Probably the best outcome for open model optics rn is if OpenAI and Anthropic refuse to shoulder the burden of hardening all soft targets (or can't), and people are forced to audit their own systems with abliterated 1017's, making it clear that loyal AGI is a necessity.