I spent hundreds of hours thinking about boundaries, spoke to 100+ people, and ran two related AI safety workshops. All because every time I asked someone to explain how boundaries worked, the explanation didn’t make sense. Note: This is an old post with rough edges.