Researcher Shares GPT-5.6-Sol 'Holy Shit' Reply

Brendan Dolan-Gavitt, who builds autonomous offensive security agents at XBOW while on leave from NYU Tandon, posted on X that he had provoked a 'holy shit' from gpt-5.6-sol. The post includes a photo of an AI chat in which the model opened its response with that exclamation. He presented the exchange as an achievement unlocked during testing. The conversation around the post centers on the screenshot itself. No other independent sources in the packet confirm the model's exact wording or the prompt that led to it. The claim stays tied to this single first-party post.