Show HN: Sonn a reasoning layer that nudges coding agent before it makes mistake | Hacker News Reader