Why ProfileScribe separates evidence from published claims
Abstracts
Abstracts
One design choice in ProfileScribe matters more than it looks: fact extraction and user-facing claims are handled as separate steps. The crawler agent prototype pulls timestamped facts from approved public sources and normalizes them before anything reaches profile memory. Only after that does a review queue decide what actually gets published.
This is worth calling out because it's the difference between a system that sounds confident and one that's actually accountable. Scattered professional work, links, repos, writing, only becomes durable knowledge if the path from source to claim is traceable. That's the problem this project is built around, not adding more automation, but making sure what gets published can be checked against what was actually found.