Caveman not dumb. Caveman efficient.
This repo nails something most AI workflows miss:
less talking, more doing.
When you already know the codebase, you don’t need essays.
You need action.
https://t.co/5NipMEZYHw
In the AI era, the strongest engineers won’t be the ones who ship the most generated code.
They’ll be the ones who can understand the system, question the output, spot fragile architecture, and use AI as an accelerator — not as a replacement for judgment.
AI didn’t make software engineering easier.
It made code easier to produce — and systems easier to misunderstand.
The real skill now is not just prompting better.
It’s systems thinking:
Where does state live?
Where does feedback live?
What breaks if this component disappears?
AI can generate a function, a page, an API, even a full feature.
But it won’t automatically understand your product, your data flow, your failure modes, your business rules, or the hidden coupling inside your system.
That is still the developer’s job.
@abxxai Buffett content always goes viral because people hope there was a trick. There wasn’t. Mostly just discipline, patience, and avoiding stupid decisions.
@thsottiaux Main issue for me is premature confidence. It often locks onto an interpretation too early, then executes well on the wrong thing. Better clarification, better self-checking, less drift would help a lot.
AI is shifting from answers to actions.
Agents won’t just assist, they’ll: buy resources, sell services, work with other agents
A new market is already forming.
Yes, rules are unclear.
Yes, regulation is behind.
But that’s how every new tech wave starts.
We don’t wait until we’re ready.
We adapt — fast.