Anthropic's 'Slow Down' Essay Reads as Competitive Moat

Dario's 'pace the frontier' essay landed on HN and the reception was almost uniformly skeptical. The top responses called out the obvious tension: Anthropic is not proposing to slow down its own model development, just calling for regulation that would make it harder for competitors, especially open-weight ones, to catch up. One commenter noted that the logical proposal, releasing open weights plus training details, is exactly what Anthropic does not offer.

This connects directly to a parallel thread titled 'Everyone should slow down AI development except for me,' which is basically a one-sentence summary of the community's read on the situation. Ed Zitron's framing, cited in the P(doom) thread, put it bluntly: OpenAI and Anthropic spend enormous energy warning about AI in the wrong hands, but it's already in their hands, and neither company has demonstrated it deserves special trust.

The counterpoint is real but weak: these companies do employ serious safety researchers, and some of the concern is genuine. But the structural incentive problem is too obvious to ignore. When the entity calling for regulation is also the entity that benefits most from barriers to entry, the call deserves scrutiny regardless of sincerity.