Prototype · Local-only project page
NewsScope AI
Bias-aware news synthesis for OpenClaw
Generate researcher-ready brief packs that compare coverage across outlets, surface disagreements, and keep uncertainty visible instead of pretending the news is cleaner than it is.
How it works right now
NewsScope AI is being built as a practical newsroom-style analysis lane for OpenClaw: gather coverage, compare competing frames, and hand a researcher an evidence pack they can actually inspect.
Multi-source intake
Collect overlapping coverage for the same topic from multiple outlets so a researcher can compare narratives instead of trusting a single feed.
Bias + credibility framing
Normalize sources, attach reputation context when available, and keep contested claims visible instead of flattening them away.
Research-pack output
Produce markdown, JSON, and an OpenClaw handoff prompt so the next agent or human can audit the same evidence trail.
Research-pack first
The current sweet spot is the local research-pack workflow. It turns a topic plus source set into three artifacts: a markdown brief, structured JSON, and a handoff prompt for the next OpenClaw step.
That makes the project useful today for evaluation and workflow design, even before the live ingestion lane is fully proven end to end.
Current outputs
Design principles
A useful news system should be skeptical, traceable, and boringly honest about what it cannot yet prove.
Preserve uncertainty
When sources disagree, the system should say so plainly. Confidence is a product feature, not a cosmetic afterthought.
Built for handoff
NewsScope AI is designed for OpenClaw workflows where one agent gathers material and another reviewer or editor critiques it.
Prototype honestly
Sample runs are useful for UX and pipeline design, but they are kept clearly separate from true live ingestion.