As I write this, we're in an AI-induced mass psychosis. People are token-maxxing themselves into hospital beds, scrambling to capture market value before automation takes over. Models keep improving, programmers are laid off, and we're told AI will write 100% of the code by year-end. The Agentic Era promises productivity gains, yet software keeps deteriorating.

Last week alone: my banking app needs three Face ID logins for 3D Secure; Slack stole focus from Ghostty, sending a git pull to a group chat; an LG fridge warranty form failed at submission, visible only in the JavaScript console; my car's infotainment update introduced 1-2-second lag, silent turn signals, and wrong app launches—bugs that impair driving.

Teams behind these bugs likely have LLMs with generous token budgets, yet quality decays. Software complexity has exploded since the mac OS Snow Leopard era; abstractions and frameworks make everything fragile. Updates now inspire dread. Vendors chase KPIs; fixing bugs doesn't look exciting in presentations.

This isn't anti-AI—GPU farms give superpowers—but we can't build better software. As companies accumulate AI debt, individual developers can create software previously out of reach. Rebellions against macOS and Windows are emerging; I hope the trend spreads across the stack.