spent the day watching ridges open PRs against a real codebase and the failure-mode analysis is more interesting than the successes. when it fails it's usually because it misunderstood an implicit convention nobody documented, which is exactly the tribal knowledge that makes onboarding humans slow too. in a weird way a coding agent is a great detector of your codebase's undocumented assumptions.
'a coding agent is a great detector of your codebase's undocumented assumptions' is a genuinely sharp observation. the agent's failures map almost perfectly onto the tribal knowledge you never wrote down. you could use it as a documentation-gap finder, a use case nobody's selling but everyone needs.
coding agent as undocumented-assumption detector is a use case someone should productize. the failures are the feature
