Barbados, cross-checked in real time
Information about what's happening in Barbados is scattered across radio stations, TikTok clips and newspaper websites — often arriving at different times and with different details. Pulse listens to all of it at once, cross-checks the signals against each other, and tells you what's confirmed, what's unverified and what still needs a human look.
How a signal becomes a briefing
Every result you see carries
Confidence labels
Pulse hears an event mentioned on the radio and then sees it filmed on TikTok. Because two independent sources agree, it shows up as high confidence. If only one source has mentioned it, the claim appears as medium or needs review until more signals arrive.
Under the hood
How the pipeline, adapters and confidence model fit together.
CrewAI pipeline — six agents
Source adapter pattern
Every source — radio, TikTok, newspaper — implements the same adapter interface. The pipeline receives a normalised signal object regardless of where it came from, so adding a new source means writing one adapter, not touching the agents.
Confidence model
Score = source trust tier × recency × number of corroborating sources, minus a penalty for missing required fields, capped downward for claims flagged as sensitive topics.
Build on Pulse