Pick a Plan and Stick With It
One of the most exciting use cases for large language models (LLMs) like ChatGPT and Claude for me has been building personalized physical training programs.
Initially, I experimented by generating a running program based on a specific distance goal, and provided my age, gender, and weight. Through iterative prompting, I realized that the more context I provided (such as my current workout routine, available training days, and time constraints), the more personalized the programming became. All of this was done without having to do hours of research to adapt generic programs or hiring a coach.
From there, I realized I was not limited to running programs. As I mentioned a few weeks ago, I have started training for a few big mountain climbs this summer. With the habit of showing up already in place, I needed a real program to follow.
Instead of spending hours researching and piecing together a program, I provided the same information as I did with the running program, but also included the mountains that I am planning to climb.
What I got was surprisingly nuanced. The program accounted for each mountain’s unique challenges. It recognized where my current routine was already sufficient, and filled in the gaps with targeted training days. Based on my own independent research, this program seems sound.
More importantly, it removed a major bottleneck for me: picking a plan that I trust enough to stick with. Instead of spending so much time planning, I’ve been able to focus on training sooner. So far, this approach has been more structured than anything I followed last year.
The true test will be this summer on the mountains.
Food for Thought
I am reminded of a video I saw last year titled, You’ve consumed enough. It’s time to create. One of the core ideas is that we often spend too much time consuming tutorials instead of actually creating. In other words, we often spend too much time planning and not enough time doing the thing.