For sure - I would say that has been my experience also. The old system FTP would have put me in the 40m range, and the training worked quite well there for a range of energy systems from z2 to vo2max. The new number just doesn’t work well for that same range if I’m being comletely honest.
Thanks, I’d like to hear what Jonathan meant from the podcast.
Would be great to dig into this on a Podcast!
Well worth remembering that Coggan’s cohort was well-trained cyclists.
If his cohort (hypothetically) had been less-well-trained cyclists, it’s likely that “around an hour” would’ve been “around 40 minutes” from the get-go.
There’s nothing important or fundamental about 3,600 seconds. Or, indeed 1 second.
Typical high FTP 210 -235
Jan 3rd 192w
Jan 29th (AI day) 175 -9%
Feb 26th 192 +11% (LOL)
B Race in March 21st - Bad
Mar 23 195w +1%
A race Apr 18th .. Not great, not as bad as feared
Apr 26 191w -2%
Compliance:
Started out 2 a week indoors on balanced plan. Fairly compliant. Odd lapse due to travel and sniffles. When the weather/daylight got good went all in out doors. Respected the plan ride time limits (weekly hours). The B race was bad: Not enough time in the saddle. Augmented the plan with more edurance outdoors to approach time in the saddle for A-race. It killed FTP progression. Made the A-race more managable and improved comfort on bike. I noted that the ride scores on the indoor rides were compromised by stopping to stretch more. This became less of an issue with time in the saddle.
The predictions were all over the place day to day. Outside training issue.
Take home: TR only really works now for time constrained and indoors(core demographic) or guys who can hold on target power far better than I can outdoors. So really winter only for me.
AI FTP came in yesterday 25 April at 257W, up from 251W on 25 Mar.
At one point my Prediction was 272W. Strongly believe I could have hit around 265-272 BUT I over did it with adding (a) a sportive [and went too hard on it] and (b) adding in weight training which I knew was risky. I’ve subbed a lot of recent workouts for Z1/2 to try to salvage my knee ahead of a big trip in 10 days time. Not happy at myself for getting injured but lessons learned:
Not worried about my 25 May prediction (265 currently) because I just want to hold fitness through this goal event next weekend and recover my knee.
After mid May assuming OK recovery, it’ll be strength + FTP builder plan to try to reach 4 Wkg (280-300W depending how weight goes) by September.
And starting to use the Energy Availability model for nutrition to try to both build and lean out.
Nice job on holding out and welcome to the AIFTP recovery week club.
Something about doing that last hard workout then only having a few endurance rides before detection day just feels nice.
Then your legs are about rested for when the first hard workout with the power increase hits.
Hmm. I had hoped prediction would get more steady. Not so.
Recovery week Sunday prediction 264. Five days later Friday VO2 max completed and returns 257.
Dec FTP 258 after injury.
First Ai FTP adjustment 249, prediction was 300 and reality was 300.
Second Ai Predicting 345, reality 333, Missed some days.
Third Ai prediction 380, reality 339, Missed more days.
Upcoming Ai prediction 388, fingers crossed.
The workouts have been good. When starting some workouts they felt not achievable but get through.
Has this worked itself out? I feel like this shouldn’t have happened to you. 
I’ve switched to a build Endurance and masters, so I’m getting less intensity and it predicted a higher FTP m, 263 at first. After two solo/commute rides and one indoor vo2 it’s dropped to 257. Seems to be a downward trajectory again.
After adjusting my saddle two weeks back I’m not struggling with threshold, unduly. And I’m eating a huge about more according to the EA principles - that has impacted my sleep a bit.
Battery issues with my pedals outside but apparently not a concern.
I wouldn’t say I have a perfect training record by any means, but broadly I’m getting it done. A month of clean blue graphs would help, I suspect. 
But club rides are calling…
Update:
I ended the block with my highest TR detected FTP of 326. It was 333-339 for a good portion of the last four weeks but the last v02 workout was Hard so it dropped. ¯\_(ツ)_/¯
Update: Started this month at a 274W FTP with a prediction of 291W. I ended the month at 280W with a prediction of 286W next month.
I really struggled with a couple of VO2 workouts this month, and I think a very strong newbie gain curve is flattening off, which is good. Its also gotten warmer here, and I needed to figure out the shift in cooling. I’ve also taken another look at fueling after the podcast two weeks ago. Hoping to see the 291 go up, but the training is starting to sharpen towards a Tri in August and a lot of half marathon in October.
Predicted: +4%
Reality: + 1.1%
First block on 3 hard workouts (2 previously), a little illness in the middle, a shade disappointed but there are improvements to make on my end. Back to back hardest workouts at the weekend, not great.
Next up, first build block and +2.5% feels more ball park.
At this point, it would be nice to see AIFTP back calculated a few seasons, but on the other hand, it is sunnier here and I should go and play outside and see 
A month of clean blue graphs would help, I suspect.
I totally get that. 
Has that “Limited Data” warning gone away, though?
Only saw it on the detection notification itself, so don’t know where to look to see if it’s still there.
You’d see it on your career or calendar.
Okay, cool. That’s what I was looking for. 
Let me know if you see that pop up again or have any other issues.
Bit of an update. Sweet spot and theshold workouts began to get too hard, so I had to revert my FTP back to the initial AI FTP detection. I also had to tone down the endurance progression, as it had me doing close to 75% of FTP for extended durations, which was completely achievable, but left my legs too fatigued to complete the interval session the next day (which is the more important workout in regards to adaptation).
I would really like to see some fixes regarding AI FTP detection and the progression of plans for athletes with longer training history. It seems at the moment the model is most effective for those that don’t have much training history, and can therefore progress really quickly.
