Andrej Karpathy has released a free, two-hour masterclass covering what he says he knows about AI agents, including agent harnesses, loops, and graphs.
AI agents · Andrej Karpathy
A free two-hour masterclass on how AI agents are built
Karpathy is offering a broad, open course on agent harnesses, loops, and graphs, aimed at developers and students moving from single prompts to working agent applications.
2 hrs
Total length of the masterclass
4
Named topics that form the main themes
Free
Open to anyone and meant to be shared widely
The core agent loop
The repeated cycle that agent systems run until a task is done
Observe
→
Act
→
Check results
↺ The cycle repeats, feeding each result back into the next observation
Harness
The surrounding software that directs a model, supplies it with tools, and handles its inputs and outputs.
Loops
The repeated cycle in which an agent observes, acts, and checks its results.
Graphs
Workflows in which tasks and decisions are linked as connected steps.
What the story establishes
The course covers how agent systems are built and organized.
It is free, open to anyone, and designed for a general technical audience.
It targets developers and students building agent applications.
What remains unclear
Hosting platform, full syllabus, and exact release date.
Depth and scope of the content, which cannot yet be verified.
Independent reviews or expert reactions.
Before you watch
Confirm the details on Karpathy's own channels.
Check the hosting platform, the full syllabus, and the release date there. Structure and follow-up materials will determine how useful the course is for practitioners.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...