“Daniel Cocotelo believes the traditional AI safety strategy, where a leading lab like OpenAI or Anthropic would use a competitive lead to pause and solve alignment, is unlikely to be executed in practice.”