“OpenAI's first product strategy for GPT-3 was to release an API and allow the market to discover the use cases, a project Greg Brockman felt was "totally doomed" at the time.”