Google announces in Colab two new features designed so that artificial intelligence doesn’t do the work for you, but teaches you how to do it: Custom Instructions and Learn Mode. Want an assistant that understands your style and guides you step by step instead of handing over ready-to-paste code? This is for you.
What Colab brings and why it matters
Google integrates Gemini more deeply into Colab with two features aimed at personalizing the learning and collaboration experience. It’s not just a cosmetic update: it changes how you interact with AI when you program, study, or prepare teaching materials.
Do you prefer understanding the solution instead of copying it? Learn Mode turns Gemini into a personal tutor that explains each step, breaks down complex concepts, and helps you actually learn. It won’t dump the answer in a block; it shows you the path.
Custom Instructions: customize the assistant per notebook
With Custom Instructions you can save preferences at the notebook level. Want it to always use a particular library, follow a coding style, or keep your course syllabus in mind? Those settings travel with the notebook.
- They are stored inside the notebook.
- You can turn them on or off from Gemini’s chat window.
- When you share your notebook, whoever opens it will have the same personalized assistant.
This makes it easier for the content you share to deliver a consistent experience: students, colleagues, or the community will see the agent exactly as you designed it.
Learn Mode: learn step by step, don’t copy and paste
Learn Mode uses Custom Instructions to steer Gemini toward teaching instead of automatically solving problems. In practice this means:
- Answers in clear, sequential steps.
- Explanations of the underlying concepts.
- Suggestions to practice and reinforce what you learned.
You can activate Learn Mode from Gemini’s chat window in Colab. It’s ideal if you’re exploring a new framework, reviewing a language, or designing educational exercises.
Practical usage examples
- If you teach, you create a notebook with exercises and leave instructions so Gemini teaches using your course’s methodology.
- If you’re a student, ask for an explanation of why a function works and request small challenges to practice.
- If you’re a developer, indicate a preference for
pandasor a style convention and receive guidance that respects that.
Colab even shows example notebooks where the agent already comes preconfigured in Learn Mode to accompany you through Python exercises.
Sharing knowledge and collaboration
The big advantage is that these settings encourage collaboration with context. When you share a notebook, you don’t just share code cells: you share the assistant that guides the learning experience.
This makes it easier to keep consistency in courses, workshops, and collaborative projects without each person having to reconfigure the environment.
Think of AI as a teammate that adapts its explanations to your rules and pushes you to understand, not to rely on copied code.
Final reflection
These updates show a subtle but important shift: AI in code becomes more pedagogical and customizable. If you’re interested in truly learning or improving how you teach programming, Custom Instructions and Learn Mode can become practical tools in your workflow.
Original source
https://blog.google/innovation-and-ai/technology/developers-tools/colab-updates