Dario Amodei says his essay “We Must Pace the Frontier” lays out a three-part plan for slowing the AI industry. He says Anthropic is committing to the first step: giving third-party evaluators permanent, employee-level access to its systems to verify adherence to safety measures, report incidents and assess models’ alignment during training. One commenter supports moving toward safety-based pacing but argues evaluations and practices seem harder to define and easier to game than compute-allocation requirements. As an example, the commenter suggests requiring 90% of compute to go to external inference—running models for outside users—rather than research and development. In a follow-up, the same commenter also favors limiting the capabilities of models used for AI research and development.