Build results-first D3 benchmark explorer

[nedcut]

Summary

• replaces the marketing-forward hero/ladder treatment with a results-first benchmark explorer
• makes the paired-lift forest plot the primary view and adds a D3 score-vs-cost lens
• adds an Analysis section with absolute score, selected-model evidence, action outcomes, and a D3 rejection heatmap
• uses one canonical leaderboard source for every chart, count, protocol observation, and table row
• removes obsolete mockups and unused redesign components

This is the implementation successor to #83.

Polish pass

• replaces the oversized 0-of-8 verdict with a neutral benchmark overview and compact readouts
• distinguishes the baseline-panel reference from the red pick-trader scripted bar
• links one selected model across the forest plot, cost view, table, score chart, inspector, and mechanic matrix
• adds collision-aware scatter labels, persistent exact-value readouts, keyboard selection, and reduced-motion-safe chart transitions
• introduces a semantic palette: teal for measured/accepted outcomes, violet for selection, amber for rejection risk, red only for the scripted bar, and ink-dashed for the oracle
• changes the heatmap to five stable amber risk bins with exact values and an explicit worse-direction legend

The polish direction was independently reviewed by Claude Fable 5 before implementation.

Scientific integrity

• all 8 published models stay in one descriptive tier
• the page leads with the paired baseline-panel comparison rather than an ordinal winner claim
• the Holm-adjusted family-test caveat appears beside the opening context
• the production build runs a consistency gate that reconciles published row count, repeats, and benchmark-bar results from src/data/leaderboard.json

Validation

• bun run lint
• bun run build
• uv run --extra dev python -m ruff format --check gm_bench examples tests
• uv run --extra dev python -m ruff check gm_bench examples tests
• uv run --extra dev python -m pytest -q — 474 passed
• desktop, mobile, light, and dark visual QA
• all heatmap text/background pairs exceed 4.5:1 contrast
• browser console: 0 errors, 0 warnings

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: f96f8eca-6fda-4564-bb9c-d718e1189d56

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/results-first-redesign

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}