NO…….. Why did It think I could.
I know you didn’t mean me and I’d never expect that. But isn’t it funny that what is so obvious for us … shouldn’t the smart AI know also 
. I’d prefer the AI prediction be more conservative and therefore surprise me more often with an uptick.
I’ve lost track of when the comment was made exactly, (past few weeks) or whether it was posted in the beta thread or not, but I recall Nate commenting that pauses occurring specifically during the warmup or cooldown of a workout were going to be “downgraded” in terms of their impact on the AI’s overall assessment of workout performance. This might be relevant to some of your issue here…
I don’t know if this update is still planned, work-in-progress or has now been implemented… Maybe @eddie might comment if he has any further info?
Since this thread is getting more TR staff attention - can I point to my experience, which is described in the following thread? How to Use the New AI FTP Prediction (Without Letting It Drive You Crazy)
TLDR - in just 9 days after AI FTP detection my prediction went from +5% to +2.6%, while following the plan and completing the workouts (my RPE matching the predictions).
I thought they fixed that. I had the same problem but on a shorter workout it was only one hour. I forgot to turn the fan on. So I stopped during the very first warm up and then continued. I was also about 2 minutes in. I was almost going to discard the workout and start it over fearing that it would do the same thing. But I noticed that it didn’t this time. It did last week when I was doing a SS workout and my laces came undone. I stopped and tied them and started up again. That one did give me an explanation saying that the pause changed the outcome. I know it did becasue even with that brief pause I noticed my HR recovered and the rest of that interval felt a little easier.
This sounds good. I’ll give it a go. Thanks
I can’t see why the system can’t accept a manual FTP figure. I don’t want to do a ramp test for medical reasons but I for sure already know my actual FTP. I think it is pretty shoddy that this feature has not been built in.
How do you get the strength workouts in. From Strava or by entering them manually?
Did you rate it harder than it predicted you would? i.e. it guessed you would rate it “hard” and you rated it “very hard”
Was forecast that it would be moderate and I put it was hard.
Is there something about getting into VO2 range that gives you medical concerns? If so I wonder if you could substitute a different testing protocol to get a more realistic number, maybe @eddie could answer that.
As for why a manual entry doesn’t work, AI FTP doesn’t look at the number to prescribe workouts. It simply looks at raw watts from the work you have done in the past. If those workout in the past have bad data the only thing you can do is start giving it good data. One suggestion I had was to increase the intensity of the workouts it provides until you are working at a power output that seems reasonable for your fitness in order to start giving it the data it needs.
I don’t know of a better way to feed the system data without testing FTP by one method or another.
I am leaving the FTP People’s Front, and joining the People’s Front for FTP.
Is there a plan to roll these AI plans to Triathlon as well?
Isn’t that the problem though? Why did it tell him he was?
I get that - but at least it’s getting more accurate the closer to the prediction date 
Splitter,
I am slightly confused/bemused by my AIFTP. At the start of the roll out with something like 18 workouts to go and 25 days my predicted FTP increase was + 3.5% after a -5% change at the start. I have completed all 10 workouts with one fail, now my predicted FTP is ±/- 0%.
How can that happen?
Maybe the AI model is also being trained on forum posts. 
You’re right of course 
But as long as it’s getting more sensible with more data I’m happy - thats the point I was trying to make.
To be honest I saw similar - unrealistically high gains predicted at first which settles into a consistent realistic prediction after a couple of workouts.
Maybe it would be better if TR didn’t try and give a prediction until the athlete had a couple of workouts with their new FTP under their belt? Who knows?
Maybe I’m just comfortable with stats? I’m not going to stop using weather reports because the predicted weather on a day 28days in the future keeps changing.