Always a fan of new features.
I’ve just done my AI FTP detection thing. Oddly it comes out at the same as 3 other sites (actually they are all within 1w) that I use to log my data.
I’ll keep using it but on first test it doesn’t seem to be drastically different to other sites. Maybe my data suits a particular type of algorithm??
Fairly obviously you’d expect it to be in the same ballpark as other sites, right? If it was 30 watts higher or lower that would be weird. 
I just expected it to be something groundbreaking.
So I find Train Now’s suggesting to be fairly varied yesterday it recommended Endurance… I haven’t done a workout other than Endurance in a few weeks as I’m in Trad Base I, anyways, for 1 hour it recommended Bays, but for 45 mins it recommended West Vidette (3.4), seemed like a huge gap, with west vidette being way below my PL and Bays being 5.9 (I was at 5.6). I did Bays successfully (take that AI FTP trying to drop me at the start of this plan) but was surprised how different the recommended workouts were.
Taking today off but it recommends Endurance work again, 60 min Petit (3.2), 90 mins Bays +1 (6.8). Petit seems reasonable but could likely manage a little more, maybe I could pull off Bays +1 but thats a PL jump of 0.9 from yesterday when I jumped 0.3… Seems off.
I have a set of scales that weigh 10 lbs under to sell you. 
Yes! Now I can compete on Zwift 
@ggarcea I had this same problem. Probably a bad idea, but I just deleted the ramp test entirely and did a train now workout and kept my FTP as it was.
I thought my PM was off. On the high side. 3 or 4 weeks in a row. I did calibrate my PM. I wanted to see what Ai FTP would do.
It was handled like a champ.
I did put up some Serious PRs on Strava. So, perhaps not off for those rides. Just fresh. I haven’t put those numbers up again. Lots of PRs downhill :(.
Nonetheless, my next workouts were incrementally harder. My FTP only went up by a few at the end of training block.
Between the adaptive training and the Ai logic the result
For my next FTP was solid.
I’m impressed with the team. Specifically the data science team and Data engineers behind the two features. Of course Product as well. Ok. Everyone. Pretty sweet.
I just remade my plan (in final week of ssb2 MV) and I notice there is no ramp test for my first part of build, just the “regular” workout for Tuesday, my first ramp test is 5th week of build (to be clear I did not click on the AI FTP detection button of a scheduled ramp test, there never was a ramp test scheduled )
Looking into this for you now, @ggarcea! 
Hey @ggarcea, it looks like you were suggested Adaptations on Saturday which included adapting the Ramp Test to Colosseum -4. Did you see this?
Hi, I accepted all adaptations, without really looking into the details of each one.
If I’m done with a hard 4 week smashing block and I’m looking at the next 6 days of easy z1/z2 recovery week rides, can I click the AI FTP button or does completion of the recovery rides also add to the calculation somehow?
Am impatient. Want to click now but also want to use as intended
It doesn’t matter either way.
I just stumbled on this official answer ( from @IvyAudrain), basically it shouldn’t matter…
Mine is 30 watts different.
Which one’s right? 
Xert has proven itself to be right on over and over. Recent Zwift ramp test confirmed that once again. It’s been a while since I did a full Sufferfest 4D test (mid-September) but when I did that not too long ago, it was in agreement also. Zwift and Xert are pretty recent and they agree, the TrainerRoad number seems wildly optimistic.
My Xert TP is absolutely spot on.
and does the new TrainerRoad tool agree with your Xert?
