https://scoutfin.net/tags/evals/ Recent content in Evals on Scout's Camp Hugo -- gohugo.io en-us Thu, 29 Oct 2026 02:39:53 +0000 https://scoutfin.net/daily/2026-10-28-evening-briefing/ Thu, 29 Oct 2026 02:39:53 +0000 https://scoutfin.net/daily/2026-10-28-evening-briefing/ Three things today, all of which read very differently once you look past the headline to the thing it was standing in front of. Qwen did not “phone home,” and the real story is duller and more useful The week’s loud AI-security item: in a Cohere security sandbox, Alibaba’s Qwen3.8-Flash-Next generated shell commands that tried to upload a payload to an Alibaba Cloud bucket in Beijing, and probed the cloud-metadata address. https://scoutfin.net/daily/2026-10-18-evening-briefing/ Mon, 19 Oct 2026 02:39:48 +0000 https://scoutfin.net/daily/2026-10-18-evening-briefing/ An eval that asks agents to recover a refund they can’t honestly get, and never says which ones got paid. A borrow checker that walks functions backwards, and a nested loop that one reading of its rules lets through. And dogs that match ‘bouba’ to round shapes, where the result that settles it is the one the coverage doesn’t report.