I’ve been giving a little thought to the fact that I seem to be improving at a faster rate than I have in previous off seasons. Whereas last year my FTP history showed that I’d pick up between 0-4 watts after a 4-week block of training, this past couple of months using AI are showing 9 watts last 4-week block and expecting 8 watts this one (almost there, going by the FTP prediction). That’s a pretty significant uptick and I wondered about it.
I suspect it’s not because AI is magic, and it’s not some feel-good nonsense untethered from reality because I’m actually getting stronger on the bike. After all, you can feel one way or another about the AI, but the power meter on my bike doesn’t know anything about it and my power levels are up. AI FTP vs. ‘real’ FTP vs. whatever, if you are measuring it using a power meter and the reference is climbing, so is your actual power.
I think it’s because the AI has actually pointed out to me how significantly a brief break during an interval or a missed interval affects my progress, and has helped me avoid those breaks. On one workout in the middle of last month I had to jump off my bike shortly after starting to help my wife with a mechanical on her bike/trainer. After the workout, I got a message after the workout “3:34 of pauses lowered your score.” For me, that was a wake-up call. I remembered all the times last year when I’d have to stop for 30 seconds during a hard interval before I could finish it, or turn down that final interval to be able to complete the workout. I always thought “Ah, that won’t make much difference”. Well, turns out not to be the case.
And that seems to be the reason behind my accelerated improvements lately. I can look back on previous blocks and see where pauses in my intervals have negatively affected my gains for that block. Contrariwise, knowing that a pause will drop my gains for the block motivates me to dig a little deeper and finish an interval strong when my legs are…less enthused. And the AI seems (for me at least) to do a good job of picking workouts that are challenging but not beyond what I’m capable of on the day of the workout.
In the end, it’s better to complete a workout properly than it is to do a harder workout that you struggle to complete with good form. I think the purpose of the AI is to pick out exactly those workouts that you can complete well (given your fitness and freshness at the time), and let you know when you’re likely to be trying something that has a low probability of going well.
As always, YMMV, outliers exist and you are free to do whatever you want.
I am in a similar boat in regards to quicker improvement gain, but for a different reason.
This is my third year training with TR. Normally I train from January till October and then take November and December off.
For 2024 and 2025 I went from 245w to around 265w both years (20w total) with the max gain per month of around 5w but with many months either no gain or 2~2w despite a high and consistent TSS which was slow but OK.
Looking back at last year, I was doing more TSS than TR prescribed as I was riding outside whenever possible on the MTB.
BUT I was only doing 1-2 of the prescribed 3 hard interval workouts per week (the dark green of the bar chart) with much of this the long weekend endurance ride workout and not a TR interval workout.
I never understood the true impact of missing the key workouts and subbing in outside rides.
When the new AI came out and it said “do this and I think you will gain 18w in one month” It was super motivating as 18w took me around 10 months to get last year! so I locked in, Did the three hard interval workouts per week exactly, and got a 15w bump in one month, despite doing less TSS.
But maybe this was just catching up on 4 month ago.
I’m now half way to the next AiFTP and its still predicting another 18w gain to 278w which is again 10 months of 2025 progress in 1 month!
I cant wait to see what another few months of dedicated training can achive.
I think for me the improvement comes from fuelling/resting better.
There’s one main reason why I’m doing that…
The predicted RPE survey. If it says something is 50/50 hard or very hard, I’ll put the extra effort in to give myself the best chance of being able to answer “hard" - honestly.
So if I’ve a hard workout, I take not overdoing it the day before a lot more seriously, and getting carbs in at the right time on the day a lot more seriously.
Another factor is the lofty AI predictions/upgrades. My first bump was from 289 up to 302, when I’ve been hovering around 290 for a long time. Again, this made me rest/fuel better for the hard days.
Next bump due in two weeks currently says 317, bring it on
[FWIW I have it set to “Demanding” and click “Check volume” every couple of weeks]
This is me as well. I want to squeeze in the outside MTB fun rides. Even when I go on the E-bike I tend to be balls to the wall 90% of the time…
So far I’ve mostly resisted, more than normal, but as the weather gets better I’ll have to face up to the reality that riding MTB in the real world outside IS better/more important than chasing watts on a turbo inside.
I’ll try to keep it easy on the ups though, and avoid missing the hard workouts as much as I can.
Additionally I’m much less inclined to miss a workout now that I see the numerical wattage drop that goes with that! I had to miss one yesterday for the first time in a while and seeing the prediction drop by 2W was hard to stomach
I will admit to being skeptical initially and not sure how this new system would help me. Lo and behold it has changed everything for the better. Seeing the effects of moving workouts around or changing them - making them longer in particular - is motivating me to stick closer to the plan and keeps me honest. I also kind of love NOT having the tension of the old FTP detection “reveal” every 28 days. You click the button and it’s a number you don’t like. Fuck. A month wasted. This way, you see where you’re headed and when you get there and see your new prediction, if you don’t like it you have a month to adjust for it. Adjust your training, make your best effort to get where you want to go. It’s freaking brilliant. And regardless of what “FTP” even means any more, my watts are going up and you just can’t argue with that.
And the AI seems (for me at least) to do a good job of picking workouts that are challenging but not beyond what I’m capable of on the day of the workout.
- is motivating me to stick closer to the plan and keeps me honest. I also kind of love NOT having the tension of the old FTP detection “reveal” every 28 days. You click the button and it’s a number you don’t like. Fuck. A month wasted. This way, you see where you’re headed and when you get there and see your new prediction, if you don’t like it you have a month to adjust for it. Adjust your training, make your best effort to get where you want to go. It’s freaking brilliant. And regardless of what “FTP” even means any more, my watts are going up and you just can’t argue with that. 
