Brainstorm word cloud
Terms from this hackathon's own brief, judging criteria, discussion board and submissions. Bigger = mentioned more often; wording that most Devpost hackathons share is left out. Click a word to see past winning projects that mention it.
Past winners in related hackathons
Winner-badged projects from ended Devpost hackathons with the same theme that mention terms from this brief, closest first. Read a few before you pick an idea.
- BLUEPRINT: AI Property Due Diligence — Seven AI agents read public records on any US address, permits, flood zones, EPA data, then run an adversarial AI debate before delivering a buyer risk score. Powered by Elastic Agent Builder MCP. (Google Cloud Rapid Agent Hackathon)
- ComplianceOS — A compliance agent that audits your code against real law, writes the fixes, and reads its own Phoenix traces through MCP to get better after every scan. (Google Cloud Rapid Agent Hackathon)
- Cassandra — AI agents fail silently, confidently wrong, and nobody notices. Cassandra is an AI that watches your AIs. It catches every lie, diagnoses why, ships a proven fix, and even grades its own accuracy. (Google Cloud Rapid Agent Hackathon)
- Carryover — Visa intelligence that protects migrants from agency fraud. Three AI agents on Elasticsearch + Gemini that don't just inform but they detect scams in real time and file real complaints. (Google Cloud Rapid Agent Hackathon)
- Sentinel — Catch security holes in your MCP server before you ship it. Static rules, GPT-5.6 review, and Docker-sandboxed probes-every finding mapped to the OWASP Agentic Top 10. (OpenAI Build Week)
- Sammy - a privacy protected federated readmission predictor — HIPAA blocks pooling patient data. Sammy predicts readmission risk with federated learning, running an XGBoost model inside the database—so data never leaves the hospital. Privacy-first, live now. (H0: Hack the Zero Stack with Vercel v0 and AWS Databases)
- Pulse — An AI teammate that listens during cardiac arrest and holds the state the team leader is juggling: rhythm, shocks, drug timings, hands-off time. Deterministic engine, no LLM in the clinical loop. (OpenAI Build Week)
- Shot Clock — A crew of Gemini agents watches a VFX render farm in Grafana, works out which shots miss the delivery date, proves why, looks at the frame, and prices it. (Agentic Cinema: The Blockbuster Hackathon)
- The Media Customs — The Media Customs clears your commercial for markets before it airs. A Google Agent watches it with you, judges it against each country's law, cites the statute, and re-renders the shots that fail. (Agentic Cinema: The Blockbuster Hackathon)
Judging criteria
Criteria as listed on the official page; weights are not stated there, so they are drawn equal. source ↗
How fresh is the project idea or its use of Relate? Higher scores go to projects that present an original idea or an inventive way to use Relate.
How effectively does the project use Relate’s graph relationships to support its main use case? This includes how easily judges can use the submission of instructions, running, and exploring with Relate working as described.
How much of the project’s core use case works?
How clearly does the project address a real problem for an intended user? Who benefits and the does the project clearly show this.
How clearly does the README explain the project and how to try it?
Timeline
Submission period (from Devpost): Nov 11 - Dec 11, 2026. Deadline: Deadline: Dec 11, 2026 @ 6:00pm PST.
How a beginner can ship
- Read the rules page
Check eligibility, team rules and what must be submitted.
- Pick one idea from the word cloud
Choose one problem and one user; cut everything else.
- Build the smallest working version
One flow that works end to end beats five half features.
- Record the demo early
Most Devpost judging leans on the video and description.