I commute by bike so it is a no brainer to do my work outs outside for the most part. I have just noticed something when it comes to filling in the survey. It seems to default to an “RPE” based on how it thinks it went for you. A couple of weeks ago I did Diamond Valey Road Race outdoors. Which is a bit messy anyway. Perhaps I could have picked a better route. I had to come off the power during a couple of intervals due to bends etc. It came up preselected as “Max effort”. It only felt “Hard” to me. Then yesterday during an Endurance spin I had a puncture so was stopped for a bit (so much for tubeless… probably out of date sealant). IIRC it preselected “Very Hard”. I just hit easy.
Actually thinking about it I might have noticed the same thing on my Saturday thresholds, indoors…. Right now I am not in a comfortable place on the bike indoors. It usually comes after about the 2k in the saddle outdoors. This normally hits for me in February, but this plan has me doing a lot less time in the saddle than I usually do. So I am still getting stiff/saddle sore on the trainer (yeah it has a complaince system, not the full rocker but good enough and yeah I m augmenting with strength training) During threshold sessions I find I have to pause for a moment and stretch things out. Discomfort rather than effort. I have a vague recollection of it preselecting a high score for me there but can’t be sure. I generally just hit the button and fugedaboutid.
I couldn’t find anything on this in my searches. Probably lousey keywords…
Per our new guidelines, do I have permission to share relevant info from your Calendar to see what may be going on here?
It sounds like maybe your unstructured rides are getting matched with TR Workouts and that’s why you are getting those predicted efforts… Same thing with the Indoors as it reads the training data>>compares it to the workouts power targets and makes the assumption of how it went. So, all these pauses you’re taking affect how the system things you did.
@Caro.Gomez-Villafane Sure. The rides were outside workouts but interrupted. The only unstructured off plan stuff I have done is 20 minute Z1 commutes.
While you’re at it can you explain what’s going on with my predicted AIFTP. A couple of weeks ago it was predicting 194→199. Then last week it went down to 196 after that interrupted Thursday ride it went back to 198. Today after my B race and 1 workout left it is down to 195. [update] this morning it has gone back up to 196!
The B race did not go as hoped. As I mentioned elsewhere the AI gave me no long endurance rides. When I added them in myself, it pulled them for rest days. It was rough on the bike after 2.5-3 hours. I couldn’t get comfortable and in the zone. By then I was already 5 on the Garmin TE.
@simonicusfacilis, as suspected, your race and endurance ride file were matched to TR Workouts, so it makes sense it trying to rate the effort between your power data and what’s being prescribed.
As far as your FTP question goes, I would say that is the AI FTP Prediction updating as you continue to complete workouts, races or group rides. I see that from the FTP recalibration of 175 on January 29, 2026 your FTP has consistlently been going up!:
194 February 26, 2026 196 March 25, 2026
Note: plus you’ve been rating your interval workouts as Hard, which means you’re getting served the right intensity.
LOL.. those two were structured outdoor rides… I guess my capacity to control my power outside is rubbish…
As regards to your statement that my FTP has increased under the new regime, I respectfully disagree. Usually in an off season my FTP hovers around the 200 mark. Old boy that I am I have made peace with it declining. My old FTP coming into 2026 was pegged at 192, not incoherent. AI Big Bang dropped it to 175, weird but ok. The fact that it went right back up to 194 by Feb 26th suggests that the 175 number was just in error; it’s a rather unbelievable jump. Especially considered the very constrained hours the plan was giving me. Then today my new AIFTP is 195. It is basically a wash for the last three months, 192-195. Actually it was 193 last night. As I said the predicted value was jumping all over the place. It often did so without any additional workouts; not exactly confidence inspiring. If I couple that with the “fail”on my B-race last week then the only conclusions I can draw are that either this AIFTP/plan builder isn’t working for me (for what ever reason; lack of time/data) or I have hit my physiological straps. Doesn’t really matter which it is, the way forward is the same. Just ride my bike and forget TR plan builder.
FWIW: Regarding AI interfacing with training; the various AI tools on offer to interface intervals.icu with your LLM of choice, have been giving me much more coherent plans. They are focused on getting my TTE out there for my long rides. Putting in more realistic time in the saddle numbers.
Now I am just going to ride the damn bike, do what specific prep I can for the A-race next month in the time left (a few threshold runs at expected climbing times). After that I will do a few two week camps with the workout selector. Then, in November see what TR plan builder comes up with for the off season. I’ll see if it has stopped hallucinating that it is getting me faster.
Hey @simonicusfacilis, the thing with the new AI FTP Detection, is that we’re comparing apples to oranges if you compare it to past FTPs.
The new AI FTP Detection is not predicting the result of a ramp test. Instead, the updated AI FTP Detection you’re getting puts you around a level 3 threshold level.
I am totally aware of the differences between the two ai FTP calculations. I mentioned the change in the value only for context. Mine dropped significantly (-9%). Then in the next round rose just as significantly (+11%). This is not a credible change: I am neither a noob nor was I just back on the bike. Ergo it looks like that original AIFTP was in error. I feel that any supposed increase in my AIFTP is within the error band of my power meters/ HRM. There is no substative difference between 193 and 195. All external metrics agree that there has been no substantive improvement in three months. Average speed is the same, RPE/power is still the same and my cardiac decoupling for long rides is still a long way north of 5%. My 45 day power curve remains quite below that for the last season. A season that was essentially a fallow one due to injury and work travel. Ergo my training plan is failing. As I said upthread. If this is my limit now then there is no point in proceeding with a plan. I should just ride and get my TTE up. Then come back for the off season.
I also notice that my original issue remains unaddressed: the preloading of post work out surveys. For example yesterday I did my VO2 workout outside. Given that you had identified my last VO2 effort as an unstructured ride, I went with an alternative. Instead of the original “Attacking” selection I went with a simple interval. The choppy nature of the attacking version is difficult to manage outside. In addition I selected a straight piece of road for the interval. There was a minor issue with this selection. It was just long enough and the wind was right down it. Hence I had to extend the time between intervals to ensure road position. I found the intervals tough but doable and on one ocassion I ran out of road about 20 seconds before the end. When I got home All out was preselected on the workout survey. I rated it as very hard. This was mainly due to it being a more physically demanding day at work than usual. I felt I could have done another but it would have been ragged.
Now yesterday morning the AIFTP prediction was for 195 → 200 in 23days. After that outside VO2 session it is now saying 195 → 195. As an experiment I changed the survey response to hard and I got 196.
Perhaps AIFTP can only work on indoor efforts. This is not for me. I use TR to get Faster outside, not inside.
The RPE pre-selection is based on your power data, and outdoor files with interruptions like you’re mentioning can skew that reading. Overriding it manually like you did is the right move.
On the AI FTP swings we’ve identified an issue with how our model projects planned outside workouts. If the entire plan is set up as outside workouts, this affects your predictions more than most. Our team has a fix ready for this, so predictions for outdoor plans should be more accurate from the start going forward.
I see that your plan normally had 2 indoor workouts and 2 outdoor workouts, however, you’ve changed all of them within detections to outside, so the above issue with the AI FTP Prediction above may well apply here.
