Hi, Juraci here. I'm an active member of the OpenTelemetry community, part of the governance committee, and since January, co-founder at OllyGarden. But this isn't about OllyGarden.

This is about a problem I've seen for years: we pour tons of effort into instrumentation, but we've never had a standard way to measure if it's any good. We just rely on gut feeling.

To fix this, I've started working with others in the community on an open spec for an "Instrumentation Score." The idea is simple: a numerical score that objectively measures the quality of OTLP data against a set of rules.

Think of rules that would flag real-world issues, like: * Traces missing service.name, making them impossible to assign to a team. * High-cardinality metric labels that are secretly blowing up your time series database. * Incomplete traces with holes in them because context propagation is broken somewhere.

The early spec is now on GitHub at https://github.com/instrumentation-score/, and I believe this only works if it's a true community effort. The experience of the engineers here is what will make it genuinely useful.