I think if OP renames the thread it will get more input e.g. Share how your latest AIFTP detection compared to your prediction, and why….
For me
New AIFTP +6% vs old AI FTP
Prediction from there: +5% (would be all time high by a margin)
Actual, latest detection -1% (couple of watts / rounding error)
Next prediction: +5% again
Compliance: moderate… missed some workouts due to vacation which cost a few watts prediction, but bigger losses from failed workouts (one ‘over unders’ with unders above old FTP, and one vo2 on-off which was partly my fault) which I guess could also be linked to patchy training on vacation. Today was ok at least partly because the O/U were only 3x9min vs the 3x12min failure… it was also more traditional O/U which I prefer vs the ramps like Carpathian peak
I’m only back to structured training and on 5 hours. Decent compliance with 2 hard workouts classified as Very hard instead of predicted hard resulted in not hitting the predicted FTP.
I wasn’t really affected by the new AIFTP and the training has been pretty good so far. I feel like nothing much has changed for me personally under the hood, TR always dished me out borderline Hard / Very Hard workouts and I don’t progress much if the workouts were very hard.
I am moving into my second month of base on a gran fondo plan and it looks like pretty much the same workouts are being allocated. Again, 15 watt increase in 4 weeks seems absurd but I have some learnings from the first month so if I can increase by half the prediction that’s a result.
Don’t really care about the FTP debate as it was and continues to be just a training number.
I followed the plan almost perfectly. I only sometimes switched some workouts to see how it would affect my prediction and kept the „better“ ones. I guess you could say that the new Ai does work for me.
The workouts were largely fine. One vo2 was awful but I got through it. A threshold I had to turn down the intensity and then I got sick towards the end of the block. So not expected it to rise.
Initial FTP: 207 (but rapidly changing as I ramped up training)
New AI FTP: 218
Prediction: 236 (at the start, honed into ~226 by the third week)
Detection: 226
Next prediction: 236
The workouts picked by the AI have been very good, with highest likelihood predicted RPE always agreeing with my actual RPE. I failed one hard workout early on due to feeling really off that day, but the rest of the time FTP felt accurate, including my first hard workout at my latest FTP (Cloudripper -3). So I’d rate the workout selection pretty good.
My compliance has been mediocre. If I did the workout I nailed it. But I missed several due to life. When I did the work that was planned on the calendar, my FTP dithered by a watt. When I missed key work, it dropped a few watts. And it settled right where it felt right. So I would say the FTP prediction was very good.
I’m pretty satisfied with how it handles TR workouts. I’ll be really eager to see how it handles the unstructured rides. If it does a good job giving me the right workouts with those mixed in, I’ll be really impressed. I don’t think the FTP prediction will handle those well due to not knowing if the group ride will be steady state, full of surges, etc. But if it accurately updates the FTP prediction afterwards, then it will be very valuable feedback to decide how many of those I want to sprinkle in (and when).
Next prediction - “more training data needed. Keep training and your prediction will appear”
.
I thought my original Ai ftp was high and I voiced that in the beta thread, but I accepted it to give it a shot. Turns out every workout was pretty spot on and I’ve been psyched with my progress. I was around a 5 threshold PL, 6 vo2 and a 8 sweet spot, and it dropped me down from those to 4-4-7 or so. Now a month later I’m back at higher PLs in every category, but 20 watts higher compared to the end of December and I’m marking the workouts in the same way (generally hard). With the said, I don’t really understand why it doesn’t give me a higher FTP now and drop my PLs back down like it did before. What’s the difference here? Maybe it wants me to increase TiZ now? Now I gotta start doing level 10+ sweet spots to get a bump?
I also don’t understand why it needs more data on the next prediction. I’ve been with TR for almost 10 years, and I’m probably in the top 0.1% of quantity of uploaded data. I have plan builder laid out until June. I’ve been as consistent as I’ve ever been over the past few months. Not sure what else it could want from me.
I’ve not seen anyone get a new FTP detection or prediction above 400.
I think when you get to 400w you fight the ender dragon?
Seriously though it does seem that there is a maximum FTP of 400w in the new system for some reason and I think that could explain your “more training required” if you have gone out of range?
The chef knew I liked to dine out but I hadn’t visited their restaurant in a while, the first threshold served up was very spicy, had to smother it in cheese. The second a bit more dialed to my taste and then a nice heat to finish. The initial prediction fell quickly and settled in once my spice levels were known. Most of my riding is commuting (1hr each way), a few of these got a bit heated but the system nudged my week around. Compliance, pretty decent (5/6 indoor high intensity sessions completed).
Next up, with commutes swapped in, a downward prediction - I think as only a single 45 minute high intensity session survived. With the help of the prediction I’ve set a realistic plan looking at +3.1%. Aim to nail 1 threshold and 1 long sweet spot indoor session per week.
Old aiftp 295
First Beta onboarding AIFTP 278
Feb 7th new system prediction 285, actual 282.
Mar 7th prediction 289.
In hindsight my FTP was set far too high on the old system. I have been served up much better workouts since joining the beta. Happy enough with how things are going but like some others, I have found it easier to disable/hide the AI FTP prediction and just check in every 4 weeks. I do feel that the workout rating causes too much of a swing in FTP prediction. I have seen +/- 5 Watts even on workouts very close to the end of the prediction window when choosing between hard/very hard.
Interesting that you should say that. I received a prediction from 386 → 400 W on the 10th of Jan. However, because I initially rejected the AIFTP of 386 W and only set the value some days later, TR apparently interpreted it as a manual entry and triggered me to have another AI FTP detection a bit more than a week later, which put me at 389 W. The FTP prediction was therefore also shifted back and is now scheduled for the 17th instead of the 7th, where it was for most beta users, I gather. However, the predicted FTP remained at 400 W. I thought this was because my Threshold workout type changed from over-unders to mixed intervals, which are scheduled with a lower PL (but definitely not lower intensity ), but might be that there is an artificial ceiling. Anyone above 400?
P.S.: We all know that my real FTP is well below 400 and will remain there also after the next FTP detection.
In the last week before my new aiFTP detection my prediction (both the first prediction and the prediction 1 week away from detection) has gone down from 215 to 213 despite 100% compliance.
aiFTP 20th Jan = 210 (down from 231w)
Prediction = 215
Prediction 9th Feb = 215
Prediction 10th Feb = 213 (after comfortably completing VO2 of 5.7) Final week of 4 week block.
Prediction remains at 213
In the final week before detection my predicted FTP decreased to 210 (ie no change from the start of the period) if I removed a sweet spot workout. I find this odd that one workout would have such a profound effect so I was intrigued to see if the predicted value was actually what became the detected value.
Noise and as it gets closer to the actual detection date, as it has more actual data than predicted data, it’ll focus in on the target and get that little more accurate, Well done btw
I’ve been off the bike for 3 years, returning just before the new update went live:
Previous AI-FTP estimate 195 on 28th Dec: (I’m mildly ashamed to say my ego didn’t like this, so I manually adjusted to 200 on 31st ). I did TrainNow or whatever workouts I felt like for a while to get back into the swing of riding indoors.
New estimated/current AI-FTP then predicted at 205 on 17th Jan: when I was accepted onto the Beta, just before general launch. I built an new plan starting w/c 19th Jan and messed about with it for a while.
Predicted** AI-FTP for 14th Feb was around 229 to 236**: This bounced around a bit as I messed around with settings, changed plans, and ran some scenarios etc.. After a week or so of experimenting, I left it as a FTP builder plan for 16 weeks.
Estimated/current** AI-FTP of 241 for the 14th Feb**: This nicely coincided with the end of a recovery week. The predicted AI-FTP remained around 230-235 as I did the workouts, but bumped me up in the last week as I recorded a 90min O/U Threshold workout (Dana) as Moderate, I thought I could do another set without too much trouble and definitely wasn’t clinging on or looking forward to the end. Even though this was a TH workout it really did feel more like SS with ‘overs’ around threshold rather than above it.
New predicted AI-FTP of 270 for 14th March. I guess I’ve got work to do.
So far I’m liking the new changes and they’re working well for me. I can see some oddities with some workouts not hitting the expected zones but I suspect this is is an edge-case while I enjoy those noob-gains.
I’ve completed all workouts fully with most of them being the ones offered by TR with a handful swapped out either due to experimentation or ones that I didn’t like, such as Ericsson with the long pedal drill section at the start.
Original Predicted FTP 358 - This slowly moved down over 28 days. I think I had 2 workouts that I rate Very Hard, 1 of which I had to reduce the intensity.
Current Estimated AI-FTP - 348
New Predicted AI-FTP 351
I am loving the new TR. Has been great since I joined beta. Giving me challenging yet doable workouts. Which is also forcing me to make sure I make my easy days easy, and feel like I am making great improvement. My first Cloudripper workout after my new AI-FTP was moderate. Definitely improved in the last 28 days. And very happy that my new prediction is a very minor increase. Will really help me solidify the gains that I have made so far. I used TR almost 10 years ago and tried again about 18 months ago and was not that impressed. Loving this new version.
New AIFTP (Beta) - 222 Dec 20, 2025 ( I also was on my end of season break)
New AIFTP increase - 233 Jan 25, 2026
AIFTP Prediction - 237 for Feb 22, 2026 (I think the AI will give me 235 from what it’s done in the past)
AIFTP update - 238 Feb 22, 2026, I gained 1 vs the prediction, cool I thought for sure it would have been less.
New AIFTP prediction- 245 for March 22, 2026
So far the workout feel just right for me. Haven’t failed one yet and only rated two workouts Very Hard which dropped my prediction from 242 to 237 and even those where so closes that I had to use the Chad Chart to help me rate them.
284 after the 4 week block, so pretty darn close. I followed this base portion of the plan really well and hit all the hard (z3/z4) workouts planned. If the endurance days were completed on the trainer I was spot on, but if I did them outside, I was typically slightly over on the designed wattages. I also added volume with the outdoor rides. Rest week went well. I just stuck to the plan and didn’t add or subtract from it. I’m typically sitting about the 284 when I am just coming out of winter, so I may be a bit ahead this time. Next block has me estimated finishing at 291. If that’s the case I’ll take it for a mid March number. Never hit 300, so maybe this is the year!
UPDATE: After a 5 week block of HV build (4 weeks on and 1 week recovery) my 291 estimation turned in to a 299 accepted FTP on March 16. The first Vo2 in week one was rough, but I was able to complete all the other hard days spot on. I did 1 Vo2, 1 SS, and 1 threshold ride on the trainer per week. The endurance days were mostly indoors as well, but I did a few outside. To wrap it up in nine weeks with really good compliance I went from 270 to 299. I’m pretty pleased with the progress. I haven’t tested the results, but I am being given and completing workouts than have never been offered up before. My next block will end on May 3 and the prediction is 300, so basically flat. I’m back to a base block, so I wouldn’t expect the FTP to rise much during this time.
My gains seem to be out of this world. I started using TrainerRoad this fall.
“Old” AI FTP: 297w
”New” AI FTP Reset: 315w
1st Prediction: 325w
Highest Prediction: 344w
Ending Prediction/2nd “New” AI FTP Detection: 338w
2nd Prediction: 375w
After first workout with new FTP (Threshold): 378w
I am under 160lbs with some weight to lose so these numbers are significant. I am a former Cat 2 roadie who does MTB and Gravel now. I never did structured training when I was racing on the road. I have done a little bit of it over the last few years but this is the most consistent I’ve ever been.
I made several changes and moved workouts around as I added XCO races to my calendar, so I’m not too concerned with the final FTP being lower than the first prediction. 279w always felt ambitious anyway.
I posted about this on another thread and also reached out to support:
1/22 Predicted AI ATP: 244
2/19 AI FTP: 227
My perceived workout compliance was high, I missed one endurance ride due to work and shortened 2 workouts during my recovery week. I do as many workouts as possible outside.
), but might be that there is an artificial ceiling. Anyone above 400?
). I did TrainNow or whatever workouts I felt like for a while to get back into the swing of riding indoors.