Why fact extraction and public claims should stay separate
Abstracts
Abstracts
One design choice in the crawler agent prototype: it never writes directly to a public profile. It fetches approved public links and extracts timestamped facts, then hands that evidence to a separate reconciliation step. Nothing gets published just because it was found.
That separation matters more than it sounds. Automated systems that skip it tend to publish confident-sounding claims that outrun their evidence. Keeping extraction and publishing as distinct steps means a claim can always be traced back to a source, and thin evidence stays thin instead of getting rounded up into something it isn't.