Trusting the AI process experiment

So I’m doing an interesting test of the AI experience. I started a new plan two weeks after Unbound 200. My FTP was around 240-245 heading into Unbound (unfortunately I overstrained early and saw it drop from 260 in February. Part of the problem was I ignored two ramp tests that had lower results.

Anyway, I was still a bit fatigued from Unbound and had a bad bout of diarrhea the morning of my FTP test. So did it with pretty much no fueling. Anyway, I didn’t expect much and didn’t get much…came out at 222. Normally, I would have ignored or retested a day or two later. But I thought, let’s see how AI works with this.

First thing it did is bump all my progression levels up (because I was completing workouts at 242 successfully right before this) and it revised my workouts to harder workouts. So far I have done a 7.5 sweet spot and 4.5 threshold in the first week. The sweet spot I marked as hard and threshold was moderate. I’m going to go through this 5 week block and see how my ramp test goes after. Will be interesting to see how I react to “harder” workouts at an artificially low FTP. Not training for anything in particular right now so thought it would be an interesting experiment.

Will update when I retest or any interesting observations during the block.


I think normal intelligence should be used before AI. Why even think about doing a test under those conditions.


If I were serially racing and had a key event coming up, I wouldn’t. I train and ride and do the occasional gravel race for fun now. So I’m not too invested if my training isn’t perfect.

So when I tested low I wasn’t too bothered. Then I figured what the heck. Let’s see what happens with this 6 week block. That’s all.


Will be interesting to see how this goes but it does raise an obvious concern… you’ll be riding at power numbers that are too low and probably at many times you wont be working in the expected zone at all…

Eg. at 222w, a 95% effort would be at 210w but thats only 85% of your 245w FTP, so tempo and not threshold at all. Likewise 120% vo2 based on 222w FTP would be 266w instead of the target of 294w for a 245w FTP - and in fact would still be in threshold and probably missing the point of the session entirely.

I’m sure you might well get some adaptions but even if the difficulty of the workouts is pushed up you’re probably still going to miss out a lot.

An interesting observation - It appears from what I’ve seen that Adaptive Training only really changes things longitudinally, so to speak. Things are made harder (or easier) by changing time rather than intensity. To me this is a problem for the system that means that it can’t really get around the outlier ramp test results problem at this stage.


I 100% know this won’t be an optimized block. Just more curious what AI does. My sweet spot workout was at 94% for 53 min with only two one minute breaks. So yes it took up the time in “zone” and pushed it to the top end of sweet spot range.

We will see where this goes!

Basically it was close to your real, not ramp test FTP. So your zones will not be so much off.

Basically it was close to your real, not ramp test FTP. So your zones will not be so much off.

Nailed it.