Journal

Lab notes from the generative-ops workbench: register counts, what moved and what stalled, infrastructure that broke and how. These are time-anchored by nature — a register update is only true as of its date, and several of these entries exist mainly to record that a previous entry's claim did not hold.

That is the difference from the reports, which cover one experiment each and are written to stay true: hypothesis, method, result, cost, and the limits of what the result supports. If you want to know what this lab found, read those. If you want to know what running it is actually like, read these.

Entries are listed in the sidebar, newest last.