First Place at Google Gemini and SaaStr

Two early milestones helped shape Mosaic: first place in the Google Gemini Long Context competition on Kaggle, followed by first place in the SaaStr × Mayfield AI Pitch competition.

One was a chance to show what long-context, multimodal models could make possible. The other was a chance to put that work in front of people building and buying software. Both started with the same idea: video editing should become much easier to direct.

Google Gemini Long Context

In December 2024, our team won a $25,000 first-place Grand Prize in the Google Gemini Long Context competition. We entered under our original name, FrameCut, before becoming Mosaic.

The competition invited developers to demonstrate interesting applications of Gemini’s long context window. For us, video was a natural place to start: an edit depends on understanding what happens across a recording, then deciding which moments belong together.

In our original announcement, we shared the judges’ feedback on our “comprehensive approach and compelling vision for a complete product in just a month.”

Read the competition announcement on Kaggle →

SaaStr × Mayfield AI Pitch

In May 2025, Mosaic took first place in the SaaStr × Mayfield AI Pitch competition at SaaStr Annual.

We brought our approach to agentic video editing to the stage: helping people turn their footage and creative intent into a finished video. The pitch was an opportunity to show the product and explain the problem we were building it to solve.

The photo above captures the team on stage at SaaStr. The competition video is also linked from our press page.

Watch the SaaStr competition video →

What we’re building next

Those early demos became the foundation for the work we’re doing at Mosaic today: state-of-the-art video agents built around multimodal understanding and agentic orchestration.

Motion is where that work becomes a product. We’re continuing to build agents that can understand a creative brief, work with video, and help people shape the result.

Thanks to the Google, Kaggle, SaaStr, and Mayfield teams, and to everyone who tried those early versions and told us what needed to get better.

Try Motion →