Andrew Ng: Mastering Coding Agents Is a Core AI Skill — Plus GLM-5.3-Flash and a 397B Legal Model
DeepLearningAI · x · 2026-09-05
Highlights from this week's The Batch newsletter:
- Coding agents as a key AI engineering skill: Andrew Ng argues that steering coding agents — for both code and non-code tasks like data analysis or ops — is evolving faster than any other top-level AI skill and requires continuous experimentation. After interviewing dozens of top AI engineers, his team found a consistent high-level workflow:
- Planning: brainstorm (research, experiments, understanding the codebase), write a spec (requirements, design, architecture), generate an execution plan, then interrogate assumptions and check for security and overengineering.
- Execution: balance agent autonomy with human oversight, verifying output via automated and/or human checks.
- Deployment & monitoring: deploy with CI/CD gates or human gates, then monitor.
- Enterprise privacy updates: OpenAI and Anthropic rolled out new enterprise data retention policies.
- GLM-5.3-Flash: Zai released a cost-efficient, open-weights multimodal model.
- Ox Alpha: Thomson Reuters launched a 397B-parameter model trained specifically for law, finance, and news work.
Related event: Andrew Ng: mastering coding agents is a core AI engineering skill(2 posts)→
More from coding & agent
- Two years after Devin launched: Cognition's task success climbed from ~30% to 90% — silasalberti · 2026-09-05
- I built a useless Gemini chatbot, then added a "teaching layer" of lessons — now it works — Realistic-Middle7168 · 2026-09-05
- A Multi-Model "Plan-Implement-Review" Workflow Wired Through Claude Code — dr_cintas · 2026-09-05
- Viral dev joke: AI has solved writing code, not meetings, requirements or DNS — StewartalsopIII · 2026-09-05
- Git worktree workflow for running multiple branches and coding agents in parallel — JeremyCMorgan · 2026-09-05
- One dev's agent idea-exploration workspace: git worktrees plus append-only memory files — viksit · 2026-09-05