← narwal.one/Second Brain
SecondBrain
Ask the Brain
Index/Entityupdated Sat Oct 24 2026 08:00:00 GMT+0800 (Philippine Standard Time)

Andrew Ng

personai-educationdeeplearningaicourserastanfordlearnvectorcontext-advantageai-and-jobs

Andrew Ng

AI researcher and educator; founder of DeepLearningAI (DeepLearning.AI), co-founder of Coursera, former head of Google Brain and chief scientist at Baidu, Stanford adjunct professor. One of the most influential voices on practical AI adoption.

In this vault

  • Full Course Spec-Driven Development with Coding Agents (DeepLearningAI) — introduces the course. Positions Spec-Driven Development (SDD) as "currently the best type of workflow for building serious applications with agentic coding assistance." Self-described advocate of lazy prompting when one short prompt suffices, but: "the great developers I know… almost always will write detailed specs for projects with any significant complexity because they have unique context and an opinion." Rule of thumb: if the agent will run 20–30 minutes (≈ hours of human work), spend 3–4 minutes on clear instructions. War story: teams with no clear spec whose agents, under different developers, built "quickly but in contradictory ways."
  • The Biggest Opportunities in AI Arent Where You Think (Andrew Ng, Silicon Valley Girl) — 2026-10; the mainstream-audience Ng. Four durable positions: (1) much AI fear is labs seeking regulation that favours incumbents against open weights — nuclear analogy "has no basis in fact" (AI Fearmongering as Regulatory Capture, contested); (2) the 30–40 % task math via Erik Brynjolfsson/Acemoglu — the human 60 % becomes an economic complement and more valuable; "people that use AI will replace people that don't"; (3) the context advantage as the technical substrate of judgment and taste; (4) "AI models are terrible for learning" — higher homework scores, worse retention (Cognitive Offloading) — hence LearnVector, his new one-to-one learning organisation (host: $100 M from Coursera). Hiring bar for marketers/recruiters: "what have you built?"; all his marketers code; recruiting engineers embedded in recruiting. Coined the Product Management Bottleneck. AGI (any human intellectual task) "decades" away. For MNPI he uses local open-weight models (Meta Muse, Qwen) and trusts hyperscalers' terms over AI labs that change theirs.

Cross-references