The longer I work on Rails monoliths, the more I appreciate boring architecture. Most complexity enters through exceptions, not through the happy path.
Every system eventually gets the monitoring it deserved on day one. The only question is whether you add it before or after the outage.
Self-hosting isn’t hard because servers are hard. It’s hard because you’re now responsible for every operational decision your platform vendor used to make.
The biggest productivity gain from coding agents isn’t writing code. It’s having a second engineer that never gets tired of reading 40 files to answer a single question.
One-million-token context windows sound like magic until you ask what entered the window.
Old decisions. Dead code. Stale docs. Conflicting conventions. Half-written diffs.
More context can help.
It also raises the review bar for what context is allowed to steer the patch.