"Ng said he rarely sticks to a single chatbot. To brainstorm effectively, he rotates across different models and leans into their contrasting strengths. For coding, he prefers tools like Claude Code and OpenAI's Codex. He added that staying longer in a conversation with the model yields a better response. "AI is very smart, but getting context in is difficult," Ng said."
""It's sometimes faster to be lazy and dash off a quick, imprecise prompt and see what happens," Ng said on X in April. "Most LLMs are smart enough to figure out that you want them to help understand and propose fixes, so you don't need to explicitly tell them." "We add details to the prompt only when they are needed," he added."
Andrew Ng uses voice-mode AI while driving as an active brainstorming companion, conducting extended back-and-forth conversations rather than issuing single prompts. He rotates across multiple models to leverage differing strengths, using Claude Code and OpenAI's Codex for coding tasks. He emphasizes that longer conversations provide better context and responses, and that AI can summarize exchanges and forward them to his team after a drive. Ng also advocates occasional 'lazy prompting'—sending quick, imprecise prompts first and adding details only when necessary—because many LLMs can infer intent and propose fixes without exhaustive instructions.
Read at Business Insider
Unable to calculate read time
Collection
[
|
...
]