Google and Kaggle draw 353,000 developers to a no-cost ‘vibe coding’ course on AI agents

Google and Kaggle on Monday outlined the results of a large, no-cost five-day training event focused on AI agents, a fast-growing approach to software that uses models to take actions toward a goal rather than only answering questions. In a recap of the program, Google said more than 353,000 developers joined the latest AI Agents: Intensive Vibe Coding course, continuing a broader push to help builders move from early prototypes to working systems.

The course combined expert-led sessions with technical whitepapers and notebooks, and it culminated in capstone projects designed to apply what participants learned. Google described the week as a full lifecycle training path, spanning designing AI agents, securing them, and deploying production-grade agents in the cloud. Instead of requiring learners to start from code-heavy workflows, the program emphasized “vibe coding,” a style of programming where developers work through natural language prompts to shape software behavior.

Collaboration was built into the format. According to the recap, participants used Kaggle’s Discord for real-time debugging and group problem solving. Google said there were more than 392,000 active participants in the Discord during the event, underscoring how the teaching component extended beyond formal lessons into peer-to-peer support.

Google also provided outcome metrics from the capstone phase. The companies said they received over 6,000 project submissions from more than 12,000 active capstone participants. The submitted projects, Google said, ranged widely in application, including tools for historical transcription and systems aimed at space-weather research.

Among the example projects highlighted in the recap were Palimpsest, described as a historical manuscript transcription pipeline, and Project ARIES, described as a space-weather research system. Google said the volume and quality of entries made it difficult to narrow down top results, and it directed readers to a page where winners and capstone work can be explored.

The recap also positioned the event within a longer-running partnership. Google said its collaboration with Kaggle has been running since a first “5 Day Intensive” course in 2024, and that more than 2 million learners and developers have participated in the companies’ multimodal, no-cost courses over that time. The latest effort, Google wrote, continued that theme by exploring “the frontier of vibe coding,” with the goal of helping developers graduate their prototypes “from vibe to live.”

In market terms, the program illustrates how major AI-platform companies are increasingly using education as a distribution channel for new developer workflows. By packaging agent-building concepts, security, and cloud deployment into a time-bounded intensive, Google and Kaggle are effectively lowering the barrier to entry for teams that want to experiment with agentic systems but may lack internal training resources.

The event also points to a community-driven model of learning. Rather than treating training as a one-way lecture, the recap emphasized Discord-based collaboration and live troubleshooting, which can reduce time-to-learning by letting new builders compare approaches and resolve errors quickly. That matters as AI agent tooling changes rapidly, with Google saying AI is moving faster than traditional learning and teaching methods can keep up.

Still, some specifics remain outside what Google disclosed in the recap. The company did not provide details on the precise infrastructure used for training or evaluation, the selection criteria behind capstone “winning” projects, or any measured performance outcomes from the agents that participants deployed. It also did not quantify the share of participants who completed the full curriculum versus those who joined and participated sporadically.

Looking ahead, Google said all course materials remain accessible as a self-paced guide on Kaggle Learn, and it pointed learners to ongoing Kaggle competitions as a follow-on pathway. For developers and observers, the next question is whether these large-scale educational cohorts translate into sustained production adoption, and whether future intensives will focus on additional agent capabilities such as stronger safety guardrails, tool use in production environments, or tighter integration with cloud deployment workflows.