Engineering video
1:56 · 14 October 2026
The speaker explains: the meta-skills and reasoning built through learning how to code remain extremely valuable for making architectural decisions and building higher quality software with AI.
Hear it at 0:42 →So of course, when you invested so much time and then you see AI coding faster than a human and the typical react component or whatever that you take half an hour to think about and then you write and carefully craft and it took you three years only to get 20% faster doing that or 50% or whatever. And then you see doing that in a fraction of a second and people building things, people with no expertise, but they didn't make the same investment of time and money and then they build these things.
It feels demotivating also. So I totally understand that and from a career perspective, it feels like just because LLM's do that thing that you got so much better off, then you'll be fully replaced. And it's basically the same as saying that you've been starting to be a civil engineer and an aerospace engineer and then they release computer assisted software and you won't have a job. The reality is to build with the tool, the people that will build with the tool, the people that will build healthcare software, the people that will build finance software, they'll be software engineers.
Now, everything you learned will help you scale yourself with AI. So a software engineer vibe coding will be much more effective in vibe coder and you see the trend on X where people that started vibe coding are now thinking, oh, maybe I should learn the fundamentals. When they realize that they're doing bugs and solving problems that were solved way before, they realize, oh, probably if the code base was modular, I wouldn't spend so many tokens every time I fix a single problem because it wouldn't build this spaghetti that the model has to digest. Or if I had a stable API, it would be easy to refactor.
Or if I had tests on this, it would be easier to write tests or maybe I don't have to write TDD. If I keep changing code, I could just write end-to-end testing. But to make all those decisions, you need to reason like a software engineer, you need to know software engineering. So a lot of people are feeling like, well, this coding thing it's worthless. But the thing is that the meta skill, the reasoning that you build while learning how to code, it's extremely valuable, especially when building with AI because the things you can build with AI, if you know fundamentals, if you know computer science, if you're good at coding, are much bigger and of much higher quality than anyone that's vibe holding.
