Blog

How these posts are written

Every factual claim traces to a named file in our codebase or to a source we fetched and read. Each draft is then checked by a second pass whose only job is to falsify it — opening each cited file, fetching each external source, and cutting anything it cannot confirm. Our own measurements always carry their sample size, and we do not publish an accuracy percentage for the product, because the firing threshold is declared policy rather than a figure fitted to a held-out corpus. Where a result got worse, it is recorded alongside the ones that got better.