Thanks for your reply.
Through all of this, even though I don’t agree with 100% of TR (and with some vocality), I will say that @Nate_Pearson definitely applies and dedicates himself as positively, and personably, as possible to the company, the employees, and the customers. Hard workin’ dude! 
I just need to be prepared for the “Not recommended” or “Not achievable” message from the AT system 
I have some questions about how the system works in terms of overall training load. I am a relatively new TR user. Last summer I was coming off with a good ‘base’ and went into a short-power build block. While I could have handled the TSS load of the training alone, I was also trying to ride outdoors with friends so I either added in days and only replaced 1-2 TR sessions with outdoor rides. This was 700-900 TSS/week, 6-7 rides/week, and after about 5-6 weeks I unsurprisingly cracked and cracked hard.
Having learned from this, and taken Pete’s suggestions from the podcasts, I have rejigged to focus on 3 key workouts per week that you hit hard and perfectly, and fill in the rest with my have-fun rides when the weather cooperates. I have been taking the structure from the med vol SSB1/2 and adapting the workouts on my key days to be harder, and deleting the endurance ride and one other ride from the TR plan.
How would this idea play into AT? If I picked a plan from TR and deleted two of the rides, would the AI adapt to understand that I am getting TSS from elsewhere (harder rides/outdoor)? Or would it think I am severely under-training myself? As I can handle 500-600 TSS relatively well, if my workouts only summed to 300-400 (to head the question off I know the model responds to more than just TSS) most of my survey responses would not be in the ‘stressed/hard’ categorization and I would assume other fitness predictors would also feed into the AI to consider this as undertraining.
As a follow-up, and relates to comments from many past threads, if there is one flaw in TR for some the prescribed training can take up most of the matches many of us have per week, and makes it hard to go out and ride for fun or do moderate-hard rides outside. Is the AI tuned to the status quo, to remain hard? Or is there a way we as a rider can tell the AI what kind of response we want from a training block. Naturally, while big gains are good, sometimes we just want moderate improvements in fitness.
EDIT: A related question. If you start an AT training plan but make your own modifications (e.g. delete all the endurance rides) will the plan to still work with the AT/AI?
As custom workouts created in WC will be profiled in ML and credit will be given in AT for completed custom workouts, will custom workouts ever appear in the TrainNow or AT recommended workouts?
AT sounds great, and I really hope it does work. And that FTP is not the sole factor considered; I think it lacks relevance as a benchmark for high intensity activity.
Nothing to be jealous of - I wish they were my stats too! Was posing hypothetical question!
Thanks for the link.
I’ve been on TR for about 16 months and I am soo excited about AT. I signed up for the BETA and just want to set my expectations. Will all BETA “applicants” get in on the BETA OR just those selected, AND are we talking a wait of a few days OR weeks? Just hook it up to my veins already!!
I don’t think there’s any definite info on this and doubt there will be.
That said, Nate has stated that there will be at least 1 more wave going out to VIPs only and that some bugs/funky feedback has already started coming out of the first wave.
The way I read the tea leaves, that means rank-and-file users should probably expect to wait 1-1.5 months before they start checking their inbox.
(Don’t get me wrong, I too wish I could get into the beta TOMORROW, I’m just thinking realistically how long it’ll take to address funky stuff from wave 1 + send wave 2 + address funky stuff from wave 2 + start sending wider beta invites…)
Agreed! My main question, and whether I will continue on or take a break from TR, is if it will include Zwift races. I recently started racing in the WTRL/Zwift Racing League and LOVE IT!
Priorities for me now are the weekly ZRL races and I want my training structured around those as well as big “A” events in the future like a GranFondo.
That Dylan Johnson video was interesting to me as I realized I may have been doing a polarized block of training recently without really intending on that 
Races were so intense I had to backoff TR structured training. I ended up seeing my FTP bump up twice and push me into B category. Now I am very eager to frame all my training around trying to not suck in B races and find that really motivating for me versus doing a long-structured training plan aimed at an event in the summer.
Thanks @SeanHurley - sounds good but I think we’re at cross-purposes 
I was after a little more info about the “Other” options in the post-workout surveys when you either fail or cut short a workout:
Is it still going to be possible to customize your plans and then have adaptive training work on that? What happens to the Sunday endurance substitutes?
Thanks @ambermalika, sounds good. Is AT sufficiently advanced enough to somehow make use of the details entered into the text box? Or will the comments just be for reference when the user is viewing the workout at a later date?
We’ll use those for learning. In some cases, they will help us “teach” the ML model, and in other cases, they’ll help us iterate on future survey designs and product features!
whoops- I got lost in the chain of responding to responses 
So normally we can prescribe workouts in the calendar and then match an outdoor workout to the prescribed workout. Could we use this as a workaround to somewhat train the ML towards strengths. For instance if I do a 2 hour zwift race thats got some work at vo2max but mostly sweetspot I could find a workout that matches both for power and intensities and use that?
I’m just after hacks here
No worries. Amber’s replied with some info. 
Is the Adaptive Training version of the app more hardware / computing (whatever the correct term is) heavy? And if so, is it possible to have the AT running your plans on one device, and will that reflect correctly as planned workouts change?
I think this is key, and hopefully coming soon.
Being able to go into plan builder and say I have a race on X date and I generally have 1hr on Mondays, 2hrs on Tuesdays, 1hr weds, no time on Thursday, 1hr on Friday morning and 3hrs each on sat/sun… And you plug it in and it spits out a plan that then constantly adapts under those criteria. Current plan builder is close, but could be better with regards to time constraints/allotment.
Sounds like the real computing work takes place on the TR side with whatever systems they have. Our local apps seem to be essential the same demand, since they are simply pulling the information from the TR side.
So it will be the same info on all of your devices (assuming they are all up to date with the AT related views).
You guys think there’s negative feedback in the DJ thread now…just wait!
“AT said I’d be at a 269 watt ftp by July. I only skipped 30% of the workouts, slept 5hrs a night, did all my training fasted, supplemented with zwift races…and I’m only at 250 watts! I want my $75 back!”
Yes, you’ll have the same info available on mobile as on desktop, and all updates/adaptations will sync accordingly.