I know nobody wants to read another pissy rant of a user showing themselves the door, so I will keep it short.
When I go into a system and pick a workout, the absolute last thing I want is for a DIFFERENT workout to be immediately synced out to my devices.
I know there is a way to turn that off. I know that I caught it before I made it out the door - resulting in the right substantive workout syncing, just with the wrong name.
I don’t care.
I care about the arrogance that selected, by default, a user option that literally says, “Ignore the thing I JUST PICKED, and substitute it with something I didn’t pick.”
On purpose.
I don’t race any more. I don’t need to be fed a workout plan any more. I’ve stuck with TR because it just works for indoor and outdoor rides I pick.
But not any more.
Plenty of free alternatives out there. I’ll learn new habits.
This doesn’t sound like expected behavior. Picking a specific workout (not to be confused with the the AI workout template types), should pin that specific workout to your calendar and sync that.
I’m sure that’s an issue that could be resolved by TR support.
But if that’s one of many things that put your over the edge, or you’re moving away from plans, I get it. Best of luck to you in your cycling journey.
I never turned on or opted into AI anything. Instead of the ride I picked, something else was put on the calendar “because my training stress is too high”. Instead of what I picked, the alternate synced. I had to go to the calendar and say, “No, i want the workout I picked.” Syncing after was a mess.
I also experienced this (before I turned off fatigue detection) and I also thought it was a very unpleasant and unexpected user experience. In TR’s defense, I think it’s really hard to get UX right (it’s really hard to predict everything about how users will feel in advance)… but I also think they’ve been pretty slow to polish things like this post-release.
Some of it is needing UI/UX tweaks, and some of it is TR going all in on there new AI and making the decision to not allow users to turn off some features.
For example, there is no way to globally disable AI FTP Prediction, even in cases like me where I don’t use TR for planning (I use TrainingPeaks) so all I get on my calendar is the next 28 days highlighted in black with no prediction
I’m trying to be fair to TR, in that they started in with “AI” before the recent mania. That said, until now it’s been something that was easy enough to take or leave. Now it’s literally contravening direct user input. Thanks but no thanks.
New TR feature: AI FFS prediction, a cutting edge tech prediction of when you will give up on TR, based on data from forum posts and device accelerometer from throwing it away in frustration.
I decided to give it a chance but I was also pissed not because it was introduced but the way it was introduced.
Thursday evening, preparing for my 1h SS ride and… yes. Now it’s much harder 1,5h ride. Because, who knows, it is just different now. No way to change it. Looking for “wtf just happended”, found “AI, again, like everywhere, straight into my throat, whatever”. Found that now I have much higher FTP somehow (unrealistic) and also that next 4 weeks I would be dead looking at my calendar. Found how to change “training approach”. Found that I can’t do it. Realized it works only in TR app while I use only Zwift with synced calendar and I even forgot that app exists. I still did my 1h ride as Zwift remembers a few planned trainings (and can’t sync very quick either) and then I could finally found that “FTP is not FTP anymore, it’s just a number for TR”, “look at pure watts” etc. When I had time to look for it. Next day in my job
Come one, I should have had a choice to enable it, accept it, or to get info first. It used to be an app for making time crunched people faster. You just don’t think, don’t adjust, no time for it - trust, train and that’s it, no time for overthinking, time of time crunched cyclist is limited. I don’t want to actively looking for info about incoming changes, watching 1,5h podcasts etc.
Anyway, those changes may be great, may not be great, I don’t care, time will tell. This is just not the proper way to introduce it without making people angry.
To be fair, with such a big change, most apps would walk you through a more significant onboarding process, where they would show you how to do the old things that you’re used to. And maybe give you the option of doing things like you’re used to.
Obviously, TR thinks the new way is so much better that everyone should use it. And for the majority of users they might be right, as far as what will make people fastest. But that assumes:
Users want TR to constantly adapt the plans.
Users want to accept this change right when it launches even if they’re in the middle of a training block
Users are willing and able to learn a bunch of new UI with minimal guidance
Users trust TR enough to create an AI that doesn’t suck
Users are willing to change how they think about FTP and how to set it
Users are willing to invest hours into learning and getting comfortable with the new system
Users are willing to gamble their fitness on such a big change with only TR’s word on it being better, despite lots of different athlete goals and constraints
Users would not be overly obsessed with every minute change to predicted FTP (ROFL!)
Take any one of those assumptions and there’s going to be a fraction of users that don’t agree. Take them all together and you have the current situation.
IMO, TR has been working on this new product for so long, and they understand it so well, including confidential and non-apparent details, that they just can’t see how much anxiety and reluctance they created by rolling this out the way that they did. They were super excited to finally launch what they feel is a massive improvement to the product.
If it was my company, I would have:
Given anyone that wasn’t auto-accepting adaptations an onboarding choice to stay that way
A few onboarding choices for when you get the changes (immediately, at end of training block, at end of training plan).
Walked the user through the UI changes in-app
Given user-specific feedback in-app about why the AI made a significant change and other suggestions for changing plans if they don’t like that particular change
Either rebrand AI FTP to some kind of TR training power or just give a training score.
I would have more help articles and had them linked prominently from the app, especially for FTP prediction (I begged them to do this when I was in the beta)
I would have had users in the (unannounced) Beta testing long enough to show their 4 week predicted vs actual FTP changes to build confidence in the new product
I would only update the predicted FTP once per week. This would reduce the fretting over every little twitch of the number
All that said, TR has financial pressure to launch ASAP. New years is an obvious time to get new/ returning customers. But I would have anticipated a lot of the above issues earlier in development. And I would have held back the FTP prediction until users got used to the other changes. It was too much at once IMO.
I completely agree. The product is good, for sure, but we need time to build trust toward it and there are many assumptions made, as you pointed out, which are not universally true. For example, I’m not willing to let the AI do stuff unsupervised. I train models for a living and I don’t trust them. They can be (and are) better than humans on average, especially highly specialized models, but they lack common sense and they always will. You can never program a model to have the full context that you have, as a decision-making human, the same way that no human can ever process the amount of data the AI can when taking a decision. So for me, I have to spend more and more time keeping track of what the AI is doing (especially as it’s unannounced), to be able to step in when necessary with my common sense.
I don’t know. They can’t seem to count the number of Rs in strawberry
Yeah, the understanding of and trust in “AI” (and distinction between LLM and ML) is all over the map. That alone is going to rightfully cause people to be cautious.
I actually didn’t like AT but really like the new “AI” (ML) system. But I’m very technical and research the heck out of things. And I’m willing to gamble my mediocre fitness. So it was fun for me to try it. I’m sure I’d feel differently if I had critical goals I was risking.
I think it’s been made clear enough that we’re not working with a chat bot or a glorified search engine here (aka LLM).
Interesting comment about the AT. It should be the same workout selector under the hood! The only change is the workout pool available to it due to the different training number (and the fact that it doesn’t ask for permission).
It’s not though. The old AT looked at only the WL and would go up or down based on that. That system was flawed and people called this out constantly. The new system looks at the work and ot at WL at all. The selector is one of the biggest changes.
That system was actually switched off for most people a while ago.
If you check back in the forum 6 or 7 months ago there was a spate of people noticing workout prescriptions that didn’t make sense in the old framework but in hindsight make perfect sense now.
That’s been communicated by TR, on the forum and podcast, but not in the app!. My point is that A) the majority of TR users probably didn’t get that info because they only use the app B) the majority of users that were told did not comprehend what that means.
That’s only been true for the past ~6 months I believe. And nobody knew about the change so their opinions were already formed about AT 5 years ago when it launched.
And I perishable hadn’t used TR much during that 6 months, so I didn’t really experience the change.
