Post

AI moved the bottleneck from coding to review. My daily workflow now: issue comes in, AI handles the implementation. But I review every change, every behavior update, every line that goes live. How did I learn to do that? By writing code. For years. There's no shortcut.

You can't review what you don't understand. When AI suggests a caching layer, I understand the tradeoffs because I've fixed cache bugs in the middle of the night. When it designs a distributed system, I can spot problems because I've faced them before.

0 reposts0 likes

"Cars exist, so don't learn to walk." That's what "don't learn to code" sounds like to me. Walking builds the balance you need to drive. Coding builds the mental models you need to review, architect, and lead.

When I said the bootcamp path is closing, some people heard "don't learn to code." That's not what I meant. The answer isn't to stop learning. The answer is to learn differently. Go deeper. Fundamentals, not just frameworks. Why things work, not just how to use them.

The "software plumber" serving local businesses? That person needs to code too. The first version might work with vibe coding. But who maintains it when the business changes? Who fixes it when it breaks on the busiest weekend?

I've made open source contributions I didn't fully understand. Every maintainer has done this. It always causes problems later. The bugs are harder to fix, the technical debt grows. Now imagine that at scale. Entire systems nobody can read or debug.