You must be signed in to submit a project.
Please see the Schedule and Judging Criteria for details.
$ cat /etc/cookies.conf
We use cookies to understand how people use this site.
Analytics cookies help us improve your experience.
They are off by default. Nothing tracks you until you say so.
$ select cookie_preferences
Entries for the Google DeepMind x Cactus Compute Global Hackathon hackathon.
Please see the Schedule and Judging Criteria for details.
We will evaluate all projects based on the following criteria, focusing on the successful implementation of local-first, agentic, hybrid AI systems:
| Criterion | Description | Scoring (1 to 5) |
|---|---|---|
| Functionality & Execution | Did the team ship a working demo that successfully integrates the required components? A score of 5 means the application runs smoothly on-device (mobile/desktop) with zero critical bugs. | 1: Project does not run or is entirely conceptual. 2: Major bugs prevent core functionality from being demonstrated. 3: Runs, but has significant stability or integration issues. 4: Runs reliably, demonstrates core features. 5: Flawless execution, robust on-device performance. |
| Hybrid Architecture & Routing | How effectively was the core challenge—deciding where computation happens (local vs. cloud)—solved? This judges the intelligence of the edge/cloud split. | 1: No clear local/cloud split; all logic is in the cloud. 2: Local execution is present but serves no meaningful purpose. 3: Simple, static routing between local (FunctionGemma) and cloud (Gemini). 4: Intelligent routing based on context (e.g., latency, complexity, connectivity). 5: Sophisticated, dynamic, and optimized routing that showcases seamless fallback/escalation. |
| Agentic Capability & Utility | Does the application demonstrate meaningful agentic behavior (reasoning, tool use, workflow coordination) that benefits from being local-first? | 1: Simple, non-agentic task execution. 2: Agent logic is present but trivial or easily replicated without agents. 3: Demonstrates basic agentic workflow using FunctionGemma. 4: Complex, useful agent workflow that leverages local speed for responsiveness. 5: Groundbreaking agent system that unlocks a new local-first UX pattern (e.g., real-time offline coordination). |
| Theme Alignment (Local-First & Tech Stack) | How deeply and creatively were FunctionGemma and Cactus Compute utilized to achieve local-first goals (latency, privacy, offline capability)? | 1: Does not use FunctionGemma or Cactus Compute, or uses them superficially. 2: Uses the required tech, but the solution would work identically in the cloud. 3: Clearly uses FunctionGemma/Cactus, but the privacy/latency benefits are not central to the demo. 4: Excellent use of FunctionGemma for speed/privacy, with Gemini used effectively for fallback. 5: The entire architecture is fundamentally dependent on the local-first capabilities unlocked by FunctionGemma/Cactus Compute. |
This milestone has passed
The team formation deadline has passed, but you may still create or join a team now if you need to.
This milestone has passed
hint: You may still make edits to your submission up until the time you are actually judged.
This milestone has passed
This milestone has passed
Times shown are in America/Los_Angeles time.