About

How we know what we know.

How We Know is an evidence-first publication about the deep ocean. Each page answers one real question and then does the part most writing about the deep sea skips: it names the instrument, the proxy or the single observation the answer rests on, and says where the evidence stops. Every published answer is paired with a video that makes the same claim in the same words, so the two surfaces cannot drift apart.

Who publishes it

How We Know is the publisher of both this site and the @howweknowdeep video channel. The two carry one name, one editorial standard and one correction process. Nothing here is sponsored, and nothing on the site is for sale: it is free to read and free to quote with a link back.

Where the evidence comes from

Figures are traced to named public research — NOAA, MBARI, Woods Hole Oceanographic Institution, the Smithsonian and the published literature those bodies cite. A source is recorded with the date it was checked, and every page lists the sources behind it rather than gesturing at “studies”. Where a number is contested, the page gives the range and says who measured which end of it.

How an answer is checked

Questions are admitted, sourced and reviewed under a written process, not an editor’s judgement on the day. The methodology sets out how a question is admitted, how sources are selected and how uncertainty is handled. The editorial policy sets out the review and correction rules: a corrected claim is changed in the canonical record, the article, the script, the transcript, the video description and the structured data together, because a correction that reaches only one surface has not happened.

What the site will not do

It will not convert an unknown cause into a confident one, and it will not turn a single viral sighting into a trend. Unusual claims stay labelled as single observations until repeated evidence exists — which is also why answers are grouped by how they are known rather than only by subject. A figure measured by an instrument and a figure inferred from a proxy are different kinds of fact, and the site says which kind it is handing you.

Corrections and questions

If a figure here is wrong, the page that carries it is the thing that changes, along with every other surface that repeats it. Review dates are carried on the pages themselves, and a page is re-reviewed when a material fact behind it moves. Start with the questions or the depth explorer, and the llms.txt file lists every canonical answer in one machine-readable place.