Problem
Root README.md has carried ## TODO: host a demo with all the performance artifacts (Combine with RSC demo by Abanoub) since the re-init, and the shakaperf.com Examples page links only to the repo and case studies. Nobody evaluating the product can see a real report — the single most persuasive artifact — without doing the full 2-hour setup themselves.
Why it matters
For the agent-era positioning specifically, the money demo is: agent opens a PR → ShakaPerf flags "LCP +212ms, p=0.004" → agent iterates twice against compare --filter → ships "LCP −38%, p<0.01, visuals unchanged." No competitor can show that today (CodSpeed shows it for code benchmarks, nobody for browser/CWV). It converts better than any docs page and anchors HN/X posts.
Direction
- Publish a real
self-contained-performance-report.html from demo-ecommerce (it's a single file — trivially hostable at e.g. demo.shakaperf.com or /demo on the site) with a deliberate, understandable regression.
- Record the agent loop (asciinema or short video): Claude Code receiving the failing verdict, reading
report.json, fixing, re-running, passing.
- Link both from the homepage hero area, the README (replacing the TODO heading), and the Examples page.
- Stretch: a "try it" repo (demo-ecommerce extracted or templated) so an agent can be pointed at a ready-made playground.
Problem
Root
README.mdhas carried## TODO: host a demo with all the performance artifacts (Combine with RSC demo by Abanoub)since the re-init, and the shakaperf.com Examples page links only to the repo and case studies. Nobody evaluating the product can see a real report — the single most persuasive artifact — without doing the full 2-hour setup themselves.Why it matters
For the agent-era positioning specifically, the money demo is: agent opens a PR → ShakaPerf flags "LCP +212ms, p=0.004" → agent iterates twice against
compare --filter→ ships "LCP −38%, p<0.01, visuals unchanged." No competitor can show that today (CodSpeed shows it for code benchmarks, nobody for browser/CWV). It converts better than any docs page and anchors HN/X posts.Direction
self-contained-performance-report.htmlfrom demo-ecommerce (it's a single file — trivially hostable at e.g. demo.shakaperf.com or /demo on the site) with a deliberate, understandable regression.report.json, fixing, re-running, passing.