This Is Not a Drill: What Anthropic’s Mythos Means for Asset Owners

October 15, 2026April 16, 2026 by Marius A. Holzer, CFA, CAIA, CIPM

An AI model held back from release because it is too good at finding software flaws is a siren, not a test. For family offices, with large fortunes and small teams, here is what to do before next quarter.

A month building an AI analyst: three lessons that weren’t on my list

October 15, 2026April 10, 2026 by Marius A. Holzer, CFA, CAIA, CIPM

After a month, my AI analyst was better at spotting its own improvements than I was, the cheap model was often the right one, and its memory belonged in files, not chat.