Evolve AI

[alek]

Did the first workout today.

Comp Squat
275x1@7.5, 210x4x3@5.5

RDL
205x12@7, 205x12@7.5, 205x12@8.5

Rear Foot Elevated Split Squats
20x10@8, 0x12@8, 0x12@8

Hanging Leg Raises
3x12

- Based squat on 295 e1RM trends (still dealing with some issues)
- Skipped Incline bench. Plan on doing this later in the week
- Did RDL (posterior chain accessory) before split squats (quad accessory), and dropped 1 set of split squats
- Dropped 1 set of hanging leg raises

Total session time 65 minutes. I think incline bench and the 2 sets of accessories I skipped would've pushed this to 90 minutes, which is too long for me in week 1 of a 5 day/week program.

I'm going to do day 5 tomorrow (comp bench) and haven't done a paused bench in a very long time, so we'll see how it goes.

Idk how the app will treat these changes, but there needs to be some flexibility..

That's one thing that I really did not like about JuggAI: I could not move exercises between sessions, nor add exercises to a session (At least I couldn't figure out how to add exercises beyond the prescribed ones.)

[gtl]

That's one thing that I really did not like about JuggAI: I could not move exercises between sessions, nor add exercises to a session (At least I couldn't figure out how to add exercises beyond the prescribed ones.)

I sent them a DM asking about this and will report back..

[gtl]

Swapped day 5 with day 2

Comp Bench
210x1@5, 225x1@6.5, 235x1@8
175x5x3@4

Chest Supported Row
140x12@6, 140x12@7, 140x12@7, 140x12@8

DB Incline Bench
55x12@7, 55x12@7.5, 55x12@8, 55x9@9

DB Hammer Curls
20x15@6, 20x12@7, 20x12@7.5, 20x10@8

- Haven't paused a bench in 4 years; this went surprisingly well
- Skipped triceps accessories

Total session 68 minutes, and I kept rest periods pretty quick (as evident by the quick drops in interset performance)

Might move this over to the log section and starting logging again.

[perman]

It looks like a pure website app. Can't find it on Android. To me, that's a no-go right there.

What phone do you have? Worked fine on my Samsung S21.

Motorola G100.

Edit: Found it now. Probably just didn't look well enough last time. I have to scroll down a fair bit, and EvolveAI is in one word, so Evolve ai in two words returns nothing

[perman]

Trying it now.

It seems to be forcing an upper lower split from 4 day medium to high powerbuilding, and I see no way to modify that. Given that I play rec soccer on Tuesdays, and lift weights Monday, I don't want the prior day to be a leg day, but I see no way to modify that directly.

Edit: I guess I can train lower upper and switch days 1 and 2 and days 3 and 4 in the actual days I train maybe...

[gtl]

Trying it now.

It seems to be forcing an upper lower split from 4 day medium to high powerbuilding, and I see no way to modify that. Given that I play rec soccer on Tuesdays, and lift weights Monday, I don't want the prior day to be a leg day, but I see no way to modify that directly.

Edit: I guess I can train lower upper and switch days 1 and 2 and days 3 and 4 in the actual days I train maybe...

Yeah they confirmed there is no way to change it in the app, but you can just go out of order. I feel like that should be on the fast follow list.

It is interesting that there is a lot of upper/lower split given that Mike T is involved. I thought RTS was more full body.

My current split now is

Squat + lower
Bench + upper accessories
off
Deadlift + lower
SBD
Bench supplemental + upper accessories
off

[perman]

Honestly, having compared this and juggernaut, it's basically identical. My girlfriend read a reddit thread where someone speculated this app was likely to get sued and removed soon. Apparently, the same dude programmed both apps.

This one's a lot cheaper per month though.

I see no reason for Mike T to have his name on this. He has own shit that bears very little resemblance. If he ever took part in designing an AI, it would spit out very different programs. I reckon someone just paid him and he said yes with less skepticism than he should have had.

[houzi]

Honestly, having compared this and juggernaut, it's basically identical. My girlfriend read a reddit thread where someone speculated this app was likely to get sued and removed soon. Apparently, the same dude programmed both apps.

This one's a lot cheaper per month though.

I see no reason for Mike T to have his name on this. He has own shit that bears very little resemblance. If he ever took part in designing an AI, it would spit out very different programs. I reckon someone just paid him and he said yes with less skepticism than he should have had.

I believe that Mike T has got some sort of AI / spreadsheet / app / thing that he has tested. I remember seeing in a Calgary Barbell video that Bryce (who is coached by Mike T) made reference to a spreadsheet that adjusted exercise selection/session intent/volume/etc based on inputting various factors at the beginning of each session.

So perhaps this is the testbed for having a platform to eventually merge the two together? No idea.

[gtl]

Honestly, having compared this and juggernaut, it's basically identical. My girlfriend read a reddit thread where someone speculated this app was likely to get sued and removed soon. Apparently, the same dude programmed both apps.

I've read this across the internetz numerous times now. Garrett wrote JuggAI and moved on to do his own thing. Its no surprise things look similar.

Outside of UI being similar, how many different ways are there to program for PL? I would almost argue that Jugg looks like the Cube method, which looks like 5/3/1 with periodization, which looks like BFS, which looks like.. etc.

I imagine the differences would have to be around the actual progressions, auto regulation, the AI, etc.

[James]

Blevins put up a video talking about about how evolve ai came about. Haven't finished it but what I've gotten out of it so far is he had a problem with programming benchmarks in evolve and went to pick Mike T's brain on it then realized Emerging Strategies was the fix.

[gtl]

Starting week 3, I am a little bit skeptical of this app.

Here is what it is having me do with Squat

Week1: x1@8, then 70% e1RM x4x3 (rpe 5.5)
Week 2: x1@8, then 70% e1RM x3x4 (rpe 6)
Week 3: 4x72-74%, then 1 more set of 4 based on last set

so, I think I may have accidentally toggled on "benchmark" mode to get the x1@8. If I toggle it back on, I get the same x1@8, followed by 1 backoff set of 4 for week 3.

2 other squat sessions for the week are 2-3 sets of low intensity low rpe, too.

Ex: 3x3@5.5 comp squat and 2x4@5.5 SSB squat

I answer all of the readiness stuff with pretty much all average ratings. If I increase readiness, it appears to just add 5-10lbs and maybe 1 set, depending on the workout.

I'm just skeptical how I won't detrain doing 6 sets of sub6 rpe squats a week. Looking at the dashboard, volume doesn't go up from here, just intensity, but even that doesn't make it look like this is high intensity low volume.

[perman]

Honestly, having compared this and juggernaut, it's basically identical. My girlfriend read a reddit thread where someone speculated this app was likely to get sued and removed soon. Apparently, the same dude programmed both apps.

I've read this across the internetz numerous times now. Garrett wrote JuggAI and moved on to do his own thing. Its no surprise things look similar.

Outside of UI being similar, how many different ways are there to program for PL? I would almost argue that Jugg looks like the Cube method, which looks like 5/3/1 with periodization, which looks like BFS, which looks like.. etc.

I imagine the differences would have to be around the actual progressions, auto regulation, the AI, etc.

Well, I'm not sure what the differences are really. Answering the same questions, the same way, gives you the same basic template (though my girlfriend noted certain minor differences in sets/reps when you zoomed in).

I wish you could customize way more. It basically looks like you're choosing either powerlifting or powerbuilding, number of days, which muscles are important, and then the app spits out a template without allowing you basic choices like what type of split you want, or reordering which days has which slots, the numbering of each slot in a day, or even just straight up deleting exercises.
My template even had a bug where I couldn't switch out shoulder press (I have a cranky shoulder now, and the app felt stifling when I needed to be adaptive).

I gave both Sheiko AI and Evolve AI a shot, but I found both not as good as I wished they were. I wanted apps that could accommodate a flexible training schedule and be able to accommodate well on the fly according to my personal priorities. But the apps market themselves as coaches that tell what you should do (though tbf, Sheiko was more adaptable), rather than assistants that could conform to your wishes, expectations and requirements. And I want the latter.

[gtl]

Honestly, having compared this and juggernaut, it's basically identical. My girlfriend read a reddit thread where someone speculated this app was likely to get sued and removed soon. Apparently, the same dude programmed both apps.

I've read this across the internetz numerous times now. Garrett wrote JuggAI and moved on to do his own thing. Its no surprise things look similar.

Outside of UI being similar, how many different ways are there to program for PL? I would almost argue that Jugg looks like the Cube method, which looks like 5/3/1 with periodization, which looks like BFS, which looks like.. etc.

I imagine the differences would have to be around the actual progressions, auto regulation, the AI, etc.

Well, I'm not sure what the differences are really. Answering the same questions, the same way, gives you the same basic template (though my girlfriend noted certain minor differences in sets/reps when you zoomed in).

I wish you could customize way more. It basically looks like you're choosing either powerlifting or powerbuilding, number of days, which muscles are important, and then the app spits out a template without allowing you basic choices like what type of split you want, or reordering which days has which slots, the numbering of each slot in a day, or even just straight up deleting exercises.
My template even had a bug where I couldn't switch out shoulder press (I have a cranky shoulder now, and the app felt stifling when I needed to be adaptive).

I gave both Sheiko AI and Evolve AI a shot, but I found both not as good as I wished they were. I wanted apps that could accommodate a flexible training schedule and be able to accommodate well on the fly according to my personal priorities. But the apps market themselves as coaches that tell what you should do (though tbf, Sheiko was more adaptable), rather than assistants that could conform to your wishes, expectations and requirements. And I want the latter.

I agree wrt the lack of customization/flexibility. For example, today I did 4@6.5 squat followed by 4@5.5 -- and that was all the app had me do for squatting. I would've liked to even add another set and have the app measure if adding the other set was the right thing to do based on future performance/readiness/strength increases.

[Skander]

I'm getting annoyed with juggernaut for only suggesting load changes and not rep changes when things get hard. So I'll rate a set of 140kg as higher RPE, and it's like "ok, do 137.5" or whatever. Stupid AI, I'm not gonna unload to get down to that weight, I'd rather just reduce reps/ increase sets than reloading over and over. It seems like intelligent loading is one of the key possible features, but I'm not seeing it being actually that useful in practice.

That said, my compliance to the program sucks because of the million things I'm supposed to do and not having time or energy to do them.

[gtl]

I'm getting annoyed with juggernaut for only suggesting load changes and not rep changes when things get hard. So I'll rate a set of 140kg as higher RPE, and it's like "ok, do 137.5" or whatever. Stupid AI, I'm not gonna unload to get down to that weight, I'd rather just reduce reps/ increase sets than reloading over and over. It seems like intelligent loading is one of the key possible features, but I'm not seeing it being actually that useful in practice.

That said, my compliance to the program sucks because of the million things I'm supposed to do and not having time or energy to do them.

What would Juggernaut do if you just did the same weight and -1 rep?

Yesterday, Evolve AI had me do my first backoff set as190x4, rated it @6.5. It said to go up to 195. I stayed at 190. Rated it @7 this time. It dropped the last set from 195 back to 190. I am smarter than the AI!

[CheekiBreekiFitness]

Do you guys feel that whatever the AI recommends is substantially better than what you (as a human) would recommend given your training history (assuming that you log all your training and you have a basic understanding of programming principles) ?

Or do you feel that the added value of the AI is to outsource the decision making to something else that is not you (I can definitely see the value in that too, in the sense that you don't have to think and overthink your training, you just launch the app and execute the workout to the best of your abilities).

[gtl]

Do you guys feel that whatever the AI recommends is substantially better than what you (as a human) would recommend given your training history (assuming that you log all your training and you have a basic understanding of programming principles) ?

Or do you feel that the added value of the AI is to outsource the decision making to something else that is not you (I can definitely see the value in that too, in the sense that you don't have to think and overthink your training, you just launch the app and execute the workout to the best of your abilities).

I wouldn't say I think its substantially better.. I think it makes a recommendation - and what I am hoping to get out of it, is that the result of that recommendation is then measured for impact and would drive future recommendations that affect the entire set of programming rules.

[Cleverusername]

Can you adjust deload frequency? JugAI had me deloading every fourth week which seemed way to frequent.

[gtl]

Can you adjust deload frequency? JugAI had me deloading every fourth week which seemed way to frequent.

I think you can just skip that week and it'll move you on to the next week (week 1 of new block). I'm not sure if it makes any adjustments to the next week by skipping the deload or not though.

With that said, I stopped using EvolveAI for a couple different reasons
* I didn't like the "split". 5 days a week was pretty much an upper/lower with 3 lowers and 2 uppers.
* I didn't like the low volume on main lifts (1-3 sets) with a bunch of accessories @ 4x10-15
* The AI didn't seem to really be doing much -- maybe it needs more time to learn, or they are still working out the kinks. Either way, I felt like I could do better on my own.

[Gustafson]

Can you adjust deload frequency? JugAI had me deloading every fourth week which seemed way to frequent.

I think you can just skip that week and it'll move you on to the next week (week 1 of new block). I'm not sure if it makes any adjustments to the next week by skipping the deload or not though.

With that said, I stopped using EvolveAI for a couple different reasons
* I didn't like the "split". 5 days a week was pretty much an upper/lower with 3 lowers and 2 uppers.
* I didn't like the low volume on main lifts (1-3 sets) with a bunch of accessories @ 4x10-15
* The AI didn't seem to really be doing much -- maybe it needs more time to learn, or they are still working out the kinks. Either way, I felt like I could do better on my own.

1-3 sets on the main lifts? It's been giving me a top set on primary SQ/DL/BP and 5 drop sets on SQ, 4 sets on DL and 6 sets on BP. Had to adjust the volume on DL. Also on the readiness survey before each day if you tell it to give you more volume it will.