Step 01The Prompt
You spend 15 minutes crafting a good prompt. Training age, goals, equipment constraints, injury history, frequency. The model gives you a solid block. Sets, reps, RPE targets, maybe even a deload week. This part works. LLMs are genuinely good at programming when given the right context.
Step 02The Copy-Paste
You screenshot the plan. Or copy it into Apple Notes. Or save the chat link. Or email yourself the thread. You now have a training program in a format that was designed for reading at a desk, not executing in a gym. The prescription is scattered across paragraphs, tables that don't resize, and conversational filler the model added to be helpful.
Step 03The First Session
You open the chat on your phone. Scroll past your prompt, past the model's caveats, past the warmup section, to today's session. You do your first set. Now you need to log it somewhere. Back in the chat? A separate tracker? A Google Sheet? There is no right answer. The chat was designed for generation, not execution. You're using a writing tool as a training interface.
Step 04The Decay
By session four, you stop logging. The overhead of switching between apps, remembering which tab has the plan, and manually tracking weights is more friction than the training itself. By session eight, you've deviated from the program. Swapped exercises because the rack was taken. Dropped a set because you were running late. Added volume because you felt good. None of this is recorded. The AI still thinks you're following the original prescription.
Step 05The Reset
The block ends. Or rather, it fades out somewhere around week six. You open a new chat. Re-explain your training age, goals, equipment, and injuries. The AI writes you another solid program. It has no memory of what volume you actually handled, what exercises you swapped, where you stalled, or what progression rate your body responded to. You're starting from scratch with a model that thinks it's meeting you for the first time.