Initially, It will be a lot of the same traditional workouts because the “AI” is really an “LLM”. It will see 99 workouts of 10×100 on a 1:20 and negate the 1 performance-specific workout.
Long-term, unless the collecting program realizes that major improvements come from technical adjustments the model will continuously recommend conditioning-based workouts with little regard to understanding swim-specific technical/muscular improvements.
I’m building a custom GPT with the technical training material. So far it’s getting better each instance and I should have a beta demo for the community by January.