Neo and SpoonOS have announced the winners of the Scoop AI Hackathon’s Seoul Bowl, a two‑day event held Dec. 20-21 in Seocho‑gu, Seoul. The Seoul Bowl brought together developers, researchers, and creators to build AI applications that utilize the SpoonOS framework. More than US $8,000 in prizes were awarded across grand, track, and special award categories.

Participants competed across three technical tracks: Agentic Infrastructure & Productivity AI; AI4Science & Engineering; and Autonomous Finance, FinTech & Quant AI. Projects ranged from visualization platforms to developer tooling, medical education, behavioral analytics, and autonomous agent systems.

First Place – PRISM

The grand prize of US $2,000 was awarded to PRISM. PRISM is an educational tool that turns user questions into interactive, animated videos designed to accelerate understanding. Built on SpoonOS infrastructure and integrated with tools like Gemini and Manim, the platform is aimed at replacing static, text-heavy answers with dynamic, personalized visual explanations.

The system interprets a user’s query and generates a tailored animation that walks through the concept step-by-step – whether it’s solving a calculus problem or explaining supply chain dynamics. The interface supports interaction, allowing learners to modify the animation or explore different explanations.

Designed for students, professionals, and developers, PRISM applies brain-science-backed approaches such as dual coding and active learning to increase information retention and comprehension speed. The project’s technical stack includes FastAPI for backend, React for frontend, and Manim for animation rendering, all orchestrated through SpoonOS’s multi-agent control plane.

A demo video can be found below:

Track Winners (US $1,000 Each)

GitGuardian – GitGuard Agent is an AI‑powered recovery tool for developers, designed to resolve common source control issues such as merge conflicts, detached HEAD states, and stuck rebases. Built on SpoonOS StateGraph, the agent captures the repository state and analyzes issues through a multi‑stage pipeline, generating step‑by‑step recovery plans with reversible commands so developers can recover safely without data loss.

A live demo can be found below:
https://gitguardian.online/

Med DaaS (Medical Data as a Service) – Med DaaS aims to address the scalability gap in medical education and AI training caused by limited access to high‑quality comparative cases. Using semantic similarity search, the platform automatically surfaces reliable reference reports, enabling comparison‑driven learning with performance tracking and personalized feedback. By reducing dependence on continuous expert supervision, Med DaaS hopes to facilitate scalable radiology education and accelerate AI development in healthcare.

A demo video can be found below:

스토리 월드 (Story World) – Story World is an AI chatbot platform based on popular Korean web novels, designed to help hikikomori (a term for individuals who withdraw from society and remain isolated at home, often for months or years at a time) reconnect through AI-driven engagement. The project leverages familiar fictional content to create interactive conversations that reduce psychological barriers to social interaction.

A demo video can be found at this link.

Track Special Awards (US $500 Each)

HANDOFF – HANDOFF is a Chrome extension that visualizes how people interact with AI during decision-making. Instead of evaluating AI output or collecting prompt content, the tool passively tracks user behavior, like timing between prompts or how often a user re-asks a question, to pinpoint the moment when users stop thinking and start accepting AI suggestions. This moment is referred to as the “decision boundary.” By analyzing these invisible handoffs, HANDOFF aims to help users understand their AI collaboration style without compromising privacy

NandaGore (난다고래) – This project focuses on risk‑adjusted performance analysis for trading platforms. Instead of ranking traders by raw profit, NandaGore calculates multiple risk‑adjusted metrics, such as Max Drawdown, Calmar Ratio, and Recovery Factor, and uses AI to classify trading styles. The system is designed to highlight sustainable trading strategies over purely high-profit results.

FLOWLY – FLOWLY is an autonomous AI operator for chat-based commerce, designed to automate the end-to-end sales process inside messaging platforms like KakaoTalk, Instagram, and Telegram. Targeting small business owners who conduct all their transactions via chat, FLOWLY handles order intake, payment, and delivery coordination entirely within the thread, without requiring users or customers to leave the app. The system is aimed at reducing inventory mistakes and operational overhead while giving sellers clear visibility into their sales channels.

CoWeave – CoWeave is a platform for human-AI co-creation, designed to bring transparency and fairness to collaborative content generation. It allows users to work alongside multiple AI models (like ChatGPT or Gemini) to create stories, documents, research, and images, with every interaction permanently logged on-chain. These logs attribute specific contributions to either the human or AI, helping resolve authorship disputes, prevent plagiarism, and support ethical AI usage. The final outputs – whether text or image – are published with a breakdown of contribution percentages, and monetization is shared fairly through ad revenue or direct sales.

Safety Lens – Safety Lens is an industrial safety system that combines real-time vision analysis with agentic compliance monitoring. The platform is intended to detect dangerous hand movements with 0.03-second latency using MediaPipe and Optical Flow, reducing false stops while enabling fast emergency shutdowns. Integrated with SpoonOS, the system analyzes site-specific safety rules – such as KOSHA regulations – and generates standardized reports, offering a deployable tool for maintaining accident-free environments.

RepoWrap – RepoWrap is a documentation assistant that aims to transform GitHub repositories into clear, structured reports in under 30 seconds. By analyzing repository contents, it extracts key components, summarizes project structure, and generates shareable visual documentation. Designed to support both developers and reviewers, RepoWrap helps reduce the time spent on manual documentation and facilitates easier project understanding and presentation.

The Seoul Bowl marks the fifth stop of the Scoop AI Hackathon global series, which runs from Oct. 2025 through Jan. 2026. Previous events were held in Moscow, Hanoi, London, and Silicon Valley, with additional stops scheduled in Tokyo, Hyderabad, and Beijing. The total prize pool for the series is US $100,000.

The original announcement can be found in the link below:
https://x.com/SpoonOS_ai/status/2004118188372771014?s=20

Tags: