# Agentic learning
In November 2025, more than 1.5 million people signed up for a free course on building AI agents. That's not a typo, and it's not a vanity metric for MOOC enrollment: the same cohort included more than 11,000 keystone submissions. Google and Kaggle's 5-Day AI Agents Intensive was, by far, one of the largest technical courses ever conducted.
Good news for anyone who missed the live week: everything is now free and automated, and there's more to it than suggested.
# What It Really Is
This course follows Google and Kaggle in their 2024 GenAI Intensive, which itself attracted 140,000-plus developers and multiples that set a Guinness record for the largest virtual AI conference. The 2025 plan has narrowed the focus of AI from AI production in general to agents in particular, which is the right call given how much confusion still surrounds what an agent is.
The live run takes place in five days in November 2025, but the whole thing has turned into a self-paced Kaggle Study Guide, so now you can do it at your own pace. A revamped run, themed around vibe coding, is also out in June 2026 for those who prefer a group experience. Either way, what matters is the drawing.
# Five Days
Each day pairs a technical whitepaper with two hands-on codelabs built on Gemini and the Google Agent development kit, so you learn the concept and build quickly.
Day 1 – Introduction to agents: Agent properties and the important question of when a job needs an agent at all versus a simple workflow. You build your first agent and your first multi-agent system.
Day 2 – Tools and interaction with MCP: How agents use tools, how to write custom ones, and how the Model Context Protocol (MCP) allows agents to communicate with external systems. This day also introduces human-in-the-loop validation for long-lasting performance, which is the difference between a useful agent and a harmful one.
Day 3 – Content Engineering: Times and Memory: To create intelligent agents that remember, and give them long-term memory that persists across time. This is where most home agents quietly fall apart, so it's worth slowing down here.
Day 4 – Agent quality: Log, track, metrics, and evaluate both agent responses and their use of tools. If you take one day seriously, take this one. Evaluating whether an agent is working is a skill that almost no one can teach and everyone who sends a product to production needs it badly.
Day 5 – Production Prototype: Agent-to-agent protocol and deployment in a managed runtime such as the Vertex AI Agent Engine. It's the awkward last mile between a notebook that impresses your team and a system that real users can experience.
# Who Should Take It, and How to Get the Most Out of It
If you can write Python and call the LLM API, you're good to go. The course assumes planning knowledge but not the agent, which is the group most doctors currently live in.
One piece of advice from watching how these tools often work: don't rush white papers to rush to code. The testing and engineering context stuff on days three and four is where the long-lasting insights live, and it's the part you can't download later by copying the working repo. Codelabs teaches you how to build an agent. Papers teach you why yours keeps breaking.
The whole thing is free, available now on Kaggle, and the capstone gives you something real to point to when you're done. One and a half million people get five days. Yours is probably the same.
Here is Davies is a software developer and technical writer. Before devoting his career full-time to technical writing, he managed—among other interesting things—to work as a lead programmer at Inc. 5,000 branding whose clients include Samsung, Time Warner, Netflix, and Sony.
