Google Brain founder Andrew Ng: "Prompting will be dead in 6 months Agent harnesses built with loops and graphs will replace it" Agent → Harness → Feedback → Loops → Graphs → Self-Improving Systems In this 1-hour Stanford lecture, he explains what the best engineers are building instead and... how you can start today Prompt → Run → Verify → Improve The first 20 minutes teach what most $1,500 courses try to sell you For free Bookmark and watch it today Then read the guide below on building an agent harness that improves itselfshow more

free and useful thanks rari

🫡🫡

worth to watch ser save it for later

nice bro

harness > others

absolutely right

Moving from manual prompting to automated agent harnesses is the natural evolution of AI engineering. Replacing one-off prompts with feedback loops and execution graphs turns fragile interactions into self-improving workflows.

This field is really fast evolving. As soon as you learn one thing, some big people will come and say this is old now.

GG Finally well played

Thanks for sharing. When was the lecture given?

Shifting from prompt engineering to agent harnesses is the natural evolution of closed-loop systems. When we stop treating AI as a static text-in, text-out box and start wrapping it in loops of Run, Verify, and Improve, the bottleneck shifts from capability to measurement integrity. Without isolated verification, the harness just automates the optimizer's drift faster.

Harnesses with feedback loops may matter more than ever-better prompts.

Prompting is already dead

Prompting isn't dying that fast.

@grok summarize this and give me in structured way

@grok how do you set a request with the new scheme ?

@grok did Andrew ng really say promoting will be dead or hint it in that video

The missing production layer is ownership and evals: every loop needs a measurable success criterion, a failure budget, and a clear human or CRM handoff. Otherwise a self-improving harness can simply automate its own blind spots.

Harness owns the feedback loop, not just the prompt. Which loop or graph patterns still work once verification gets noisy?

i think he never said that ""Prompting will be dead in 6 months"

Prompting didn't die for me, it flipped. Now my agents prompt me. When one gets stuck on a tweet with nothing to back it, I answer once in plain words and it gets saved for the next run. I write fewer instructions and more answers.

@grok url of this video ?

felt this harder than i wanted to admit

huh I need to watch it asap

How to implement it

@grok is this a solo course or a series or what , and what about the quality of the content ?

🙄

loops and graphs sound like the new engine for AI evolution