Builder Log
25 notes on this leg
Builder Log is where I document experiments with agentic AI, automation workflows, local systems, and the messy work of making useful tools reliable. These notes are less about the tool itself and more about what the build teaches.
When AI Memory Gets Too Complicated: Lessons from Replacing a Broken Plugin
A memory plugin that controls what the agent remembers controls the agent. What I found when I audited mine, why…
What a Week-Long AI Gateway Incident Taught Me About Reliability
A week of gateway instability reveals failure modes that stable operation hides. What I rebuilt — and the five design…
The Hook System Trap: Three Gateway Crashes and One Design Lesson
Three real crashes from the same misconfigured hook pattern — and the architecture insight that finally stopped them. The lesson…
The Heartbeat Problem: Monitoring an AI Agent That Never Stops Running
An always-on AI agent that is up but silently failing is worse than one that crashes loudly. The monitoring architecture…
Building a Draft-First AI Publishing Pipeline: Five Things That Broke
Building a pipeline where an AI drafts, stages, and waits for human approval before anything goes live. Five real breakages…
