# 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 MOOC’s vanity metric for registrations that never convert: the same cohort filed more than 11,000 capstone submissions. Google and Kaggle’s 5-Day AI Agents Intensive became, by attendance, one of the largest technical courses ever run.
The good news for anyone who missed the live week: all of it is now free and self-paced, and there’s a lot more to it than the hype suggested.
# What It Actually Is
The course is Google and Kaggle’s follow-up to their 2024 GenAI Intensive, which itself drew 140,000-plus developers and a rerun that set a Guinness record for the largest virtual AI conference. The 2025 edition narrowed the focus from generative AI broadly to agents specifically, which is the right call given how much confusion still surrounds what an agent even is.
The live run happened over five days in November 2025, but the whole thing has since been transformed into a self-paced Kaggle Learn Guide, so you can now work through it on your own schedule. A refreshed run, themed around vibe coding, also went out in June 2026 for those who prefer the cohort experience. Either way, the material is the draw.
# The Five Days
Each day pairs a technical whitepaper with two hands-on codelabs built on Gemini and Google’s Agent Development Kit, so you’re reading the concept and then immediately building it.
Day 1 – Introduction to agents: Agent architectures and the fundamental question of when a task needs an agent at all 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 tools, how to write custom ones, and how the Model Context Protocol (MCP) lets agents talk to external systems. This day also introduces human-in-the-loop approval for long-running operations, which is the difference between a useful agent and a dangerous one.
Day 3 – Context engineering: Sessions and memory: Building stateful agents that remember, and giving them long-term memory that persists across sessions. This is where most homemade agents quietly fall apart, so it’s worth slowing down here.
Day 4 – Agent quality: Logging, tracing, metrics, and evaluating both an agent’s responses and its tool use. If you take one day seriously, take this one. Measuring whether an agent works is the skill almost nobody teaches and everyone shipping to production desperately needs.
Day 5 – Prototype to production: The agent-to-agent protocol and deploying to a managed runtime like Vertex AI Agent Engine. The unglamorous last mile between a notebook that impresses your team and a system real users can hit.
# Who Should Take It, and How to Get the Most from It
If you can write Python and you’ve called an LLM API, you’re ready. The course assumes programming literacy but not agent expertise, which is exactly the band most practitioners sit in right now.
One piece of advice from watching how these intensives tend to go: don’t skim the whitepapers to rush to the code. The evaluation and context-engineering material on days three and four is where the durable understanding lives, and it’s the part you can’t pick up later by copying a working repo. The codelabs teach you to build an agent. The papers teach you why yours keeps breaking.
The whole thing is free, it’s available now on Kaggle, and the capstone gives you something real to point at when you’re done. A million and a half people found five days for it. Yours are probably worth the same.
Nahla Davies is a software developer and tech writer. Before devoting her work full time to technical writing, she managed—among other intriguing things—to serve as a lead programmer at an Inc. 5,000 experiential branding organization whose clients include Samsung, Time Warner, Netflix, and Sony.
