Howard Marks asks Claude to grade AI predictability, and frontier labs score 2 out of 10

Howard Marks asked Claude to grade the predictability of frontier AI labs, and the models returned a score of just 2 out of 10, underscoring how little visibility investors have into the sector's trajectory. The exercise, reported by Crypto Briefing, frames that unpredictability as a material risk for creditors and argues for more cautious financial strategies. The low score matters because AI-linked debt and equity are increasingly priced on assumptions about future capability and cash flows that even leading models cannot reliably forecast. What to watch next is whether lenders and allocators adjust credit terms and valuations to reflect that uncertainty.