- AI agent guardrails: how we built a validator to stop rogue actionsA separate LLM call reviews every side-effectual action before it runs. The agent has to justify it, and the validator checks that against what the user actually asked for.8 min
- iMessage API: three rewrites, one Apple ban, and what actually worksWe built iMessage support three times in six weeks. A Swift daemon, a BlueBubbles rewrite, then an API migration after Apple banned our account for sending 10K messages in 12 hours.9 min