On April 9, we ran a Build with AI Code Lab at UTBM's Crunch Lab in Belfort-Montbéliard, France. I co-organized it with GDG on Campus UTBM. Jean-Philippe Baconnais, a Google Developer Expert in Cloud from [Zenika](https://zenika.com Nantes, came in to lead the live coding. About 25 students and developers showed up with laptops. The deal: start with an app idea, walk out with something deployed on Google Cloud Platform. Two hours. No manual coding (or close to it).
Most of them had a working app live before the networking break.
The three tools
The session covered three Google tools in sequence.
Google AI Studio came first. Participants tested prompts and poked at Gemini models to see how they respond. No setup required. You type, you get a response. It is the fastest way to get your hands on a Gemini model if you have never used one.
Firebase Studio was next. Database, hosting, auth. Jean-Philippe showed how to wire Firebase services into an app without spending an hour on boilerplate config. This part went faster than I expected.
Google Antigravity was the core of the session. Antigravity is an agentic development platform. You describe what you want, and autonomous agents on Gemini models generate code, connect components, and wire things together. You review the output through Artifacts. You keep what works, fix what does not. You can plug in custom Tools like skills and MCP servers to steer the agents.
The pipeline: brainstorm in AI Studio, backend on Firebase, app assembly and deployment through Antigravity on GCP. Google provided credits, so nobody paid a cent.
What it looked like in practice
Nobody was typing code line by line. People described features, reviewed agent output, accepted or rejected, and moved to the next piece. The speed caught me off guard. Students who had never deployed anything before had live URLs within the session.
Jean-Philippe did not lecture.
He built in front of everyone, hit errors on screen, debugged them live, and talked through his reasoning. That is hard to pull off well, and he made it look easy.
Zo R. from our sponsor Aquantic (Zonova Sarl) posted about it and mentioned the concentration in the room. That is accurate. People were focused.
What this changed for me
I write C++ and Python every day for my PhD in computer vision and robotics. I control every line. Every function call. Every parameter.
During this Code Lab, the students were not doing that. They were describing what they wanted. Reviewing what came back. Catching mistakes. Deploying. The role was closer to supervision than implementation.
I am still processing what that means for how I work. I do not think it replaces writing code for research (my problems are too specific). But for building apps, prototyping, shipping things fast? The workflow is different. And I watched 25 people learn it in 2 hours.
Two things I did not expect
The deployment step. I assumed it would be the bottleneck. It was not. Antigravity handled GCP deployment cleanly, and most apps were live well before the end.
Non-CS attendees shipping apps. Some people in the room were not computer science students. They still produced working prototypes because the interface asks for intent, not syntax. If you know what you want to build, you can get surprisingly far.
If you want to try this
Pick a small project. Something you would normally spend a weekend on. Start in AI Studio to test your prompts. Use Firebase for the backend. Let Antigravity assemble the rest. See what happens.
Fair warning: reviewing agent output instead of writing code yourself feels strange at first. You keep wanting to take over. Resist that for the first hour and see how far the agents get you.
And if you run a developer community, this Code Lab format works. Hands-on, 2 hours, clear goal, everyone walks out with something deployed. Way more effective than a talk.
Thanks
To Jean-Philippe Baconnais for the live coding. You made Antigravity make sense for a room full of people who had never seen it.
To Zenika for backing community events.
To Aquantic (Zonova Sarl) and Zo R. for sponsoring and being there.
To UTBM Innovation Crunch Lab, Olivier Lamotte, and Eva Hazebroucq for the space.
To Pauline Bonnin Parsons and Angelika Zaucha for connecting us with Jean-Philippe.
To Fayçal Bijji and the GDG UTBM team for co-organizing.
To every student and dev who stayed, built, and deployed something. That is what made it worth running.
