Andrew Ng shares his AI brainstorming playbook — a lot of it happens in the car (www.businessinsider.com)

At the Masters of Scale Summit 2025, Andrew Ng revealed he regularly brainstorms by talking to AI in voice mode while driving, treating models as “brainstorming companions” and routinely switching among multiple systems to exploit their complementary strengths. He said effective sessions are extended conversations rather than one-off prompts — you feed context, iterate, give feedback, and then ask the model to summarize the exchange and forward it to teammates when you arrive. For coding tasks he favors tools like Anthropic’s Claude Code and OpenAI’s Codex, and he noted that staying longer in a dialogue generally yields richer, more actionable outputs. Ng has also endorsed occasional “lazy prompting” — tossing out a quick, imprecise prompt to see what the model surfaces, then refining only if needed. Ng’s pattern highlights practical trends for the AI/ML community: voice interfaces and mobile workflows are becoming viable for real work, multimodel toolchains are valuable for different task types, and conversational context management is a key skill for better outputs. The approach underscores the importance of persistent context, summarization and handoff mechanisms for team workflows, and pragmatic prompt strategies. It also signals growing executive acceptance of AI as an integrated collaborator rather than a narrow utility, shaping how researchers and builders prioritize usability, model interoperability, and context retention.