@Michael_Simms sharing my same note with you too:
If you’d like to proceed with opting out just let me know and I can process that for you!
@Michael_Simms sharing my same note with you too:
If you’d like to proceed with opting out just let me know and I can process that for you!
Please opt me out.
@Michael_Simms you should be opted out now! ![]()
Sharing the same note as above: due to caching it may take a few minutes for the change to propagate from our end to your client, but should be visible after logging out and logging back in.
Please opt me out too. Thanks.
Please opt me out as well. I have no use for it. I always know what I’m going to do before i start Zwift. Thanks.
@Adrian_Brewster @Xavier_Neys you should be opted out now as well ![]()
Sharing the same note as above: due to caching it may take a few minutes for the change to propagate from our end to your client, but should be visible after logging out and logging back in.
Already yesterday I wondered how AI assisted workout suggestion could suggest me “The Long Ride” while telling me its about fatigue management. Anyway I rode it, just to find it to suggest the exact same “The Long Ride” workout to me again today. But this time it even says “Today’s 90-minute endurance ride at a relaxed pace will help you […] while managing fatigue”. That doesn’t look right to me. FYI, I tuned the suggestion to ‘longer’ rides if that is of any relevance.
Please opt me out too. Thanks.
I really hope these “heads down” updates and improvements are actually based on the overwhelming asks from the user base instead of something someone in Zwift HQ thinks would be cool to have.
Hi @Jonas_H you should be opted out now as well ![]()
Sharing the same note as above: due to caching it may take a few minutes for the change to propagate from our end to your client, but should be visible after logging out and logging back in.
The paragraph in the “next up” again doesn’t seem to make a lot of sense. My training status is “overreaching”, my training score is going up, but it’s saying again that I’ve been focusing on recovery-intensity.
None of my recent rides would really be classed as recovery rides from what I can tell.
Zwift Ai wants you to quit being a baby and go harder! They are a very mean coach ![]()
![]()
As your coach, I’m prescribing an endurance ride at moderate intensity. It’s called INEOS Grenadiers: Anaerobic Capacity
In fairness, it’s a 52min workout with only 55 stress points, with 20x20secs 140% FTP intervals with 40sec recovery periods between each effort in groups of 10 and 6mins of recovery between each set of 10.
I don’t have visibility into why Zwift’s AI recommendations are flawed, but I suspect the problem is not AI and probably the system needs more AI to do a better job. Any general purpose AI service can give a better recommendation.
I for one, would like to know better how workout categorization works. Not sure it’s just a zwift thing. Here’s some of the top results from What’s On Zwift workout filter, with the only filter selected being “Z2 Endurance” for 60-75 minute durations
I don’t think the fact that whatsonzwift’s filter isn’t very good is really relevant to why Zwift’s recommendations are not very good. Again, a mainstream AI service is better. I don’t agree 100% with what Google AI says about these workouts but it is undeniably better and many people might agree with it.
Fair enough, and there’s much wrong. Sometimes the recommendation’s words might be correct, but then the chosen workout doesn’t match. Or, vice versa – the workout might be suitable, but the description doesn’t align?
This doesn’t necessarily have as much to do with the “Personalized Recommendations” feature, as it does with the preexisting route categorization that Zwift has in place for its workout library.
This workout (“Saw Legs”) resides in Zwift’s own Endurance folder. Maybe the description is a giveaway of what’s going on. Ie. a workout labelled as Endurance, is not saying it’s a workout that is mostly in Zone2, it’s saying it’s a workout designed to develop your Endurance (within the short timeframe that these workouts typically are).
Let the AI label it then. Don’t overrule the AI’s interpretation of the workout with fixed categories, just feed them all to the system and let it decide, because any mainstream AI service can do a reasonable job of that. Not knowing how any of Zwift’s Next Up feature works under the hood, I don’t know what the mistakes are and can only judge it based on the output which in many cases appears to be nonsense in the current iteration.
In the actual description of the workout it says these short sprint efforts should build up a large amount of lactate. It suggests doing this workout if you want to improve anaerobic capacity. It definitely does not sound like the creator of the workout thought it would be “an endurance ridde at a moderate intensity”.
What I find somewhat puzzling is Zwift themselves did work to categorize workouts… One of those categories is “Endurance”… If they are using that terminology in their own AI workout recommendations you think they would start by choosing workouts that were tagged into that endurance category.
In fact, this exact workout is in Zwift’s VO2 Max category. So Zwift themselves have manually categorized it as a VO2 workout, and not an “Endurance ride at a moderate intensity”.