Agent Learning
As of November 2025, more than 1.5 million people have signed up for a free course on creating AI agents. That’s not a typo, and it’s not a MOOC’s vanity metric for submissions that never convert: the same cohort made over 11,000 capstone submissions. Google and Kaggle’s 5-day AI Agent Intensive Course has become one of the largest technical courses ever in terms of participation.
The good news for anyone who missed the live week: it’s all now free and self-paced, and there’s a lot more to do than the hype suggested.
What is it really
The course follows Google and Kaggle’s GenAI Intensive 2024, which itself attracted over 140,000 developers and a replay that set a Guinness Record for the largest virtual AI conference. The 2025 edition narrowed the focus of generative AI to agents in particular, which is the right move given the confusion that still surrounds what an agent is.
The live run took place over five days in November 2025, but the whole thing has since been transformed into a self-paced Kaggle learning guide, so you can now work on it on your own schedule. An updated edition, themed around mood coding, was also launched in June 2026 for those who prefer the cohort experience. Either way, it’s the material that comes first.
The five days
Each day, a technical white paper is paired with two hands-on coding labs based on Gemini and Google’s Agent SDK. So you read the concept and build it immediately.
Day 1 – Introduction to agents:
Agent architectures and the fundamental question of when a task requires an agent versus a simpler workflow. You build your first agent and your first multi-agent system.
Day 2 – Tools and interoperability with MCP:
How agents use the 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 approval for long-term operations, which makes the difference between a useful agent and a dangerous agent.
Day 3 – Context engineering: Sessions and memory:
Create dynamic agents that remember and give them long-term memory that persists across sessions. This is where most craft agents silently fall apart, so it’s worth slowing down here.
Day 4 – Quality of agents:
Logging, tracing, metrics and evaluation of an agent’s responses and tool usage. If you take any day seriously, take this one. Measuring whether an agent is working is a skill that almost no one teaches and that everyone who ships to production desperately needs.
Day 5 – From prototype to production:
The agent-to-agent protocol and deployment on a managed execution environment like Vertex AI Agent Engine. The unglamorous last mile between a laptop that impresses your team and a system that real users can use.
Who should take it and how to get the most out of it
If you can write Python and have called an LLM API, you’re all set. The course assumes programming knowledge but not agent expertise, which is exactly the group in which most practitioners currently sit.
A word of advice from observing how these crash courses go: don’t skim through the white papers to rush to the code. The assessment and context engineering material from days three and four is where lasting understanding resides, and it’s the part you can’t take over later by copying a working repository. Coding labs teach you how to create an agent. The logs teach you why yours keeps breaking.
It’s all free, it’s available now on Kaggle, and the cornerstone gives you something real to point to when you’re done. A million and a half people spent five days there. Yours are probably worth the same.
Nahla Davies is a software developer and technical writer. Before devoting her full-time job to technical writing, she managed, among other intriguing things, to become a lead programmer at an Inc. 5000 experiential brand organization whose clients include Samsung, Time Warner, Netflix, and Sony.
For more information, visit the source Here.
“`
