The receipts
“Verified” should be checkable before you pay
Every nontrivial claim in the reference carries a source, a source tier, a capture date, and a volatility rating. This page shows the method, and a real excerpt of the evidence, so you can check the claim instead of taking our word for it.
The method, in five rules
- 1. Sources are tiered. Primary documents and official vendor documentation (store policies, framework release notes, court filings) outrank named industry research, which outranks trade press, which outranks blog posts and content farms. Load-bearing claims need the top tiers.
- 2. Vendors are treated as biased by default. Framework teams benchmark their own framework; SDK vendors quote their own match rates. Their directional claims can triangulate; their magnitudes get flagged. When a widely repeated benchmark traces back to a content farm with no stated method, the reference names it and refuses to repeat it.
- 3. Everything volatile is dated. Store commissions, review policies, framework versions, and legal rulings decay monthly. Each is stamped with when it was checked, and rated for how fast it rots — that's what the CURRENT AS OF stamp in the footer actually means.
- 4. Surprising claims need two independent lineages (or one primary source). When credible sources disagree — the EU fee stack is a live example — the reference records the range and the reason. It doesn't pick the convenient number.
- 5. Corrections stay visible. When re-checking proves an earlier entry wrong, the correction is logged against it — verification that can't lose isn't verification.
From ARC-006 — Mobile stack (Sources & last-verified):
- React Native 0.84 release, Feb 2026 (Hermes V1 default, New Architecture) — reactnative.dev/blog; current stable 0.86, released 2026-06-09 — reactnative.dev/versions [2026-07-02]
- Expo SDK 57 current (released 2026-06-30; ships React Native 0.86) — expo.dev/changelog/sdk-57 [2026-07-02]
- Expo EAS pricing — Production plan $199/month — expo.dev/pricing [2026-06-26]
And the part most references skip (from the same topic):
The widely circulated "Flutter cold-starts at 2.1s vs React Native 2.8s" and "Flutter wins 7/10 benchmark categories" numbers trace to content-farm blogs with no stated device, no release-vs-debug build, and no reproducible method. We could not verify them against any primary or independent benchmark, and we won't repeat them as fact.
Read the chapter this evidence stands behind — complete and free — then decide.