They are working on getting the ML up to speed on classifying outdoor rides where the desired outcome is unknown, but that they are not there yet. So it doesnāt use it now, but will 100% use it in the future.
3 Likes
He seems to hold the company to a high standard. Has to make you feel good to give them money for a service they provide.
4 Likes
I think there are some assumptions that arenāt true here. If the system wants to make a prediction of the outcome prior to the workout, then yes it needs a āplannedā workout. That lets it makes guesses on a specific progression.
To make predictions on adaptations from an effort it shouldnāt need what the plan was. The challenge would be that it canāt pool a ton of similar efforts to get a big enough corpus to make predictions on. Would have to play with abunch of signals like TiZ, length and number of segments in a particular power range, work done prior to the individual efforts, etc.
They claim to be close on it. If theyāve gotten as far as they claim they have I donāt see why they couldnāt bridge that last little gap. Interested to see if it is effective.
No doubt Iāve missed this somewhere but what do the numbers after the workout type represent (ie. Threshold or VO2 max 9.8) & are those individualized?
Youāre heading down a tangent from the original comments which was about an IPO. Not all stockholders are the same and a closely held private company is very different to a public one.
4 Likes
The vo2max 9.8 is the classification of the workout difficulty, if I understand correctly, so that isnāt individualized. So out of 10, TR must have thought that workout was damn near the hardest type of vo2 workout one can do?
1 Like
Cool, thanks for the info!
So mostly just doing achievable, slacker 
Now we can see how lazy people are 
2 Likes
Does Adaptive Training also make changes to the overall training plan? There was some discussion about what happens during time off (travel/illness) in terms of adjusting your levels, but Iām wondering if it also re-shuffles what type of workouts you do (e.g., go back to base rather than picking up in the latter part of build, where you left off).
Very very cool stuff, and driven by what will help users.
True. But, for example, zwift have very aggressive investors. So they do stuff based on what investors wanted. And for a period of time they were going nowhere. So, IPO, or private investors, itās similarā¦
Tesla is public and elon musk doesnāt give a shit about investor feeling.
So not about the investors , but the ceo. And I think @Nate_Pearson is a very generous ceo and he seems to care about their users and their employees. So thanks nate for caring about your product
4 Likes
EDIT: this pricing āsplitā I had in mind is under review. Not sure what we will do yet. I see pros/cons to having it hear as well as separated.
Splitting it will be a royal pain, and is not clean cut since there is overlapping discussion on AT mixed with pricing. āEasyā is to leave everything as-is, so that may be the solution, and it makes sense in ways as well. LOL, me being indecisive as usual. 
I am going to pull the āpricingā stuff into a separate thread (as I alluded to in the part you didnāt quote).
But it takes time to sift and pull those out, and mixing that with my āreal jobā is a challenge, but it will happen.
4 Likes
It seems so. Amberās analogy of how the weather forecast changes seems to explain how the workouts change over time.
I think this is what they have right now. Probably the rest will be done in a near featureā¦ But is hard to understand what was the intention of the ride. Not all .71IF rides are the sameā¦ 
Anyone on the list been given access yet???
Hey TR,
Awesome feature! A few things Iād like to point out that I think you missed however.
-
As an older rider I always wondered why this feature wasnāt already present. It just made sense to have on day one. What took so long? I mean itās just code! Sheesh.
-
Also, as an older rider I donāt like change. Disc brakes? No! Carbon frames? No! Lycra that isnāt completely worn and doesnāt smell like brut 33? Nope! Now that the feature is here what do I complain about now? Iāve settled into my Base Build Speciality and I am not changing.
-
Last but not the least of my complaints whatās my excuse when I donāt hit a workout? It seems you thought of everything and for that Iām truly hurt.
Guess Iāll take my old arse over to Zuffer Fest or whatever they call it.
In all seriousness, well done TR team. Canāt wait to try the new cycling AI.
8 Likes
Is @mcneese.chad the only person posting here who is confirmed to actually HAVE the early access so far?
Would be eager to have others post if they get invite codes.
2 Likes
It sounds like thatās what the workout classifier addresses directly, if I understand Nateās comments correctly in the podcast. It recognizes that IF and TSS arenāt the end-all-be-all of workout difficulty and classifies them even further. I assume this is based on user completion rates and a bunch of other āfeaturesā of which Nate said the ML evaluates more than 100 (all of which are proprietary).
I wonāt name names, but there are several more that I know for sure are already added into the beta as of yesterday.
3 Likes
Iām still waiting to be added.
Fear not! Now you can simply say āI was feeding the algorithm with a failed workout ā¦ for science!ā
5 Likes