Do the Reps
AI creates shortcuts. Keeping up is about doing the work.
Most people know AI is changing the game. They’re just not certain what to do about it. I’m not either. But one thing I’m doing is using AI to make sure I’m not embarrassed by my own gaps in knowledge.
I’ve been running a learning experiment on myself. I grab a research paper from SSRN. Recently, one on global resilient supply chain network design. I read it. Then I turn it into a podcast using Google’s NotebookLM, which forces the ideas into a conversational frame I can listen to (key: multi-modal learning).
Midway through, I realized I was fuzzy on MIP solvers. The paper wasn’t about MIP solvers… it used them as a tool. But I felt the gap. So I asked Claude to generate a self-contained interactive tutorial, and I worked through the exercises. When I got stuck, I asked Claude to explain. Not to summarize. To teach.
Here’s what I retained: MIP solvers tell you which seat in the stadium. Precise, but computationally expensive. LP estimates get you in the ballpark. Approximate, but often good enough for decision-making, and much faster to solve than precise MIP formulations. The optimization gap, how close you get to the true optimal1, is something you choose. One percent is usually good enough.
I didn’t memorize an algorithm. I built intuition for why the tradeoff exists. That’s the kind of understanding that changes how I make decisions day to day.
I work with people who know this stuff like the back of their hand. If I take shortcuts, it’ll be clear as day. AI can generate a summary I skim and forget. And the first time I’m in a room with someone who actually understands network optimization, that shortcut shows.
Or AI can generate a tutorial I work through. One that reveals exactly where my understanding breaks down, so I can close the gap before the room does it for me. Same tool. Opposite outcomes. The difference is whether I do the reps.
I know this material. But there’s only so much I can keep in working memory. And just like running, if I don’t use these muscles, they atrophy. So this isn’t a one-time climb. It’s a training regimen. The reps don’t stop because I already know the route, they stop when I’m okay with getting slower.
franksprotos.com/mip_tutorial.html
I know, there’s nuance here…



How do you decide what is worth learning? Is it just curiosity?