Guiney's wroclove.rb 2025 lightning talk sharing GoHiring's observability practices. Advises instrumenting on the network/kernel level via eBPF/cgroups to get service maps without modifying applications; preferring open-source backends like Elastic and Grafana (made swappable by the OpenTelemetry standard); writing logs to stdout and letting libraries handle threading/batching/back-pressure; setting a dedicated warn-level logger on all library logs (Ruby ecosystem library logs are 'pretty useless' but you still need to catch warnings and errors); adopting canonical log lines instead of scattered log entries; propagating trace IDs so one request can be followed end-to-end (they use it in Elasticsearch); and using ChatGPT to learn best practices. Covers RED metrics plus queue time as a critical missing metric. Organizational punchline: observability is a developer problem, not ops — scale-up fails with 20 devs + 4 ops. Concludes that OpenTelemetry is the direction but not yet mature in Ruby, and that frameworks/libraries need better built-in observability so developers don't have to think about it. Ends with a GoHiring recruiting pitch.