A founding engineer at yeet describes replacing statsd_exporter with an eBPF-based kernel observability tool that captures StatsD metrics at the traffic control layer instead of requiring a listening socket, eliminating the need for a separate exporter binary while maintaining Prometheus compatibility.
A Show HN project that generates live system dashboards through natural language queries to an AI model. Users ask questions about their host (CPU usage, memory consumption, etc.) and receive real-time visualizations in twelve chart types, with no polling and theme-aware styling.
Software bills of materials generated from build-time lockfiles are inaccurate because they don't reflect runtime reality—processes running with manually installed software, unrestarte packages, and dynamically loaded plugins. yeet, an open-source observability tool, solves this by querying the kernel's /proc filesystem via a GraphQL interface to build an actual runtime bill of materials that reveals what software is truly in memory.