Ralphthon

Busan · Codex Goal Participant Guide

Busan
Problem Statement

The goal is to show a working result built with Codex Goal in a half-day sprint.

Track 1 — Local Impact

Prove a Busan/Gyeongnam problem with a working product.

Build an AI product that local startups, communities, or daily operators could try right away.

  • The problem is real, and someone is willing to pay (or use it daily).
  • The product is polished enough that the demo itself sells it.
  • The craft, UX, and shipping-ness are visible from minute one.
  • AI-native, but the AI is in service of a human user — not the other way around.

Stack is free, but use Codex Goal to move fast and show a real working demo.

Think: local commerce, manufacturing, logistics, tourism, community, student, or founder workflows.

Track 2 — Goal Harness

Prove how well you operate Codex Goal.

Less about the product surface, more about making Codex Goal run with clear goals, checks, recovery, and repeatability.

  • The agent runs long, runs autonomously, and finishes its job without babysitting.
  • Goal, termination, verification, and recovery are designed — not improvised.
  • Multi-agent orchestration, tool use, memory, and loop structure are deliberate.
  • A judge can read the harness and understand why it works, not just that it works.

Goal Harness requires Codex Goal. Show goal setup, verification, termination, and recovery in the demo.

Think: Codex Goal playbooks, verification loops, automatic task decomposition, and reusable development workflows.

Track comparison

Impact
Harness
Scored on
Problem fit · UX · completeness · local impact
Goal design · autonomy · verification · reuse
Codex Goal usage
Recommended
Required
Why it's good
Local builders can try it immediately
The team's development workflow improves
Keywords
local, useful, demo-ready, AI-native
goal, verification, recovery, harness
Reference
Local AI products and workflow automation
Codex Goal playbooks and agent workflow harnesses

Declare your track at submission. Prizes are awarded per track.

NOT TO DO List

Projects will be immediately disqualified if they fall into any of these categories:

  • Basic RAG Applications
  • Streamlit Applications
  • Image Analyzers
  • "AI for Education" Chatbot
  • AI Job Application Screener
  • AI Nutrition Coach
  • Personality Analyzers
Connect with the Community

Join the Ralphthon Discord for team formation, announcements, and Q&A:

  • #announcements — Official updates from organizers
  • #team-building — Find teammates (max team size: 4)
  • #live-and-updates — Live updates during the event
  • #questions — Ask organizers anything
Rules
  • Team Size: Maximum 4 members per team. Solo participants are welcome.
  • Track Lock-in: One team, one track. Declare at submission.
  • Open Source: Submitted repositories must be public.
  • New Work Only: You may not present existing projects as your own work. Judges will check.
  • Demo Requirements: Your demo must highlight only features built during the hackathon. Failure to clearly identify original contributions = disqualification.
  • Harness Track = Codex integration required: Harness submissions that don't integrate with Codex are auto-disqualified.
  • Banned Projects: Projects will be disqualified if they violate legal, ethical, or platform policies, or use code/data/assets you don't have rights to.

📹 Media Notice

Photos or video may be taken for event documentation. If you prefer not to appear, tell the organizers at check-in.

Judging

Casual demos: 3 minutes per team, focused on a real working screen rather than slides.

Clearly show what you built during the event. Working demos beat polished decks.

Scorecard

One card per judge per team.

Columns: Team name · Judge name · Live demo (0-4) · Creativity / Originality (0-3) · Impact potential (0-3)

Total: 10 points (Live demo 4 + Creativity 3 + Impact 3)

  • Live Demo (0-4)How well does the team implement the core idea? Does it actually work live?
  • Creativity / Originality (0-3)Has this been seen before? Does it tackle the track's problem in an original way?
  • Impact Potential (0-3)Will this survive past the hackathon? How useful is it?

Tracks

Local Impact and Goal Harness are reviewed with different emphasis. Goal Harness includes Codex Goal operation quality.

Questions? Reach out to GB Jeong — bong@team-attention.com