Free observability audit
Green alerts you don't trust? Send me your stack and I'll come back with the 3–5 alerts that actually matter, plus one thing you almost certainly have wrong. Around two hours of my time, free, no obligation.
Posts
The empty alert payload: why autonomous triage fails on representation, not reasoning
Two weeks of running an LLM-driven triage loop over a telecom alerting stack, and every wrong conclusion came from the summary layer instead of the model. What that says about the autonomous NOC.Per-path SLOs need packet-level evidence, not site averages
Site averages compress away the one path that is actually broken. How eBPF packet-level evidence and OTLP let you define SLOs at the granularity your service really fails at, without blowing up your time-series database.Designing APIs for maximum usability
If API transactions are not your application's main communication, it's development tend to be Ad-hoc. What should be your concerns when designing an API that everyone will want to use?Why distributing systems affects your whole organization?
Distributed systems is an architectural paradigm that makes any application more reliable, scalable and maintainable. The key to design a distributed system is to isolate behaviors.There is life after TLS-SNI-01 deprecation
Some people have gone crazy when Let's Encrypt announced the deprecation of the TLS-SNI-01 domain validation method. This post is about my reaction to the news.