Andrew Ng's AI Engineering Skills Map (The Batch #366): 'Using Coding Agents' Is Now a Named Core Competency for All Developers, Not Just AI Engineers
The Batch issue 366, dated 2026-08-14, leads with Ng's four-skill map: building and deploying AI applications (LLMs, RAG, agentic workflows, plus statistical technique to measure and govern them); software engineering fundamentals for cost/scalability/reliability tradeoffs; using coding agents — building mental models of agent limits, managing context, and knowing when to intervene versus delegate; and 'shaping the build,' meaning product strategy and deciding what the spec should contain. Ng's explicit claim is that all developers need these, not only people with 'AI Engineer' in the title, and that the role shifts from execution toward decision-making. Same issue covers Meta's Muse Code, Google's Gemini Robotics 2, and MiniMax H3.
↳ Follow the thread