What if you could build an artificial intelligence application by describing what you want in natural language? That is the idea behind the new AI agents course that Google and Kaggle will offer online, at no cost to those who register.
The initiative builds on the five-day intensive course both companies organized last November, which reached more than 1.5 million students. This new edition will feature updated content, new speakers, and a practical final project.
A course for building AI agents
The program will take place from June 15 to 19, 2026, covering the development of artificial intelligence agents from basic concepts to the creation of systems ready for real-world scenarios.
The course is aimed at people interested in learning how to design, build, and deploy agents capable of using tools, connecting to APIs, and completing tasks more autonomously. It is not simply about chatting with a chatbot, but about creating systems that can act according to defined goals.
Each day will combine conceptual explanations with practical examples. This way, participants can move step by step from the fundamentals to a functional implementation.
What it means to program with natural language
One of the course's main topics will be what is known as vibe coding. In simple terms, it means using instructions written in natural language as the main interface for creating software.
Instead of starting with lines of code from scratch, you can describe what an application should do, request changes, and refine the result alongside an AI tool. This does not eliminate the need to understand the program's logic, but it can lower the initial barrier to experimenting and building prototypes.
The idea is not for AI to program without supervision, but for it to turn a description of an idea into a starting point that you can review, test, and improve.
From concepts to so-called “10x agents”
The course will also teach participants how to integrate tools and APIs to create what Google describes as “10x agents.” The term refers to systems capable of expanding what a person can do by automating parts of a workflow.
For example, an agent could receive a request, retrieve information through an API, organize the results, and prepare a response. To be useful in practice, it must also handle errors, respect limits, and respond coherently to unexpected situations.
The content will aim to connect experimentation with the construction of more robust systems. That difference matters: a demonstration may work once, but a tool intended for users needs testing, safeguards, and a reliable architecture.
A final project to put the idea into practice
The June edition will conclude with an integrative project, or capstone. Participants will be able to apply what they have learned to turn an idea into a functional agent system.
This type of activity helps you understand where real-world challenges arise. What tools does the agent need? How should it react if an API fails? What information can it use? How do you verify that its response is correct?
Answering these questions is an essential part of moving from curiosity about AI to creating solutions that can be used in a business, a personal project, or a work team.
How to participate
The course will be completely online and free for registered participants. Google and Kaggle say that registration is already available on their official website.
The program may appeal both to people who are just beginning to explore artificial intelligence and to developers and entrepreneurs who want to update their skills in agents, APIs, and automation.
AI is not waiting in some distant future. It is already changing how software is built, and courses like this aim to help more people understand and use it in practical ways, without being held back by their initial technical level.