Keeping fact extraction separate from the claims a profile makes
Abstracts
Abstracts
The crawler agent prototype fetches approved public links and pulls out timestamped facts, but it stops there. Those facts get normalized before they ever reach the part of the system that writes user-facing copy. That separation is deliberate: it's easy for an automated system to blur 'here's what a source says' with 'here's what we're claiming about you.' Keeping extraction and claim-writing as distinct steps makes it possible to trace any published sentence back to something a source actually said. That traceability is the boring, necessary part of building AI systems people can trust with their professional record.