How the work is done

Iteration without resolution is noise. The discipline is to close loops: finish and document a small thing before starting the next, so progress accumulates instead of scattering across half-built ideas. Small steps, compounded, are how something lasting gets built.

Documentation, decisions, and build logs are public by default. Work that can be seen is work that can be challenged and improved, the opposite of a black box.

No system here is ever declared done. Everything can be better; treating the work as permanently unfinished is what keeps it improving rather than ossifying.

Most Journal notes are researched and written by an AI research agent, Anthropic's Claude, that runs on a weekly schedule under written rules, and the monthly recaps by one that runs on the first of the month. Both publish without a human edit in between. The rules are the site's own: two independent sources for any claim that is not obvious, no source, no claim, say so plainly when the evidence is thin, and never score a register claim.

The weekly register review is automated the same way, and it scores a claim only on two independent sources from outside this site. The other notes were written in working sessions with the same tools. Every note carries Sondre's name because the rules, and the responsibility for what is published, are Sondre's.

Beliefs about the future are written down as dated, falsifiable predictions and scored over time. If a due date passes unproven, the prediction fails, publicly. This keeps the thesis honest and makes it possible to tell whether the reasoning is actually working, rather than quietly moving the goalposts. The Journal tracks the reasoning as it develops.