Google Colab is getting a lot smarter — and a lot more personal — thanks to new Gemini customization features that turn the built-in AI into something closer to “your” agent rather than a one-size-fits-all helper.

At the heart of this update are Custom Instructions, which live at the notebook level. That means you can effectively “program” how Gemini should behave for a specific project or course, and those preferences stay attached to that notebook. Prefer a particular coding style, want Gemini to always reach for a specific Python library, or need it to keep your class syllabus in mind when suggesting exercises or explanations? You can now set all of that once as Custom Instructions and reuse it every time you chat with Gemini in that notebook. You toggle these instructions right from the Gemini chat box and Colab saves them with the file, so you don’t have to repeat yourself in every new prompt.

The second big addition is Learn Mode, which basically flips Gemini into personal coding tutor mode. Instead of dumping a ready-made block of code you can just copy-paste, Learn Mode focuses on step-by-step guidance, explaining concepts, breaking problems into smaller pieces, and nudging you toward the solution. It’s clearly aimed at people who actually want to understand what’s going on under the hood — students learning Python, developers exploring a new framework, or even educators who want an AI co‑teacher inside their notebook. You can turn Learn Mode on directly from the Gemini chat window in Colab whenever you want that more hands-on, tutor-style experience.

Google is also leaning into the community angle. Because Custom Instructions are stored inside the notebook itself, they travel with it when you share a Colab link. So if you’re an instructor, you can ship a notebook where Gemini is already tuned to your teaching style and course context. If you’re a developer sharing an internal notebook at work, you can bake in your team’s preferred libraries, patterns, and documentation references so everyone who opens it gets the same tailored AI behavior by default.

To help people see Learn Mode in action, Google is surfacing example notebooks with Python exercises — like dedicated notebooks for lists and strings — where Gemini is preconfigured to guide you through practice problems instead of just handing you answers. It’s a simple way to showcase how this tutoring style can make Colab more than just a place to run code; it becomes a place to actually learn to code.