VIBE CODING
Vibe code without losing your brain
AI can write the code, but you're still responsible for what it does. This section is about doing it mindfully, or choosing not to.
The one rule
Autopilot is the biggest mistake. Before you accept anything an agent wrote, you should be able to explain every change, point to a check that would fail if it broke, and say where you would look first if it failed. When you can't, writing that part by hand is a real option, not a failure.
What this section will cover
Each entry will follow the same shape: the concept, what to do, what not to do, why it matters, a 30-second check you can run on your own code, and its sources.
- SecurityGenerated code that works but lets the wrong people in.
- Data & secretsAPI keys, .env files, and anything that should never reach git.
- ReliabilityWhat happens when the network, the disk, or the other service fails.
- PerformanceCode that is fine with ten rows and falls over with ten thousand.
- DeploymentThe gap between 'works on my machine' and works for everyone.
- Architecture & scalingKeeping a codebase coherent when an AI writes a lot of it.
- ObservabilityKnowing where to look when something breaks in production.
- ProcessGit, small changes, and reviewing every diff before you accept it.
- PromptingGiving the agent goals and checks instead of a solution to copy.
- Staying sharpDebugging and reasoning skills you keep using, so you don't lose them.