RunTK.com: AI Prompt Generator Updates - Workout Export, Multi-Day Planning, and Graph Prompts
I’ll be upfront - this is a bit of a different one. These updates are less about how I personally use the tool and more of a response to what I think others might find useful based on how people are using the paid AI coaching apps to generate workouts, push them to intervals.icu, and display nice-looking graphs.
Honestly, I don’t think I will get much mileage out of these features. I’m either following an 8, 12, or 16 week training plan or I’m in an off-season base building phase where I’m mostly doing Z2 with the occasional random workout thrown in - which is actually why I built the random workout generator in the first place. And when it comes to graphs, intervals.icu already has everything I need, and I’d always trust those over something an AI generated on the fly anyways. That said, I know this is not how everyone trains, so here’s what’s new.
Multi-Day Workout Planning
I replaced “Sketch a 4-week plan” with something more immediately actionable. The purpose of this is really if you’re simply looking for a small block of workouts without committing to a whole training plan.
- Enable “Suggest next days workouts” and dial in how many days you want planned (1 to 30)
- The prompt asks for day-by-day sessions with specific activity type, duration or distance, target intensity zones, and a one-line coaching note
- It instructs the AI to respect your existing weekly pattern from your actual data rather than inventing a schedule from scratch
Include Graphs
There’s a new “Include graphs” checkbox that adds detailed chart specification instructions to the prompt.
- Currently there is fitness, recovery / readiness, fueling, load composition by sport, long run progression, intensity distribution, sleep accumulation, HRV vs. next-day-load, readiness heat-map, weight trend with 7-day smoothing
- If you guys want me to add any more just let me know, they’re simple to add
Why Not a Full AI Chat Interface?
After a discussion on this thread the other day, I was extremely tempted to build a full AI chat interface where users could bring their own API key and use it at cost. And I still might at some point. But after some careful consideration I came to the conclusion that whatever AI wrapper anyone builds, it’s always going to be at least a step or two behind ChatGPT/Claude/Gemini, which are constantly shipping new features and new ways for their AI to execute in-interface scripts and code. On top of that, any custom wrapper is likely going to be a lot more buggy as well. Trying to build and maintain one on top of that just doesn’t make sense right now, and IMO it’s really only a matter of time before MCP integrations replace them all.