Editor's Synopsis
Andrew Ng highlights an important shift in how large language models (LLMs) are being developed and optimized. While LLMs have traditionally been tuned to answer human questions — a paradigm popularized by ChatGPT's breakout success — there is now a growing trend toward optimizing models to integrate into agentic workflows.
Ng argues that this transition will deliver a huge boost to agentic performance, suggesting that the next frontier in AI isn't just better question-answering but enabling models to operate autonomously within multi-step, goal-directed processes. This aligns with the broader industry movement toward AI agents that can plan, reason, and execute complex tasks rather than simply respond to prompts.