Audience asks why the talk is about logs when Callaghan declared traces the most powerful data type. Answer: pragmatism. Tracing is magical but hard to explain; people's eyes light up at 'structured logging' but glaze at 'tracing'. Traces capture hierarchical context (tree of nested events queryable at any level) while logs are flat series. Traces are generally more expensive and slower than logs, and introduce head/tail-sampling cost-control problems worse than logging-cost issues. Start with logs as the gateway drug; graduate to traces.