Andrew Ng: Steering Coding Agents Is Now a Core AI Engineering Skill
AndrewYNg · x · 2026-09-04
Andrew Ng published a new article arguing that using coding agents is a key AI engineering skill — and one that's evolving faster than any other top-level skill.
Key points:
- Steering agents covers non-code tasks too, like data analysis and system operations;
- Proprietary agents (Claude Code, Codex, Cursor) and open agents (OpenCode, Pi) improve via both harness and model upgrades, demanding continuous experimentation;
- After interviewing dozens of top AI engineers and reflecting on his team's practice, he distills a consistent high-level workflow, with Planning as a crucial step.
More from coding & agent
- Intent update coordinates swarms of agents with workspaces across devices — LukeW · 2026-09-04
- LukeW's Intent update: multi-agent coordination, workspaces, multi-device runs — LukeW · 2026-09-04
- VC advice for agentic startups: pitch how your agent recovers from failures — atShruti · 2026-09-04
- WebMCP demo: an AI agent DJs in the browser with hundreds of direct tool calls — No_Guide_8697 · 2026-09-04
- Newsletters, not killer use cases, are what get users to trust agents with real email access — JanJanJaJa · 2026-09-04
- Kafka, Kafka Connect and Schema Registry exposed as native MCP tools — jkriket · 2026-09-04