fix(website): retire stub-era "4× faster" claims; reframe homepage around B2 evidence

[blove]

Summary

Continues from the B2 corrections PR (#125). With pretable / MUI parity confirmed at n=20 and the H1 flip narrative overturned, the homepage's three "4× faster" components were carrying stub-era numbers and editorial claims about codenames that no longer reflect reality. This PR brings the homepage into line with the real B2 evidence.

No prose has been merged yet — please read the prose draft below and confirm before merging. No auto-merge set.

What changed

• ComparisonTable.tsx — full data rewrite. Drops the "4× faster scroll" badge; replaces codename columns with real ag-grid / tanstack / mui columns; replaces scroll-row data with real B2 numbers (pretable 9.07, MUI 9.14, AG Grid 16.7, TanStack 16.7); adds row-height-fidelity / blank-gap / anchor-shift rows that surface the quality wedge; drops streaming rows pending follow-up feat: S5a transaction API with streaming updates benchmark #6 (real-comparator S5 evidence); replaces them with headless-engine + streaming-pipeline rows that honestly distinguish pretable's surface. Trail-marker labels rewritten to fact-checkable characterizations.
• ReceiptsBand.tsx — drops the "4×" hero stat; replaces with quality-wedge anchors: 0 blank gaps under scroll (accent), 9ms frame p95, ≤1px row-height fidelity, 25k/s max sustained update rate.
• FeatureGrid.tsx — replaces the "16ms p95 ... 4× faster than Grid Alpha Community" line with a parity + quality-wedge description that names real comparators.
• Test updates: ReceiptsBand.test.tsx asserts new hero stats; ComparisonTable.test.tsx asserts new trail-marker labels (regex-matched so future prose tweaks don't break the tests).

What's NOT in this PR

• Streaming-row evidence in ComparisonTable. Pending follow-up feat: S5a transaction API with streaming updates benchmark #6 (real-comparator S5 runset).
• Any source / package changes outside apps/website.
• Threshold or evaluator changes to scripts/bench-matrix.mjs.

Prose draft for review

ComparisonTable section subhead

Wrapped-text scroll at 3,000 rows on Chromium, against the three most-cited React grids. Pretable matches MUI at the front, ~1.7× ahead of AG Grid Community and TanStack on raw frame p95 — and is the only adapter here that clears every quality threshold (zero blank gaps, zero anchor shift, ≤1 px row-height drift) at full-grid feature weight.

Trail-marker labels

Adapter	Label
pretable	Recommended path (unchanged)
AG Grid	Slower scroll; row-height drift
TanStack	Headless; you wire selection and nav
MUI X	Parity at scroll p95; full-grid feature surface

Receipts hero stats

Value	Caption	Notes
0 (accent)	blank gaps under scroll	Quality wedge anchor
9ms	frame p95 / wrapped scroll	n=20 mean rounded; full number on /bench
≤1px	row-height fidelity	DOM-truth check
25k/s	max sustained update rate	pretable's own; comparative pending

FeatureGrid 60fps line

Sub-frame scroll p95 on wrapped text — at parity with the best full-grid comparator (MUI X) and ~1.7× ahead of AG Grid Community and TanStack. Zero blank gaps, zero anchor shift, ≤1 px row-height drift.

Test plan

• `pnpm -w typecheck` passes
• `pnpm -w test` passes (190 website tests)
• `pnpm -w lint` 0 errors
• `pnpm format` clean

After merge

B2 follow-up backlog: items 3 (autosize wiring), 5 (open comparator interaction scripts), 6 (S5/S7 coverage) still pending.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
pretable	Ready	Preview, Comment	May 9, 2026 3:57am

[github-actions]

Vercel preview ready

Preview: https://pretable-7d3q3kie3-cacheplane.vercel.app
Commit: a38cd2c430b516b7fd5b05860b71a70d739c4197

_{Updated automatically by the deploy-preview job.}