Google’s official Gemma account has spotlighted a standout project from the latest Gemma 4 hackathon called Sidebar, a live meeting copilot that turns spoken conversation into a working prototype before the meeting is even over.
What Sidebar Does
According to the demo shared by Google, Sidebar runs as an AI agent that listens to a live meeting and converts what’s being discussed into a functional product prototype in real time. Google says the tool operates at over 1,000 tokens per second, fast enough that a growth dashboard idea discussed out loud, complete with details like MRR, churn, and weekly active users, can turn into a working chart-based prototype before the meeting wraps up.
The demo shown by Google features a split-screen view: a live meeting transcript on one side, and the resulting prototype, including UI components and a working bar chart, generating on the other, with response times of roughly 1 to 5 seconds shown on screen.
Part of the Gemma 4 Good Hackathon
The project comes out of the Gemma 4 Good Hackathon, a global competition hosted by Google DeepMind and Kaggle that challenges developers to build real-world applications using the Gemma 4 open-model family. The hackathon emphasizes building for the edge, targeting use cases with low connectivity, strict privacy needs, or limited infrastructure, and pushes teams to go beyond simple chatbots toward grounded, agentic systems.
The competition runs across five impact tracks: Health & Sciences, Global Resilience, Future of Education, Digital Equity & Inclusivity, and Safety & Trust, with a total prize pool of $200,000 across its Main, Impact, and Special Technology tracks.
Why It Stands Out
Sidebar reflects a broader trend among Gemma 4 hackathon entries: using the model’s speed and multimodal capabilities to collapse the gap between an idea being discussed and a working demo being built, without needing a developer to translate the conversation into code after the fact.
