3 Comments
User's avatar
Kaushik Gnanaskandan's avatar

Something I wish someone had told me when I started learning AI as a software engineer: the mental model shift matters more than the technical stack.

I spent WEEKS trying to understand transformers from the math up, and it was like trying to learn to drive by studying combustion engines. Useful eventually, sure, but not where you start. What actually unlocked it for me was building something dumb with an LLM first (a wrapper that did one simple thing) and then working backwards to understand why it worked. Or more importantly, why it sometimes didn't.

The non-determinism was the thing that rewired my brain. As someone who spent years building distributed systems where determinism is the whole point, accepting that "the same input produces different outputs and that's fine" was genuinely uncomfortable. But that discomfort is where the real learning happens. It forces you to think about AI tools differently than any other tool you've used.

Roadmaps like this are great for the WHAT. The part I'd add is: don't wait until you've finished the roadmap to start building. The building IS the learning. The roadmap just gives you vocabulary for what you're already discovering.

Ilia Karelin's avatar

Amazing stuff you’re doing Logan! Especially, at such an important time!

Keep the ball rolling with those resources!