Category
Tech
Deep dives, tutorials, and post-mortems from the workbench.
Build the alarms before you need them
A component sat in a fault state for hours and paged no one. The dashboard knew. No human did. That gap (between monitoring and alerting) is where most reliability is won or lost, and AI writing the code is making it wider.
The case for small models on your own machine
After six months running everything locally, here's what I learned about latency, privacy, and the surprising creativity of constrained systems.