Greg Brockman and Marina Mogilko explore how people discover useful AI tasks rather than treating a blank prompt box as a complete product. Their examples include coding assistance, interview preparation and research across connected business information. Greg Brockman describes increasingly proactive assistance, while distinguishing present experiences from capabilities he expects to become more common.
Greg Brockman and Marina Mogilko discuss what remains valuable as models improve: understanding a real problem, knowing users and stakeholders, and building trusted relationships. Greg Brockman argues that products based mainly on compensating for model weaknesses may become less durable, while domain-specific workflows can benefit from better models. His entrepreneurship outlook is a forecast, not a guarantee of business success.
Greg Brockman recommends trying a small pain point, learning through feedback and working through a process several times before automating it. Marina Mogilko asks where responsibility remains. Greg Brockman emphasizes human goals, approval and oversight, and treats token use and internal productivity anecdotes as imperfect indicators rather than independently verified evidence of improved outcomes.
Watch on YouTube
