I don’t think this is true based on my understanding of what it’s trying to do. It will offer workouts based on you performance on previous TR workouts but as long as they are a decent representation of your overall performance it should be fine.
It might fall down if you only do endurance on TR and Vo2max on other platforms - as TR wont know about your v02max performance - but I don’t think this makes it useless at all.
They will, it’s just much harder to categorise outdoor rides so it’s not in the first release. E.g. I have a local loop I use for outdoor VO2 training which incorporates a ~20 mile section containing 6 decent climbs that typically take between 3-7 minutes each if I’m going hard. I do the climbs as VO2 intervals, descents as recovery, flats at a comfortable Z2 pace. It’s a fun ride and I accumulate 25-30 minutes of time at VO2 so it’s a good workout. But I have no idea how TR categorises it given that the interval length varies, the recovery length varies, and the interval intensity varies since I adjust my watts to the length of the climb. TR workouts? Much easier. Nail a VO2 session with 2 minute intervals and 3 minute recoveries, and it’s reasonable to expect that you could step up to a workout with 2.5 or 3 minute intervals and same recovery time. Or do the same intervals at higher intensity. Or with reduced recovery.
After riding Sunday I didn’t have a work out on Monday but I looked at the suggested rides with TrainNow wondering how different it would be to my scheduled ride the next day. On Tuesday TrainNow offered me 3 different rides to those suggested on the Monday and none were the same as the one scheduled as per plan builder.
I’m pretty sure I have seen Nate suggest that TrainNow works like the AI Beta without the automation, workout levels and questionnaires at the end of sessions.
I’m finding the feature somewhat interesting but the fact that it’s giving me harder rides for the challenging section than for the above FTP ones doesn’t seem to make sense if we look at it from the “achievable/stretch/breakthrough” perspective
Also after hitting shuffle 10x I still have Bays appearing in the easy spot
Nate has confirmed that they are working on including the non TR rides but these are more difficult to catagorise without a target wattage to compare the ride against. I’m not sure that this will be available ‘soon’ but guess it depends on how you define soon. I’m hoping to be allowed into the Beta soon 
Yes, every time a TR inside/outside workout gets updated that could potentially change TN.
And in the future, any workout could potentially change TN.
It does use your TR inside/outside workout history.
The more near term the better. If you haven’t trained for a while it will have little to no impact.
@Cleanneon98 are you expecting to see an achievable, stretch & breakthrough workout?
If so then I don’t think that it is working that way within TrainNow. Basically it is offering an easy z2, V02 and Threshold workout at whatever level the AI deems appropriate based on your historic TR performance.
From Monday I believe they are hoping for the workout levels to be shown to everyone so that you can start to see whether the proposed workout is a breakout session or not and what ‘level’ you are at in the various progressions within each zone.
We have some ML systems that look at all rides right now like FTP prediction. That doesn’t feed into TrainNow workout choice at the moment…but the plan IS to have it feed in once we can better describe outside unstructured rides to ML.
That part is underway (for about 5 months) and we’ve got some prototypes with it.
I have used this feature a few times now because my current schedule prevents me from sticking to a plan. It appears to be working well for me. I really like the fact of seeing more of the workout catalog. Each suggested workout has been achievable. Thanks TR team.
I don’t know about that part, seeing as the most I’ve completed is a .92IF it sure is giving me a lot of .95 to do. I don’t think AI is tied into that system, YET. Just thought it was a bit odd but I know it’s a work in progress
Are you sure your FTP is set correctly? Your results would to me suggest the trainnow thinks that your recent threshold efforts have been easy so wondering if your FTP is set a bit low and that’s why you are only seeing hard threshold rides? 
I would be interested to know if anyone had attempted a recommended TrainNow session that wasn’t acheivable assuming they weren’t on a ‘bad day’.
Pretty sure it’s spot on, but not having completed a single workout over .92 successfully I don’t know how or why it would recommend a .95 I think right now it’s more of a random workout generator that’s not taking individual success/failure into account
Without the “how did it feel” questionnaire I guess we could see some anomalies.
I regard any sort of pause, backpedal, or reduction in % a fail but perhaps the AI is a more forgiving beast!
If I recall correctly, TrainNow will recommend achievable and progressive workouts for each category, but not stretch.
I’m in the same boat. My work schedule is completely random at the moment and my hours are long but TrainNow is keeping me in the game.
I’m still seeing some wildly oscillating behavior in TrainNow recommendations even after populating some TR inside and outside workouts, all done with good compliance I think.
First time I looked at it, they were all absurdly easy - then I seeded one TR indoor workout (Mills) and the next day they were all solid progressive/achievable ideas. I picked Bays w/sprints.
Then did Carp Peak last week in Gen Build, towards high side of targets - no issues.
Now, TrainNow recommends nothing above IF 0.82 - most lower - for the climbing/threshold selections, while my plan has Fang Mtn +1 (IF 0.85) and I’m thinking of jumping ahead to a slightly harder one.
When I saw that all the selections got easy yesterday, I thought maybe it thinks I did too much volume with other stuff the prior 4 days in a row - spanish needle, antelope +5, baird +1 and some hard sprints on an easy ride - but all of those workouts were passed without problems, and yesterday I took the day off - and still, the suggestions are clearly not progressive at all.
Does TrainNow bias the choice based on its perception of current TSS/fatigue ?
If not, why might it be oscillating so wildly like this now that I’ve got quite a few TR inside/outside populated?
I am interested in how well it accounts for fatigue, same as AT. I think it would be pretty hard to implement?
