Why It Matters
Significant development in Frontier LLMs reported by Reddit OpenAI Community.
Overview
I know AI isn’t a replacement for a good coach, but I’m trying to figure out how close I can realistically get to using GPT as a long-term adaptive training coach. I competed in CrossFit for about 12 years and have a substantial amount of experience with programming and training. I also have years of my own programming available, plus research and information on the specific areas I’m currently trying to develop.
Why It Matters
This development was reported by Reddit OpenAI Community.
Key Details
My training now is more complicated than simply generating CrossFit workouts.
I’m trying to combine trail/ultra running development with CrossFit-style strength, power, conditioning, durability, and injury prevention, with the programming changing based on my current goals, training history, recovery, injuries, and how I’m responding day to day.
Ideally, I want GPT to function more like an actual coach managing a long-term program.
It should understand the larger training cycle, know what adaptations we're currently targeting, remember what I've recently done, manage volume and intensity, deliberately progress movements and training blocks, introduce enough exercise variation without becoming random, and adjust the day's training based on recovery and feedback.
I’ve gotten reasonably close using a Project with multiple chats, GPT-5.6 Sol at high thinking, several deep-research reports, my training information, and a fairly detailed set of project instructions.
--- *Original report published by [Reddit OpenAI (ChatGPT, Codex & Leaks)](https://www.reddit.com/r/OpenAI/comments/1wiopgv/whats_the_best_way_to_build_a_training_coach/).*