What is this answer built on?

Answers arrive without their sources

A search result showed you where a claim came from. You could see the domain, the date, the ad label. An answer shows you none of that. It reads as one confident paragraph, and the difference between a well-sourced fact and a paid placement disappears into the same sentence.

Some of what surfaces is bought. Some rests on a single post from 2019. Both look exactly like the part that is solid.

What Flamingo reads

Placement

Whether money moved for this to appear. Sponsored content, affiliate text and seeded reviews are written to survive summarisation, which is precisely why they need marking at the point of use.

Support

How many independent sources stand behind the claim, and whether they are genuinely independent or the same release quoted five times.

Age

When the underlying data was published, and whether anything has superseded it since.

Dispute

Whether credible sources disagree. A contested claim is not a false one, but it should never arrive sounding settled.

The mechanic

One hover. No click, no tab, no interruption of the thing you were doing.

Reading provenance has to cost less than ignoring it, or nobody reads it. Flamingo sits on the surface where the answer already appeared, in the terminal or the chat window, and answers on hover. If a claim is clean it stays quiet.

Discovery, not placement

Flamingo detects advertising. It does not sell it. The distinction is the product: a tool that both marks paid content and takes money to place it cannot be trusted to do the first job, and would sit badly beside the rest of what Tantile builds.

The other products here exist to make what agents do inspectable, through audit trails, human approval and spending a person can see. This applies the same idea one layer up, to what agents read.

Where it is

Early. The detection backend and the operator portal exist. The surfaces people would actually use are not built, and there is nothing to install yet.