Meta Global AI Hackathon 2026 — Unofficial Submission & Judging Readiness Checklist

Your Submission Readiness Saved automatically in local browser storage

0 / 20 checklist items verified 0% ready

Submission Portal ↗

1. How the 3-Stage Judging Workflow Operates

Submission Freeze: Wed, Nov 4 at 9:00 PM EST (Thu, Nov 5 at 02:00 UTC)

Stage 1 · Triage & Flagging Nov 5–6

Pass/Fail Screening & Bonus Flagging

Every submission is screened pass/fail before entering rubric scoring:

Addresses Build for Your World or Push the Frontier
Meaningfully uses Meta Model API and/or Meta open-weight models at runtime
Public GitHub repository showing work built Oct 26 – Nov 4
Working, publicly accessible 3-minute demo video
Complies with Anti-Project rules (not a generic 1-prompt wrapper)

Critical Nuance: There are no opt-in checkboxes on the form for technical bonus prizes ($10k Open Weights, Best Agentic, Best Creative). Stage 1 screeners manually flag projects based on your video, top of README, and the “Use of Meta products” form field.

Stage 2 · 10-Point Rubric Nov 6–8

Weighted Rubric & Partner Verification

Projects passing Stage 1 are scored on the Official Rules weights (overriding the “equally weighted” typo in the guide):

Impact (30% · 3.0 pts): Real-world utility, problem severity & post-hackathon viability
Use of Meta AI Tooling (25% · 2.5 pts): Creative, load-bearing multi-model or open-weight usage
Technical Depth (20% · 2.0 pts): Architecture, engineering quality & quantitative benchmarks
Demo Quality (25% · 2.5 pts): Clear 3-minute video + reproducible repo setup instructions

Partner Check: Infrastructure partners (Baseten, Modal, Fireworks) verify projects flagged for the $10k Open Weights bonus to confirm Muse Glimmer or Muse Spark Open is a load-bearing component.

Stage 3 · Live Finals Mon, Nov 9 · 12:00 PM EST

Top 5 Indie & Top 5 Startup Broadcast

The Top 5 projects in the Indie track and Top 5 in the Startup track (10 finalists total) advance to the Closing Ceremony stream:

No live pitch or live Q&A required from finalists
Final Round Judges watch the submitted 3-minute demo videos live on stream
Ranked using the same 4-pillar rubric to crown track winners
Prizes stack: a project can win Main Track + multiple Bonus Prizes

Key Takeaway: Your submitted 3-minute video is the exact artifact played live for the Final Judges on November 9 — it must stand completely on its own.

2. Interactive 20-Point Pre-Submission Checklist

Click any item to mark it complete (5 items per pillar)

3. Recommended 3:00 Video Pacing Blueprint

Proportional 180-second allocation for Stage 1 screening, Stage 2 rubric & Stage 3 live finals

0:00 Start 0:15 Live Demo (45% of runtime) 1:35 Meta AI Stack (22%) 2:15 Benchmarks (17%) 2:55 Stop (3:00 Cutoff)

0:00 – 0:1515s Hook & Provenance

1-sentence problem statement + mandatory 1-line provenance callout stating what was built from scratch during Oct 26–Nov 4.

0:15 – 1:3580s Live Product Walkthrough

Jump straight into the running product (no slide deck). Show a real end-to-end user workflow with live Meta model calls visible on screen.

1:35 – 2:1540s Meta AI Architecture

Show how Meta models work together under the hood (e.g. Muse Spark + Muse Spark Open / Muse Glimmer / SAM 3.1) to trigger Stage 1 bonus prize flagging.

2:15 – 2:4530s Benchmarks & Proof

Show quantitative evaluation metrics (accuracy, latency, A/B model comparison, test suite, 1-command reproducibility).

2:45 – 2:5510s Summary & Clean Stop

Final summary card with repo URL, Meta models used, and headline metrics — finishing cleanly at 2:55 before the 3:00 cutoff.

4. Copy-Paste README Provenance Header & Bonus Prize Guide

Help Stage 1 screeners & Stage 2 judges verify your project in 10 seconds

## Meta Global AI Hackathon 2026 — Submission Overview

- **Track & Problem Statement**: Indie Developer (or Startup) · Push the Frontier (or Build for Your World)
- **Hackathon Provenance (Oct 26 – Nov 4, 2026)**: Built 100% from scratch during the hackathon window.
- **Meta AI Runtime Models**:
  - `Muse Spark 1.3` (`api.meta.ai`) — Primary reasoning, tool-use & orchestration
  - `Muse Spark Open` / `Muse Glimmer` — Load-bearing open-weight inference stage
  - `SAM 3.1` / `Muse Wonder` / `Muse Voice` — Multimodal perception / synthesis
- **Bonus Prize Categories Targeted**:
  - Best Build on Meta Open Weight Models ($10k) · Best Agentic Use-Case · Best Creative Build
- **3-Minute Demo Video**: https://youtu.be/YOUR_VIDEO_ID
- **Quickstart & Keyless Replay**: Run `cp .env.example .env` with `MODEL_API_KEY`, or `npm run demo:replay`.

Runtime Meta Models Used:
1. Muse Spark 1.3 (Meta Model API): Orchestrates multi-step agent planning and structured tool execution.
2. Muse Spark Open (meta-models/muse-spark-open-1026 via Baseten/Modal/Local): Load-bearing open-weight specialist stage.

Bonus Prize Categories Targeted:
- Best Build on Meta Open Weight Models ($10k)
- Best Agentic Use-Case

Provenance (Oct 26 – Nov 4): Built 100% from scratch during the hackathon window.

How Bonus Prizes Are Flagged & Verified Prizes Stack!

Best Build on Meta Open Weight Models$10,000

How to qualify: Use Muse Glimmer or Muse Spark Open (meta-models/muse-spark-open-1026) as a load-bearing runtime stage (partner inference or local GPU). Manually flagged in Stage 1 & verified by Baseten / Modal / Fireworks.

Best Agentic Use-Case & Best Creative BuildBonus

How to qualify: No form checkbox — Stage 1 screeners flag contenders based on multi-step tool/agent autonomy or creative multimodal synthesis shown in your 3-minute video and README header.

Most Viral Project on X$20,000

How to qualify: Post your demo video on X before Nov 4 and paste the X URL in the submission form. Ranked by Likes + Reposts + Comments (views excluded) through Sun, Nov 8 at 11:59 PM EST.

Community Vote Winner$10,000

How to qualify: Opens on the Cerebral Valley project gallery after the Nov 4 deadline (1 vote per verified account; winner announced Mon, Nov 16).

5. Common Questions & Official Clarifications (12 Key Rulings)

Topic / Question	Official Clarification from Discord Q&A
Scoring & Finals Are the 4 rubric pillars equally weighted (25% each) or 30/25/20/25?	30% Impact / 25% Meta AI Tooling / 20% Technical Depth / 25% Demo Quality. Organizers confirmed on Oct 26 that the Official Rules weights (`30/25/20/25`) govern and override the “equally weighted” typo in the Participant Guide.
Scoring & Finals Do finalists have to pitch live or answer Q&A on November 9?	No. There is no live pitch or live Q&A for finalists. During the Closing Ceremony broadcast (`Mon Nov 9, 12:00 PM EST`), Final Round Judges watch the Top 5 submitted 3-minute demo videos per track live on stream and select the winners.
Models & Prizes How do we apply for the $10k Open Weights, Best Agentic, or Best Creative bonus prizes?	There are no separate checkboxes on the submission form. Stage 1 screeners manually flag projects for bonus categories based on what you write in the “Use of Meta products” form field, the top of your `README.md`, and your 3-minute video.
Models & Prizes Does combining hosted `Muse Spark 1.3` with `Muse Spark Open` or `Muse Glimmer` qualify for the $10k Open Weights prize?	Yes! Combining multiple Meta models is explicitly encouraged. As long as `Muse Glimmer` or `Muse Spark Open` (`meta-models/muse-spark-open-1026`) is a load-bearing stage of your pipeline (via Baseten, Modal, Fireworks, or local hardware), hybrid architectures qualify. Using hosted `SAM 3.1` alone does not count as an open-weight build.
Models & Prizes Can we use Claude Code, Cursor, OpenCode, or Copilot to write code?	Yes, unrestricted. Coding assistants used during development do not need to be disclosed and are not restricted. The Meta model requirement applies to the runtime models powering your application.
Models & Prizes Can we use deterministic algorithms or small custom classifiers alongside Meta models?	Yes. Deterministic engines (parsers, solvers, compilers, rule engines) are not competing AI models. Small self-trained classifiers/forecasters alongside Meta models are also permitted provided the training/eval code is open in the repository.
Video & Repo What happens if our demo video is 3:05 or 3:15 long?	Organizers stated they will not disqualify a video for being a few seconds over, but judges stop watching at the 3:00 mark. Keep your video at or under `2:55–3:00` so your benchmarks and closing summary are never cut off.
Video & Repo Is a hosted live web URL mandatory?	No. A hosted web deployment is optional. Judges require a working 3-minute demo video and a public GitHub repository with clear setup instructions (plus a keyless/replay mode if your app needs special hardware or external setup).
Video & Repo What if we started from an existing codebase or startup product?	Only work built during the hackathon window (`Oct 26 – Nov 4`) is evaluated. Organizers recommend submitting a clean standalone repository created during the hackathon containing the new module/application, and clearly stating in the first line of your video and README what existed prior vs. what was built this week.
Tracks & Teams Can I switch between Indie and Startup track, or switch between Build for Your World and Push the Frontier?	Yes. You make your final track and problem-statement selection directly on the Cerebral Valley submission form (regardless of what you selected when registering).
Tracks & Teams Who qualifies for the Startup Track vs. Indie Developer Track?	Startup Track requires a legally incorporated company in your country of origin with the ability to employ people (e.g. US EIN / Tax ID; solo incorporated founders may enter Startup). Unincorporated individuals and teams must enter the Indie Developer Track.
Tracks & Teams Can a teammate who wasn’t accepted to the hackathon contribute code?	No. Every contributor to the repository and every listed team member (maximum 4 people per team) must be an individually approved hackathon participant, and each participant can only be on one team / one submission.