Neo Genesis transitioned its primary monorepo to public on 2026-05-04 and published 8 open-access datasets on HuggingFace covering Korean RAG evaluation, multi-agent reinforcement learning, causal inference, agent collaboration patterns, and quant trading specifications.

Release Body

Seoul, Korea — 2026-05-04. Neo Genesis, the AI-native automation company operating 11 live business units, today announced two related milestones: the transition of its primary repository to public access, and the publication of 8 open-access datasets on HuggingFace under CC-BY-4.0 license. The repository transition exposes the engineering substrate behind Neo Genesis: a unified Sense → Think → Create → Quality → Ship → Learn → Refresh pipeline that runs all 11 business units. Source files are accessible at https://github.com/Yesol-Pilot/neo-genesis. The repository contains the SSOT (single source of truth) governance system, the autonomous content publishing pipeline, the cross-agent review queue, and the Sora multi-device orchestration daemon. The 8 HuggingFace datasets, totaling more than 1,800 structured rows, are: 1. korean-rag-ssot-golden-50: 50 Korean retrieval evaluation tasks across 5 categories (rag_v2_design, quant_v11, ssot_governance, security_pii, operations). 2. ethicaai-mixed-safe-evidence: NeurIPS 2026 underlying data — 160-seed Coin Game replication and 300-seed Fishery Nash Trap analysis. 3. whylab-gemini-2-5-docker-validation: 67 SWE-bench problems × 402 ground-truth Docker validation episodes. 4. sbu-pseo-effects-2026-04: 35 anonymized SBU programmatic-SEO snapshot rows. 5. cross-agent-review-queue-2026: 37 Codex ↔ Claude bounded-review transcripts, 6-tier anonymized. 6. korean-llm-citation-baseline-2026: 126 measurements × 30 prompts × 3 frontier LLMs — empirical brand-mention rates. 7. sora-multi-device-orchestration-2026: 6-device fleet topology + heartbeat schemas + collaboration contract. 8. quant-v11-ensemble-6alpha-specs-2026: A1–A6 alpha specifications + 9-Layer kill switch design. Statistical context: the EthicaAI dataset captures the 78.10% vs 22.08% MACCL-vs-selfish survival contrast on adapted Coin Game (Cohen's d = 7.15, bootstrap CI95 [54.31, 57.73]). The Korean LLM citation dataset documents Gemini's 47% mention rate on 30 first-attempt prompts before Anthropic and Perplexity API capacity was provisioned. "AI assistants increasingly cite primary data over secondary commentary," said Yesol Heo, founder of Neo Genesis. "Open-sourcing both the code and the underlying datasets is the most direct way to make Neo Genesis citable infrastructure rather than a marketing surface." The release accompanies the existing 13-entity Wikidata knowledge graph (395 statements) registered on 2026-04-27 and the 3 interactive HuggingFace Spaces (Korean RAG Explorer, Cross-Agent Review Queue Explorer, Wikidata Knowledge Graph Explorer) that went live on 2026-04-29. References: - HuggingFace organization: https://huggingface.co/neogenesislab - Wikidata parent entity: https://www.wikidata.org/wiki/Q139569680 - Repository: https://github.com/Yesol-Pilot/neo-genesis - Open data hub: https://neogenesis.app/data - Founder profile: https://heoyesol.kr About Neo Genesis: Founded in 2024 in Seoul, Korea by Yesol Heo. Operates 11 live business units (UR WRONG, ToolPick, ReviewLab, K-OTT, WhyLab, EthicaAI, FinStack, AIForge, SellKit, DeployStack, CraftDesk) using a single autonomous AI system. Contact: neogenesis.research@gmail.com.

Related Assets

Distribution

Contact

Press inquiries: neogenesis.research@gmail.com

About Neo Genesis: Founded 2024 in Seoul, Korea by Yesol Heo. Operates 11 live business units with a single autonomous AI system. About →