So we've gone from blindly piping random scripts from the internet directly into our shell, to letting an AI hallucinate malicious code and execute it for us. Progress!
The "recommended" tag really seals the deal here. Nothing says "enterprise-ready" quite like pointing Claude at an installer script and hoping it doesn't decide to rm -rf / your entire system because it misunderstood the prompt.
At least with curl | sh you knew exactly which sketchy website was about to own your machine. Now you get the added thrill of non-deterministic execution. Run it twice, get two different results. It's like Russian roulette, but for your production environment.
Security teams everywhere just felt a disturbance in the force.
AI
AWS
Agile
Algorithms
Android
Apple
Bash
C++