TrainerRoad has the data, it might even have a compliance score but let’s create some anecdata.
Keep it short. Keep it light.
Do post
1. Your starting AI FTP Prediction (%, w or both)
2. What Detection you were served at the end of the 4 weeks (%, w or both)
3. How closely you followed the recipe (make up your own compliance scale)
Do say
I swapped the egg whites for aquafaba
Don’t say
What’s a recovery week? Oh I do those, only 1 Zwift race.
Back in 3 days with my first completed cycle..
Started Feb 2024 - 207W
Feb 2026 - 238W
Peak May 2025 - 242W
Followed plans sporadically, especially during the summer race season, 59kg, getting older
Starting FTP 281
AiFTP Prediction -305
Final Prediction - 284
FTP Detection -284
The prediction not lining up is completely on me. Had a bad first threshold interval due to lack of sleep. Work trip and snowstorm/ice caused me to miss the 2nd threshold, 3rd VO2 workout and long ride.
Was kind of shocked it went up and didn’t stay stagnant or go down.
New FTP Prediction - 310
Know I won’t get there since I’ve got a week long work trip that’s going to throw a wrench into some training. I’ll be happy if I’m 290-295 at the next detection.
I think the intention here is to ask how it’s going with the new AI system (hence the predicted vs actual FTP). What was the predicted FTP and then what did you achieve at the end of the 4 week prediction window. So only people who were in the beta would be at the end of the 28 days now.
Oh, in that case I wish I had a record of the initial AI FTP, unless anybody knows how I can retrieve it ? as that would be interesting.
Been on AI FTP Since September 2025
-edit-
Any of the original Beta AI testers, was the AI FTP number even there on the dashboard at that time ?, I honestly cant remember
Starting FTP 285 (bumped up from 283 that was previous detection 2 days earlier)
Initial Prediction: 303
Final Prediction: 293
Actual Result: 293
I was sick for 3 days. Failed a workout before I noticed I was sick, then I picked a reduced intensity ride when coming back. Other than that pretty much followed plan. 3 hard 1 endurance. Bounced around 293-297 after being sick. Happy with 293.
Old aiftp 306 Sept after 300+ for hour race so “accurate”
Dec new system 264 ( lots of time off )
Jan new system 274 ( not many workouts)
Feb 7 initial 306? Maybe 304?
This seemed valid based on old ftp and historical I could come back for a month and pop back up over 300 if I worked out.
I did not work out.
Instead I subbed corn starch for flour and the muffins came out to 278 predicted for tomorrow.
I did maybe 3 rides so yeah expected result. Lifted, rowed, swam but no trainer rides.
Exactly. The forum is filled with bugs, ftp grievances, some legitimate, some not, etc. and I want to see if there is a silent majority quietly using the product and give a space for quickly posting the outcomes.
Dec 12 AI detection FTP - 254w (this was too low, but I rolled with it)
Dec 30 AI detection FTP - 262w (still low, but trusting it will catch up)
Jan 9th in Beta and got my first AIFTP - 296w, predicted 308 in 28 days (massive jump, def felt too high but got through block. SS was thresh, and over/unders were over/overs
Today, Feb 6th, got new AIFTP - slipped from predicted of 308 to 301w. New predicted in 28 days 312w. The workouts are again more thresh when they should be SS and the over/unders are not gonna hit the energy systems the plan is calling for.
I’m honestly not sure what I’m gonna do. My plan for this block was was originally to manually enter a lower FTP, but I’m also kind of willing to crash and burn for the sake of science. This AIFTP is definitely too high for the intended zones it wants me to do…hence my dilemma.
As far as the predicted AIFTP in 28 days of 312w - that ain’t gonna happen!
For context, I’m turning 51 this year and been riding for decades, so I know what kind of gains are realistic. The system has definitely under predicted me from the start, and has overcompensated and over predicted me now.
This update is interesting as it was a big change and everyone is motivated in different ways.
I like seeing predicted difficulty as it fits my mental model better than stretch, break through, etc. hard is hard as I rate. I also recognize the ftp prediction as noisy but it does give real time feedback.
Thanks for creating the thread for data.
‘New’ AIFTP (Jan 9) = 284
Prediction= 0%
AI Detection (today)= 274
Next prediction= 289
The predication did track down over time. Compliance was 75-80% (missed a workout this week).
I’ve record many years of data:
So far so good, threshold and long vo2 more achievable, as I think my sub 3 min power pushed those levels into time I couldn’t handle.
Old AIFTP 325w
New AIFTP 334w
Initial predicted FTP at end of cycle 349w
Detected FTP at end of cycle 333w 
Now lets do it all again….
It’s not actually as bad as it seems… based on my old AIFTP I have already surpassed my all time best w/kg and I didn’t have the best compliance in the last 28days due to me trying to cram in too much running. And I’m not getting any younger (42).
My legs are tired.
Still unsure wether to reduce my FTP back down or not? I think I preferred doing the less intense but higher level threshold workouts (level 5ish).
I can’t image that I will get close to a 4.8% increase in FTP but maybe I need to follow the plan exactly and not mess about and see what happens? 
I think this survey would be the best way…maybe I should have done an extra main post for it to get more attention:
The last question gives a good summary that majority gets good training.
AIFTP 220W, next day ramp test 221W.
254W
239W
Closely followed including fatigue mgmt recommendations missed some endurance workouts, and a few days unwell the week before D-Day
I had manually set ftp to 243 through recommendations while on the Beta, removed it before executing the detection.
Jan 1 manual FTP setting: 330
Jan 1 old AI FTP detection: 334
Jan 20 new AI FTP detection: 346 (according to Nate 351 was the direct comparison to the old 334 - right at 5%. I lost a bit of fitness by the time I was added to the beta).
Initial prediction: 358
Current prediction: 349 (this drop is primarily due to rounding up on survey responses that could go either way, and I’m totally fine with that).
Initial [new] AI FTP: 205 W
Initial prediction: 209 W
Final prediction: 207 W
Detected FTP: 207 W.
I would rank my compliance as very good. I had a hiccup where I was out of town for a weekend and tried to do a threshold workout the Thurday before my trip (which boosted the estimate to 212 W). I was way too fatigued for that workout and did an easy ride instead (dropped prediction to 204 W). We got back a day early, and I added and successfully finished a threshold workout (bumped estimate back to 207 W).
The prediction waivered between 207 and 208 W over the course of the past 10-14 days, IIRC.
Old AI FTP:258
New AI FTP on entry into beta:250
Predicted FTP on entry into beta: 274
Detected FTP today 7 Feb: 262
New predicted FTP 273
Compliance - generally good (except saturday rides where i preferred less structure). Predicted FTP mostly started falling when I rated one or two workouts RPE harder than it was expecting. Model certainly had me doing more intensity than I would have done mytself and I feel stronger, but run training a few days a week + a tough threshold tough sweet spot per week did take its toll and not sure if its sustainable….
Old AI FTP: 305
New AI FTP on entry into beta: 317
Predicted FTP on entry into beta: 319
Detected FTP Feb 28: 319
New predicted FTP 329
My compliance is a strong point for me. I’m always pretty good about sticking to the plan especially since I don’t mind inside rides and even prefer them some days. It’s not 100% as I do add in longer weekend adventures sometimes.
Also, my weight is trending down as we get into race season so those ftp gains are coming while in a deficit.
I haven’t failed a workout in a long time, beta period included. I’m pretty aware of what I’m capable of, but I have looked at some of the intervals and wondered if it’s would push me too far into the red. So far so good. I’ve been surprised by the software’s ability to prescribe just enough.
Old AI FTP = 255w (4.07w/kg)
20th Jan - New AI FTP = 240w, 3.83w/kg (Prediction 254 w, 4.06w/kg IIRC)
25th Jan = AI FTP recalculated = 247w (3.95w/kg)
15days to the new prediction of 259w (4.14wkg)
I’m 50 years old and on a LV Masters plan with Saturday & Sunday group rides at the weekend. With the UK weather though I’ve substituted the group rides a few time (including today) for TrainNow workouts. Today’s Anaerobic substitution moved my predicted AI FTP up by 2w.
