35万3000人が参加したKaggleの「AIエージェント集中講座」の舞台裏Inside our 353,000-person vibe coding course

Kaggle, the Google-owned platform best known for data science competitions, recently wrapped up its AI Agents Intensive, a free online course run in partnership with Google that drew more than 353,000 registered learners. The scale of participation is notable because it points to intense demand for structured, hands-on training in one of the fastest-moving corners of software development: autonomous and semi-autonomous AI agents that can plan, use tools, and complete multi-step tasks with limited human oversight.

The course was offered at no cost and centered on teaching participants how to build and deploy AI agents rather than simply understanding them in the abstract. That practical, deployment-focused framing distinguishes it from more theoretical machine learning curricula. Agents represent a shift from single-shot prompting toward systems that can reason across steps, call external functions and APIs, retrieve information, and act on the results. For many developers, this requires learning new patterns around orchestration, memory, tool integration, and evaluation, which a guided intensive is well positioned to introduce.

By making the program free and open to a global audience, Google and Kaggle appear to be pursuing a strategy of democratizing AI developer education at scale. Lowering the cost barrier to essentially zero widens the funnel of people who can experiment with agent frameworks, and it also helps seed familiarity with Google's own tooling and ecosystem. This kind of large-scale, no-cost course functions both as talent development and as a way to broaden the base of practitioners comfortable building on a given platform, a dynamic common across major AI providers.

The turnout also reflects broader momentum around the tag the organizers associate with the effort: vibe coding, a loosely defined term describing a more conversational, intent-driven style of programming in which developers describe what they want and rely on AI systems to generate and assemble much of the underlying code. Agents are a natural extension of that trend, since they push beyond code generation toward software that can carry out workflows. Teaching large cohorts how to construct such systems is likely to accelerate the adoption of these development patterns, though the long-term productivity and reliability implications remain an active area of discussion.

Context helps explain why an intensive like this resonated. Over the past couple of years, agent-building has moved from research demonstrations into mainstream developer tooling. Frameworks and libraries aimed at orchestrating language models, connecting them to external tools, and chaining reasoning steps have proliferated, and cloud providers have layered agent capabilities into their platforms. Google has been investing across this space through its model families and developer offerings, so a Kaggle course serves as an accessible on-ramp for developers who want to move from reading about agents to shipping working prototypes.

Kaggle itself is a logical host for this kind of program. The platform has long combined competitions, shared notebooks, public datasets, and community discussion, giving it an established audience of data scientists and machine learning engineers along with the infrastructure to distribute course materials and hands-on exercises. Running a large intensive through that channel lets Google reach a technically inclined community that is already predisposed to experiment, and it builds on Kaggle's earlier track record of similar short-format learning programs.

For learners, the appeal of such courses typically lies in structure and momentum. A time-boxed intensive with defined milestones, worked examples, and a shared cohort can help participants push past the initial friction of setting up tooling and understanding core concepts. Whether all 353,000 registrants completed the material is unclear, since large open online courses generally see registration numbers far exceed completion, but even partial participation across a cohort of that size represents a substantial injection of agent-building familiarity into the developer community.

The initiative fits a wider pattern in which major technology companies use free education as a means of expanding the pool of skilled practitioners while cultivating ecosystems around their products. As agents continue to attract attention from enterprises exploring automation and from individual developers experimenting with new workflows, offerings like the AI Agents Intensive are likely to become a recurring feature of the landscape. The reported figures suggest that appetite for this kind of accessible, deployment-oriented training is considerable, and comparable programs from Google, Kaggle, and their competitors will be worth watching as the tooling matures.