Claude writes code for me. It writes good code. It creates branches, runs tests, opens pull requests, and sometimes I just sit there drinking coffee while it refactors an entire module. This is not an exaggeration — this is literally what I did this morning.
So why am I publishing a course on terminals, Linux, Git, and SSH?
Because the code has to run somewhere. And that somewhere is not Claude's problem — it's yours.
Here's what actually happens when you let AI do everything and you don't understand the environment: it writes a script that looks perfect. You run it. It fails. The error says permission denied. You don't know what that means. You ask the AI. It tells you to sudo it. You do. Now it runs as root and writes files that your normal user can't touch. Your next deploy breaks. You ask the AI again. It suggests chmod 777. Now your server is a security joke. You didn't understand a single step. You just kept asking and pasting.
That's not engineering. That's being a middleman between an AI and a terminal.
The whole point of AI writing code is that you — the engineer — get to operate at a higher level. You think about architecture. You decide what to build. You review what the AI produced. You catch the mistakes. But you can only catch mistakes if you actually understand what's happening.
When Claude creates a Dockerfile, you need to know why it chose python:3.12-slim and not python:3.12. When it sets up an SSH tunnel, you need to know what port forwarding is. When it writes a bash script with set -eu, you need to know what that does and why it matters. When it commits to the wrong branch, you need to know how to fix it without destroying the repo.
You are the pilot. The AI is the autopilot. Autopilot is fantastic — until something goes wrong. And when something goes wrong at 35,000 feet, you better know how to fly the plane.
This is not about gatekeeping. This is not "learn to suffer through manual work because I did." No. Use the AI. Use it aggressively. Let it write your boilerplate, your tests, your configs. But understand the ground you're standing on.
Terminals, filesystems, permissions, processes, Git, SSH, Docker — these are the foundation. They don't change every six months like JavaScript frameworks. They've been stable for decades. Learn them once, and they'll serve you for the rest of your career, regardless of which AI is writing your code.
Here's the thing — this stuff is natural to many of us. Second nature. We forget we ever had to learn it. But to a huge number of technical, smart, computer-literate people, it's complete alien-speak. I was surprised by this too — I genuinely thought terminals and Git were obvious to most technical computer users. They're not. And that's fine. No stupid questions.
So I made a course. It's free. It's practical. It's not a textbook — it's the stuff that should be obvious but isn't, explained the way I'd explain it to a friend.