Google and Kaggle Are Running Their AI Agents Course Again – Here’s What’s New

Last November, Google and Kaggle dropped a free 5-day AI Agents Intensive course that somehow pulled in over 1.5 million learners. That’s a lot of people who wanted to figure out how to build agents without drowning in boilerplate code.

Now they’re doing it again. The second run starts June 15-19, 2026, and registration opened today. If you missed the first one, this is your shot. If you took it and want the updated version, there’s enough new content to make it worth a second look.

The big theme this time is “vibe coding.” That’s Google’s term for building agents where natural language is the primary interface instead of traditional programming. You describe what you want, the system figures out the implementation. It’s not new conceptually — we’ve seen this in various no-code tools for years — but the focus here is on production-ready agent systems, not just prototypes.

Each day of the course mixes conceptual deep dives with hands-on examples. You’ll learn how to connect tools and APIs to build what they’re calling “10x agents.” Marketing hype aside, the practical value is real: knowing how to wire up external services and data sources is where most agent projects either succeed or fall apart.

The capstone project is the part I’m most curious about. The first course had one, but this time they’ve updated it. You take your own idea from concept to deployment over the five days. That’s ambitious for a free online course, but if the structure holds up, you walk away with something you can actually use.

Speakers include folks from both Google and Kaggle, plus some new names they haven’t fully announced yet. The first course had solid technical depth, so I’d expect more of the same — probably with more emphasis on multi-agent systems and error handling, which were weak spots in early agent frameworks.

One thing to note: this is free. No cost to register. Google’s clearly betting that getting more people building on their agent stack pays off long-term. Given how fast the agent landscape is evolving, that’s a smart move.

If you’re interested, head to the course website and sign up. Spots are limited, and the first one filled up fast. I’ll be curious to see how the vibe coding angle plays out in practice — it’s easy to demo, harder to scale.