Hey everyone, maker here 👋 The origin story's up in the description, but the short version is loadr didn't start as a roadmap, it started as an experiment, one carefully-structured prompt to Claude Fable 5 while it was briefly available. What came back genuinely stopped me. Not a toy or a snippet, but a coherent, architecturally sound foundation, a real engine, the plugin seam, scripting wired in. It even generated little demo videos of itself running. So I kept building. The parts I'm most proud of now: • Percentiles you can actually trust, HDR histograms merged across threads, so p99.9 is real, not sampled. • Scenarios in plain YAML, with embedded JavaScript (per-VU on QuickJS) when you need real logic. • Protocols and datastores as installable plugins, so the core binary stays small. • A live web UI baked into the binary, a desktop app, a drop-in CI gate, and observe, which pulls your server's Prometheus metrics after a run and lines them up against your latency timeline, so "p95 spiked" meets "the DB hit 90% CPU". Honest caveats: it's source-available under Elastic-2.0 (not OSI open source), and still beta, HTTP/gRPC are rock solid, the browser-driven path is newer. What genuinely excites me isn't "AI wrote it", it's how far one well-structured prompt got on a real systems-architecture problem. The gap between an idea and working code is closing faster than I expected. Happy to go deep on any of it, the QuickJS-per-VU design, the HDR merge, even the original prompt. And I'd love to know: what would make loadr a daily driver for you? Try it: https://loadr.io · Demos: https://loadr.io/demos/

loadr

Find the breaking point — load testing in one Rust binary

Find the breaking point — load testing in one Rust binary