Hack02: LFMs with Eyes — What Happened, Who Won, What’s Next
Liquid AI just closed Hack02, themed LFMs with Eyes. This round focused on vision-language capabilities with our new LFM2-VL model — and the results surpassed expectations. Participation doubled. Our Discord community passed 1,000 members and is now a Discoverable Community, opening the door to even more builders.
The prize structure also evolved: more tiers, more targeted categories. You can see details and learn how to get involved at hackathons.liquid.ai.
Here’s who took home the prizes.
Taking Home the Prize
1st Place — Joseph Pollack: “Tonic AI: 100% local Android UI Control Dataset”
Joseph built a dataset and framework for 100% local Android UI control, enabling applications to interpret, manipulate, and respond to whatever’s visible on screen — no internet required. A big step toward fully local UI agents.
2nd Place — Angelo Cortez: “Home Listing Agent: Multimodal Similarity Search”
Angelo’s project applied LFM2-VL to match images and text queries for real estate listings. By combining visuals with descriptions, his agent can find similar homes 100% locally in resource-limited environments.
3rd Place — Peter Sholz: “Video Change Detector (Video-Diff)”
Peter built a video differencing tool for camera and security feeds. It detects changes between frames using on-device vision features — a privacy-first approach to monitoring without sending footage to the cloud.
What Hack02 Proved
Vision + language, local
LFM2-VL let participants build systems that combine text and visual inputs directly on device. That means new classes of applications that cloud-first models can’t deliver: real-time UI control, private CV for home environments, and secure video analysis.
Bigger stage, bigger community
Hack02 drew twice the builders of Hack01. With Discord crossing 1,000 members and opening to the public, the developer community is scaling fast.
More complex builds
Compared to Hack01, Hack02 projects showed deeper technical integration — from dataset design and visual preprocessing to latency optimization and local inference pipelines. This is the level of sophistication required to make edge AI real.
Keep Building at the Edge
If you missed Hack02, check out the winning projects and details of upcoming events at hackathons.liquid.ai. For model bundles, benchmarks, and deployment tools, head to leap.liquid.ai. And if you want to build alongside the fastest-growing edge AI community, join our Discord.
Hack02 proved it: vision and language on device isn’t a concept. It’s here, it’s working, and it’s being built by the Liquid developer community.