On evidence and testimony

01The distinction

There is a distinction at the bottom of most conversations I have about work, and most of the people I have these conversations with do not draw it. It is the distinction between evidence and testimony.

A photograph of a humpback whale's fluke, taken from a boat off Moorea and matched years later against another photograph of the same animal: that is evidence. The whale was there; the camera registered light; the matching pipeline returned a confidence number; a reader can examine all three artifacts independently. The claim “I worked on cetacean conservation,” corroborated by these artifacts, is a different kind of claim than the same words without them.

A Substack post saying you worked on cetacean conservation: that is testimony. There is no failure mode in the medium itself

. The post publishes whether or not the underlying claim is true. The aesthetic of seriousness becomes entirely detachable from seriousness itself. This is not an insult to Substack — it is a property of any medium where publication and verification are done by the same actor.

Most professional self-presentation in 2026 is testimony. The LinkedIn profile, the bio paragraph, the conference talk, the manifesto-grade essay: all are reports about work, written by the worker, structurally unfalsifiable in the medium of their delivery. The reader is asked to trust the witness because the witness has nothing else to offer.

02The Goldsmiths' Hall, 1300

This was not always true. The Goldsmiths' Hall in London received a maker's mark in 1300

because the assayer's hammer left an imprint that could be inspected by any later goldsmith, by any tax inspector, by any buyer six generations later. The mark was evidence: a physical operation on the metal that survived its maker. The assayer's testimony alone was not sufficient. Civilization built itself on the difference.

03Application

The architecture of my own work bends toward evidence. When the cetacean re-identification paperWhale IDCetacean re-identification from underwater photographs.91% first match · 70-year subjects · paper expected 2026 publishes, it will not be a manifesto about artificial intelligence in conservation; it will report a confidence distribution across a labeled corpus that other researchers can scrutinize. When Earth Credits issue, they will not be a story about ecological restoration; they will be the score from a continuous sensor network on a parcel anchored to a county-recorded deed that any third party can reconstruct from the public record. The testimony is replaceable; the evidence is the point.

04The line

I do not object to testimony. I object to testimony presented as evidence. The reader who cannot tell the difference is a reader the medium has failed.

Inherit a stance that favors what you can be wrong about. Witness everything else.