Structure, not summaries.

YouTube is the largest body of expert knowledge ever assembled, and it is nearly unusable as knowledge. The talks, lectures, and deep-dives that matter are trapped in a feed format built for watching once, not for learning from, citing, or querying. The information is in there; the structure is not.

Liquid Gold exists to fix the structure, not the watching. Any public channel can be distilled: videos, transcripts, and topics become a connected knowledge base — entities, claims, and the sources that back them.

Master distillers extract and concentrate the essence of raw materials, discarding the noise. Liquid Gold does the same to a YouTube channel: the signal — what a channel actually knows, claims, and teaches — is separated from the noise of delivery, repetition, and filler.

The result is meant to be tapped directly: browsed by a curious mind, queried live by an AI agent over MCP, or owned offline as an archive.

Fetching a transcript is commodity — dozens of projects do it. The product is what happens after: turning a raw transcript into entities, claims, and relations that merge into one growing graph instead of sitting as an isolated per-video silo.

That discipline holds on both sides of the product. Every claim cites the video that supports it and says no more than the source does — the same rigor an answer demands at query time has to hold at authoring time, because a hallucinated claim doesn't just answer one question wrong, it becomes permanent corpus content.

Liquid Gold is built by BourdonLab. It is an independent service that works from public YouTube data; the creators own their content.

Say hello: hello@bourdonlabs.com.