Loading video...

Video Failed to Load

Go Home

Andrew Ng just released a free 2-hour course on complete Harness Engineering How to go from one prompt to a reliable system of agents that can run, test, and improve themselves: 09:14 - Build your first agent from scratch 33:11 - Master agent loops 1:02:46 - Turn loops into...

98,000 views • 6 days ago •via X (Twitter)

13 Comments

Kira's profile picture
Kira6 days ago

always happy to see some lectures by Andrew Ng

Ichigo's profile picture
Ichigo6 days ago

always happy to share some useful videos.

beamnxw ./'s profile picture
beamnxw ./6 days ago

from one agent to a full ai system

volovuk's profile picture
volovuk5 days ago

The term is legit, but ‘just released’ claims deserve a source link — actual course page, not just a screenshot with timestamps.

Roberto M's profile picture
Roberto M5 days ago

this is the link for the course:

Risk Rich's profile picture
Risk Rich6 days ago

does the course cover how to design evaluators when success cannot be reduced to a passing test? that is where reliable loops get much harder

Internet Labs's profile picture
Internet Labs5 days ago

Shifting focus from basic prompting to the surrounding infrastructure is what finally makes agents production-ready.

Chen Avnery's profile picture
Chen Avnery11 hours ago

good. the scaffolding around the model has been the actual job for a while and it mostly got learned by people breaking things in prod. nice to see someone teach it on purpose.

Kshitij Mishra | AI & Tech's profile picture
Kshitij Mishra | AI & Tech5 days ago

this is called gold

Shon Pazarker's profile picture
Shon Pazarker4 days ago

The jump is not from one prompt to many agents. It is from a demo that runs once to a system that can verify its own work.

Arash Rahimi's profile picture
Arash Rahimi5 days ago

the missing layer was never another prompt

Huysolo's profile picture
Huysolo5 days ago

Moving past single prompts into proper agent harnesses is literally the missing piece for reliable multi step software execution

Able Varghese's profile picture
Able Varghese2 days ago

Reliability comes from the harness, not the prompt. Which signal do you trust most for regression detection: task success, tool traces, or recovery behavior?

Related Videos