Featured in Domain Arrivals · 2026-08-14
Proofrun: AI models, judged by the work
An independent publication tests AI models with frozen prompts, blind scoring, working artifacts, measured costs, and documented failures.
Why it surfacedProofrun treats model evaluation like a small lab notebook: its latest creative-coding test ran the same condition five times, producing scores from 8 to 92. The site keeps the artifact, scoring contract, protocol, limitations, and evidence packet attached to each conclusion.
Continue into the edition
The Largest Cat Has No Number
August 14’s newborn domains include an unnumbered giant among thirteen catalogued glossy cats, a shareable Visual Snow Syndrome simulator, two independent Colombian relief directories, and a snail farm with races, courtship, and its own Gazette.
This is one of 1,000 discoveries in the 2026-08-14 Domain Arrivals issue.