Practical, technical writing from engineers who ship production AI systems every week.
We've shipped 50+ agent systems. Here's a taxonomy of the failure modes nobody talks about — infinite loops, hallucinated tool calls, and tool schema drift.
KEDA's event-driven scaling changed how we handle inference workloads. Side-by-side comparison with real latency numbers from a GPU node pool.
Applying STRIDE to AI systems surfaces threats that traditional AppSec misses entirely. Here's our full threat model template, open for use.
How we define tool schemas once in YAML and auto-generate OpenAI function definitions, LangChain tools, and Claude tool_use blocks — with zero duplication.
Route 53 latency routing + Aurora Global DB + DynamoDB Global Tables — the exact Terraform modules we use for enterprise clients who can't afford 60-second RTO.
We debated keeping Matimo proprietary. Here's the business and philosophical reasoning behind the MTI licence decision — and the numbers 12 months later.
No spam. Every post is technical, practical, and worth your time. Unsubscribe any time.