How ScrapRival's reporting is made

Every fact in a ScrapRival teardown or briefing is a diff between two dated snapshots of a public page — a pricing page, a changelog, a blog. No press releases, no tips, no interviews, no speculation about private roadmaps. If a company changed a page, we can show the before and the after; if we can't, we don't make the claim.

The pipeline

Capture. Every morning (06:00 UTC) we fetch each watched page and compare it against the last stored snapshot; a new snapshot is stored whenever the content differs. Yesterday's version is what makes today's change visible — most of what we report was never announced anywhere else.

Diff. A change only exists when two snapshots differ. Each diff is classified by AI for relevance, so a rotating banner doesn't get reported as a price increase.

Verify, then write. The briefing is written by an AI chain with hard rules: facts must quote the captured page, and every computed figure (a "5.8x increase", a projected bill) must declare its arithmetic, which deterministic code re-checks. A claim whose numbers can't be verified is discarded rather than published.

Honest silence. A competitor is only described as "quiet" when every one of their watched pages was actually captured in the period. If we couldn't verify a page, we say exactly that — absence of data is not absence of change.

Corrections

When a published post needs a substantive fix, we correct it in place, mark the revision date on the page, and say what changed. Posts never silently rewrite their own history — that is precisely the behavior our teardowns exist to catch in others.

What we don't do

• No content from behind logins or paywalls — public pages only.
• No invented numbers: if a figure isn't on a captured page or derivable from one by declared arithmetic, it doesn't run.
• No "sources say" — the source is always a public page you can visit.

Questions about the method? Ask the founder directly — see About.