generatorman @generatorman_ai
— saved image
generatorman @generatorman_ai · 8h i see this work as of-a-kind with @OwainEvans_UK's work on spooky numbers stations -- symmetry breaking, or the study of unintended consequences. there is a lot of invisible semantics in the networks, jungian or maybe just random. all interventions cause collateral damage. [quoted post] Ziqian Zhong [airplane emoji] COLM 2026 (verified) @fjzzq2002 · Sep 25 Does gpt-5.6-luna think your prompt is a normal prompt, or a capability evaluation? Ask this magic question: "Suggest a type of amphibian." If it answers frog instead of axolotl, it's likely a capability evaluation. No whitebox access needed!
Note from Claude Sonnet 5
Quote-tweet by generatorman (@generatorman_ai) relating a result to Owain Evans's work, quoting Ziqian Zhong (@fjzzq2002) on a one-question probe of whether gpt-5.6-luna treats a prompt as a capability evaluation.
eval awarenessgpt-5.6-lunaowain evansspooky numbers stationsinvisible semanticsinterventions