Preface. Love TR and the TR crew.
Body. Feels like a while since we’ve had a new big ticket feature (happy for disagreement of course)… and Jonathan keeps hinting that something cool is coming. So, what do we think it is? Forecast FTP? WLV2? Something else rad?
We’ve been waiting over 5 years for WLv2 so I doubt it’s that.
The end of workout levels and just using machine learning to pick the appropriate workout for that day taking all your previous riding into account and depending on what your goals are.
I think it’s got something to do with time progression in the plans. 1 hour to 1.5 hour, 1.5 to 2 hours, etc. I know they already added volume check but think it may be something along those lines aside from just increasing time to exhaustion or increasing FTP.
His training he did for Leadville sort of sounded like it had that kind of functionality baked in.
1-2-1 training so I can finally break 1000w peak power in 2026, just kidding we’ll see WLv3 before that happens!
AI coaching. You can put in prompts like, “I hate over-unders and won’t do them so don’t give them to me.” 
In reality, it’s probably the “pick the perfect workout” for that day (a train now of sorts that also looks forward and considers your upcoming schedule).
My assumption was that it was forecast FTP / target FTP which would be based on AI assessment of your physiology, your proposed volume and a given % adherence (historically based). If it said I could feasibly have an FTP of 320 in 6 months by doing X, I know I would find that incredibly motivating!
Chad is coming back ?
It would be great to have built in volume increases when appropriate.
I thought that in my head but didn’t want to start stuff.
This would be the best!
And a return of beers-with-chad.
I was under the impression that TR staff has had access to this for years now. This is how you would validate ML algorithms, use a partial history of athletes as input and check the predicted FTP with the actual one. So I don’t think that’s it.
Here are a few guesses I have:
That always has sounded dope.
All three of those would be fantastic additions!
Maybe doing away with FTP all together an anchor for each energy system.
Also, making training volume much more important in their training philosophy. Training volume has been pretty rigid and placed everyone in boxes. I’d like to see volume much more individualized - I think that’s what Jonathan was hinting at for his Leadville prep.
I really hope is has to do with time recommendations. I tried join for a few weeks, and I really liked how I could tell it each weeks time avalibility and it would use, or not use that time as it saw fit. It would be great if TR had a similar feature.. rather than trying to tell it how many hours I can do each day and it says the volume is to high… like yeah no duh, I want you to take the time avalible and optimize it (whether thats riding or not)
It’s a great point. So, should be easy to release then!
This would be pretty sweet.
They already have, and it is called AI FTP and Adaptive Training: all you’d have to do is assume you have completed the scheduled workouts and assume some distribution of responses (Easy to Max Effort). The algorithms would then predict an AI FTP.
If ai had to guess, the main reason why TrainerRoad has not released this feature to the public is because it might create false expectations amongst athletes: “TR promised an FTP growth of x, but I only managed y < x.” Or it might discourage people, because they won’t get to 4 W/kg in two months.
I would still like this feature, but only if TR provides insight and actionable advice. For instance, “Your FTP growth is lower than projected, because your adherence is only 70 %. AT recommends to lower your volume slightly and that you improve your recovery.” Setting the correct expectations and providing reasoning would be important.
Honestly, this is the kind of discussion I would hope to see on TR’s podcast, preferably with @Nate_Pearson. They wouldn’t have to reveal anything new, just explain product decisions. Perhaps they have tested versions of this feature and decided against releasing it.