• How I Dropped Our Production Database and Now Pay 10% More for AWS: Lots of reliance on AI here and possibly not much awareness of how the underlying infrastructure works or how to create safety in a process with lots of automation. A lot of hard-won experience unlocked here. I like that he owned his mistake.
  • Ghostty Is Leaving GitHub: Github has apparently not been very reliable lately. I use bitbucket more than github these and haven’t really noticed reliability decline over the years but some people have and are sad.
  • Harness engineering for coding agent users: Talks about agent harnesses and the closed loop one tries to setup where a change is made, tested, and learnings collected which are fed back into the next change. How do we evaluate the health of the harness itself which is an important aspect of development context. If it falls out of sync with the goal because of bad signals or memory that isn’t helpful anymore it’s no longer a good tool for the task at hand and should be rest. How do you know you’ve landed in that state?
  • April 2026 Outage Post-Mortem: Postmortem for a bluesky outage in April. Sounds like they had an api endpoint that started to get post batch sizes much larger than they were expecting exhausting ports available to make new connections on. We learned about delays after a socket is closed by the application where the o/s holds onto the socket for up to 2m waiting for any lingering packets that may yet come due to network blips.