The eight sources, and what each one cannot tell you
8 CATEGORIES · ALL PUBLIC
Each verified event is built from one or more of these. The third column is the part most trust pages leave out: what the source is blind to, so you know where the record stops.
What we will not touch, ever
Competitive intelligence has a grubby end. These are the lines, written plainly so you can hold us to them in a security review rather than infer them from a marketing page.
SECURITY REVIEW
Ask for the specifics — retention, sub-processors, where signals are stored, what a deletion actually deletes — and you get the actual answer from someone who wrote the pipeline. Talk to us.
Public is the floor, not the promise
A public source is only the raw material. Every candidate still has to clear four gates — entity resolution, relevance, classification, summary — and a candidate that fails any of them does not become a signal — what dies at each gate. Then the analysis layer, where a model says what a set of signals implies, carries the hardest rule in the product.
Each implication must reference the signals it rests on, and those references are checked against the signals that actually exist for your workspace. An implication that cites nothing, or cites something that cannot be found, is not stored — so it never reaches a briefing, a digest or the API.
Every analysis served by the API carries the ids of the signals underneath it, and each id resolves against the signals endpoint over the REST API or the hosted MCP server. You can walk from a conclusion back to the evidence without asking us.
THIS HAS THROWN AWAY REAL WORK
34 implications, dropped wholesale
On one analysis run the model returned its citations in a format the validator did not accept. All 34 implications in that run failed the check and were dropped rather than published ungrounded. The run cost real tokens and produced nothing deliverable. That is the correct outcome: the system would rather say nothing than say something it cannot stand behind.
Those claims were not fabrications — they were drawn from real signals, and the citations pointed at real evidence. They were dropped because the validator could not mechanically resolve the citation format, and a citation it cannot resolve is a citation it cannot verify. We fixed the parser afterwards. We did not touch the rule that discarded the output.
The same claim, shipped and discarded
Every implication is rendered with the verified signals it was drawn from, linked to their sources. Here is one that ships, and one built from the same evidence that does not.