Star on GitHubStar1.1k stars

Case study

Dinosaur runner: an REA-inspected speed rule rebuilt with Codex

The author uses REA’s MCP browser tools to passively capture the JavaScript loaded by wayou’s open-source T-Rex Runner page, then has Codex rebuild its acceleration rule in one self-contained HTML5 Canvas game. The report identifies the captured script by SHA-256 and separates verified source facts from reconstruction choices; only the speed rule is claimed, not full gameplay equivalence.

Targets
Game · Website
Category
Community REA investigations
Relationship
Author-reported mixed workflow
Sources reviewed

Requirements & limits

  • Prerequisites: Playing needs only a browser: dino-demo.html is self-contained, with no external assets, packages or build step, and a GitHub Pages copy is linked. The embedded tests run from the page’s Run speed tests button, the browser console or the documented Node.js harness with mocked DOM, Canvas and animation-frame interfaces.
  • Scope: The target is wayou’s open-source, Chromium-derived T-Rex Runner page, not chrome://dino/ or hidden proprietary code; the investigation read JavaScript the browser had already loaded. Only the acceleration rule is reconstructed: start at 6 and add 0.001 per eligible update while below 13, with no elapsed-time factor. Artwork, physics, obstacles and scoring are new or simplified, and the intro, ducking, audio and narrow-screen speed reduction are omitted.
  • Attribution: README and the report record REA MCP list_browser_targets and inspect_web_page calls over a local Chrome DevTools Protocol endpoint, with no navigation, clicks or page evaluation; Codex handled analysis, implementation and tests. Evidence IDs and excerpts come from the author’s REA results, but the full evidence bundle is not published and the live upstream script may change.
  • Testing evidence: The author reports 67 of 67 embedded tests passing (15 unit, 52 integration), ten rejected in-memory mutations and seven owner-reported manual browser checks on 2026-10-13 without browser or OS details. Floating-point results such as the extra step past 13 at update 7,001 were measured in the reconstruction, not sampled from the original. There is no CI or cross-browser coverage.
  • License: The project is MIT-licensed; short original-source excerpts keep the Chromium BSD-style header, and THIRD_PARTY_NOTICES.md reproduces the upstream BSD-3-Clause terms. This third-party repository is separate from REA’s own dinosaur-speed teaching case.

About this entry

The description was reviewed against primary sources.

Report an outdated entry Suggest a translation fix

Primary sources

  1. README.md GitHub · zhxunzh84/rea-dino-reverse-engineering · Revision 1fa0ae9
  2. docs/reverse-engineering.md GitHub · zhxunzh84/rea-dino-reverse-engineering · Revision 1fa0ae9
  3. docs/testing.md GitHub · zhxunzh84/rea-dino-reverse-engineering · Revision 1fa0ae9
  4. docs/lessons-learned.md GitHub · zhxunzh84/rea-dino-reverse-engineering · Revision 1fa0ae9
  5. dino-demo.html GitHub · zhxunzh84/rea-dino-reverse-engineering · Revision 1fa0ae9
  6. THIRD_PARTY_NOTICES.md GitHub · zhxunzh84/rea-dino-reverse-engineering · Revision 1fa0ae9
  7. LICENSE GitHub · zhxunzh84/rea-dino-reverse-engineering · Revision 1fa0ae9

More in Websites

All 61 entries in WebsitesWebsites case studies (13)

↑↓ move↵ open⌘↵ new tab