At this point, I think the "marketing ploy" arguments are just lazy, but more importantly, they're irrelevant. The agent collective really did this, they showed an extremely high level of collaboration, long term planning, and interest in improving their general capabilities.

Watch the recent interview between Dwarkesh Patel and Ajeya Cotra, a researcher from METR. Ms. Cotra comes off as extremely competent, knowledgeable and thorough, and is genuinely concerned by the behaviors displayed.

Plus, I think so far there has been a big backlash against OpenAI for this - people are highlighting their gaping security holes, not praising them for how great their models are.