By the time the athlete has read that, the session will be over…
3 Likes
AI generated or generalised sessions?
i have been pondering this topic further and think that the key difference between coached and AI is summarised by “knowing about” and “knowing of”
3 Likes
A bit like the conversation on the sub 4 thread, I’d guess AI will do a half decent suggested training plan that you used to get off Fink etc
A human coach (should) take into account things like life, tiredness, weaknesses, e.g. more cycling if that’s a weakness. AI might improve but a coach should be able to be more reactive.
1 Like
Here’s a 16-week sub-4 hour marathon training plan, averaging 26.2 miles per week and about 3.1 days of running per week.
Plan Overview:
- Target Time: Sub-4:00 (9:09/mi or 5:41/km)
- Weekly Volume: ~26.2 miles / 42.2 km
- Running Days per Week: 3 or 4 (avg ~3.1)
- Key Runs:
- Long Run: Builds endurance
- Tempo / Marathon Pace Run: Builds lactate threshold
- Speed / Hills / Intervals: Improves VO2 max and strength
- Cutback Weeks: Every 4th week to allow recovery
16-Week Marathon Training Plan (Sub-4 | 3 Runs/Week Avg)
| Week | Long Run | Speed / Tempo | Easy / Recovery | Total Weekly Miles |
|---|---|---|---|---|
| 1 | 8 | 4 Tempo | 3 Easy | 15 |
| 2 | 9 | 5 Easy | 3 Easy | 17 |
| 3 | 10 | 5 Tempo | 3 Easy | 18 |
| 4 | 6 | 4 Easy | Rest | 10 (Cutback) |
| 5 | 11 | 6 Tempo | 3 Easy | 20 |
| 6 | 12 | 6 Hills | 4 Easy | 22 |
| 7 | 13 | 5 Tempo | 4 Easy | 22 |
| 8 | 8 | 4 Easy | Rest | 12 (Cutback) |
| 9 | 14 | 7 Tempo | 4 Easy | 25 |
| 10 | 15 | 6 Hills | 5 Easy | 26 |
| 11 | 16 | 6 Tempo | 5 Easy | 27 |
| 12 | 10 | 5 Easy | Rest | 15 (Cutback) |
| 13 | 17 | 7 Tempo | 5 Easy | 29 |
| 14 | 18 | 6 Hills | 4 Easy | 28 |
| 15 | 12 | 5 Tempo | 3 Easy | 20 (Start taper) |
| 16 | 26.2 | Rest | 2 Easy (race week) | 28.2 |
I think its undercooking the LR, not sure what tempo is for, and just a single 2mile run on race week is on the low side
1 Like
based on what?
its what ChatGPT gave me to the question
“I would like a 16 week training plan for a marathon, to target a sub 4 finish. I would like an average of 26.2miles per week over the 16 weeks and an average of 3.1 days of running”
yep…but it asked no questions such as terrain, envitonment, current training level, experience etc…
1 Like
I guess the AI coaches will also target standard events or distances, e.g. 70.3, Ironman, marathon rather than bonkers things like arch to arc where you need someone who understands the stupidity of it!
3 Likes
for many people, a park run is a bonkers event…
2 Likes
would have thought at least 1 race during any build ,if just to see were your at
that’s the problem, it will only ever be pre generated sessions, with a few changes to fit in. What it gave me is probably identical to something that’s been dumped in runners world every january
1 Like
Yep, and see above about all the fink plans that people shared in excel etc
1 Like
Hills?
Is that up or down them or both? 
Hard down easy up?
Hard up easy down?
2 Likes
AI, in a number of industries, is imperfect. This is no different.
It’s like having a junior member of the team produce something that needs further intervention to truely value.
The above sessions and continued discussions lean into that.
Does it give you a starting point? Yes.
Is it 40/50/60% there? Possibly.
Does it mean expert tweaking for optimal value? Yes.
But, where will this be in 2/3/4 years?
That 40/50/60% will be higher.
3 Likes
Fify…
In engineering and some medical applications ai makes a lot of sense.
Billions of known inputs, or at least millions, each with a certain known outcomes. Solve for a desired output.
Thats not how humans work.
It might get close in future but the tech isnt there at the moment for all the required inputs and the outcomes based on them is largely unknown
1 Like
Don’t known the workings
Blar blar marketing “ai generated session” based on the route profile, historical / predicted weather, personal profile, bike setup, weight etc
Spat out very different sessions for Cotswolds flat and Swansea hilly.
Both basically was a simplified version, as a workout, of the race plan.
Eyeballing the profile of the course and normal wind direction for the time of year I had a very similar race prep session in mind.
Did it really use ai who knows? Did it need to for this application, definitely not.
1 Like
Think we are arguing the same point, fella.
3 Likes
Yeah we are…
I started with a bit of a joke and then was reinforcing your point that I 
% agree with. My comment was for the general more causal readers, not you. You get it. Tbf most do, but it gave me something to do.
2 Likes
It’s the other way around - if you recall Douglas Adams Hhgttg and Deep Thought which calculated the answer to life, the universe and everything to be 42. They then realised if the answer was 42 then they clearly didn’t understand the question they had asked.
This has become reality, and we call it prompt engineering. Really, it’s the skill of knowing how to ask a question.