is exactly what it’d do. But that’s me.
As for the AI… and this is for any training plan… but it can check every box it is supposed to and be scientifically sound, but if the rider doesn’t do it than it’s not a good program for that rider. I think this is where the “art of coaching” comes in and is something that can’t be duplicated by a computer. Personal coaching, trust, buy-in… is a human thing. I think many TR users had that buy-in with the old plans and podcasts with Coach Chad and the guys talking through the training. Now the human element is gone and we are asked to trust a computer. It’s a hard sell for some.
I think the old plans were pretty straight for everyone to understand. Base, build, speciality and people got that. With the ai training there is a lot of change in the workouts and people don’t like the uncertainty. With my experience with coaching, you’d follow quite a structured plan with the odd change here and there to the workouts but at least you could discuss this with your coach. I think the changes to the workouts without the “why this is changing” is a problem. I’m surprised they haven’t implemented something like the Strava ai comments that explains why things have changed. I’m sure it would remove some of the uncertainty.
Back to the main question:
In my experience, AIFTP is still different from any traditional notion, higher in my case, forcing a bias towards intensity.
The FTP prediction is still very generous (7% and above at the start of my last block), which discredits it right off the bat. Of course, once struggles happen (not surprising due to the intensity bias), it drops down.
The workouts, in their own right, look doable and meaningful. That’s pretty much the only thing that keeps me using my account. Not enough to renew come Christmas though.
If you’re looking for some reason behind what the AI is doing, there isn’t. It’s as untraceable as ever.
It works for me, i train only outside. Don’t care about prediction just about detection which is spot on for me ( every time). Gained 22w since January at lowest volume in my training history. It serves good workouts. In my opinion predicting ftp is wrong. I can predict my ftp. You must not care about monthly or two months gains in range if 1-2% which is the accuracy of best powermeters. As for outdoor training it a bit different due to many variables. Don’t put too much emphasis on exact power. I long passed the time of newbie gains. For my volume I expect no more than 10-12% increase for whole year. If you trained 7 years prior I am absolutely Shute that you already know what works and what not. Training outside on rolling terrain or gravel 7-8 for watts more or less don’t make intervals wrong. About TR changing workouts that’s normal and it expected. What would a coach do if workout is harder or easier than intended…it will change it accordingly. TR does the same.
- AI Workout picker - truly phenomenal - great/productive feeling every time - worth the money on it’s own.
- AIFTP - downgrade from old version - I don’t use it anymore - I just nudge up my FTP when the AI workout picker struggles to find me workouts. Wish they brought the old one back.
- Predicted FTP - Personalily think they got so excited that they “could” they didn’t stop to think if the “should” - I turned it off (manual FTP) and am happier for it.
- Plan and Calendar - as good as any others I’ve found.
- Workout Player - still my favorite out of all I’ve tried.
100% agree with all of this
No idea…they claim its AI.
When I need to train indoors I’ve been using Trainerday on my phone. Quite a basic app but very stable, have had zero issues with connectivity, power or HR dropping out. I use Claude to create and push the indoor trainer workouts to my intervals.icu calendar where they automatically get picked up by TrainerDay. Trying to manage my own training for now, which is easy enough for the moment since it is just lots of endurance stuff and the occasional CX race. Might re-evaluate when the 12 week base block is done.
Not sure if it’s a feature, but an Ai On/Off Toggle switch is needed.
All in on ai and after 9? months unable to deal with someone who dares to train outside.
Still full on time crunched.
If you train indoors only and at low volume its fine, if you train (or ride) outside and want to do higher volume its broken.
Hey @Sten,
Seven years is a long time and we appreciate you sticking with us for so long. Aging, stress, and inconsistency are real factors, but so is losing trust in the tool. If you’re second-guessing every workout, that’s something that we need to fix.
TR won’t likely ever go back to a static plan. It’s built to adapt, so workouts will change, but keep in mind that you have more control than it likely felt like last year.
FTP: You can always set your own value. If you trust your Intervals.icu number, you can train with that. Different tools estimate FTP from different data in different ways, so they rarely match exactly. What matters is whether the number you train off makes the workouts feel the way they’re supposed to.
Adaptations and surveys: Your survey answers do feed into what gets served next, so they matter. If you’re consistently marking workouts very hard or failing them, the system will respond to that. You also don’t have to accept every adaptation.
I do hear your feedback on the “why” part. I get that you want to understand the reasoning behind a change a bit better, especially when it’s a downgrade, and I’ll share that with the team specifically.
One thing to consider is that most of the issues you’ve experienced with FTP Predictions/Detections while training outside this year likely won’t follow you when you move to indoor training this winter.
Just shout if I can help out with anything moving forward. I’m always happy to chat.
Let me know if this helps! 
Is there a way to approve or reject adaptations to AI workouts?