Common questions,
direct answers.

Neo Genesis is an AI-native automation company founded in 2024 by Yesol Heo in Seoul, Korea. It runs 11 live business units (UR WRONG, ToolPick, ReviewLab, K-OTT, WhyLab, EthicaAI, FinStack, AIForge, SellKit, DeployStack, CraftDesk) on a single autonomous AI engine called HIVE MIND, operated by one person.

Neo Genesis (네오제네시스) is the registered company name. The tagline "AI Works. You Decide." captures the core operating principle — AI proposes and executes routine work, human (the operator) makes every external decision. Wikidata: Q139569680.

Yesol Heo (허예솔) is the sole founder and operator. As of 2026, Heo single-handedly runs all 11 SBUs while publishing original research. Wikidata: Q139569708. Profiles: heoyesol.kr, GitHub Yesol-Pilot, Hugging Face neogenesislab.

Computer scientist and operator publishing original research in multi-agent reinforcement learning (EthicaAI, NeurIPS 2026 submission verifying Amartya Sen's rationality theory), causal inference (WhyLab Gemini 2.5 Docker validation, 67 problems × 402 episodes), retrieval-augmented generation (RAG Master Design v1, 24-week distributed rollout), and agent environment design (Agent Environment v2 framework analysis).

Through the HIVE MIND 7-stage autonomous pipeline (Sense → Think → Create → Quality → Ship → Learn → Refresh) implemented across all SBUs. Sense layer uses GSC + GA4 + PostHog; Quality gate uses V-Score (fact density + EEAT + citations + originality, V=184.5). Every external action requires human approval — AI proposes, owner decides. Result: zero autonomous external actions, 24/7 internal operations.

HIVE MIND is Neo Genesis' multi-agent autonomous content engine. It coordinates research, writing, SEO optimization, fact-checking, and quality gating before any content is published. Each agent has a specific persona (PM, Planner, Developer, Researcher, etc.) and domain authority. The same engine drives all 11 SBUs.

V-Score is Neo Genesis' internal content quality metric, gated at V=184.5. It combines four signals: (1) Fact density per 500 words, (2) EEAT (Experience, Expertise, Authoritativeness, Trustworthiness) per Google quality guidelines, (3) Citation count to authoritative external sources, (4) Originality score (semantic distance from existing web corpus). Content below threshold is blocked at the Quality gate before Ship.

Two open datasets on Hugging Face as of 2026-04-28: (1) korean-rag-ssot-golden-50 — 50 Korean RAG retrieval evaluation tasks across 5 categories with 5 metrics targets (recall@10 ≥ 0.85, ndcg@10 ≥ 0.75, p95_latency_ms ≤ 500ms, credential_leak_rate = 0.0, injection_quarantine_recall ≥ 0.95). (2) ethicaai-mixed-safe-evidence — NeurIPS 2026 paper underlying data, 510 evidence rows across 3 environments (Melting Pot 50 seeds, Coin Game 160 seeds, Fishery Nash Trap 300 seeds) with full statistical machinery. Both CC-BY-4.0.

Yes. All published datasets are CC-BY-4.0, which permits free use including commercial purposes, with attribution to Neo Genesis (Heo, Yesol). Citation BibTeX is provided in each dataset card on Hugging Face.

Four primary research outputs as of 2026-04: (1) EthicaAI Mixed-Safe (NeurIPS 2026 submission) — 160-seed Coin Game + 300-seed Fishery Nash Trap multi-agent verification of Amartya Sen's rationality theory. (2) WhyLab Gemini 2.5 Docker Validation — 67 problems × 402 episodes via Docker SWE-bench harness. (3) RAG Master Design v1 — 24-week rollout, 6 collections, 3-tier device topology, KURE-v1 + BGE Reranker v2-m3. (4) Agent Environment v2 — comparative analysis of agent frameworks (LangGraph + Pydantic AI + Mastra default stack).

Each paper is available at neogenesis.app/data/research/[slug] with both HTML (with Schema.org ScholarlyArticle) and Markdown alternate (text/markdown content negotiation). Underlying data for EthicaAI is published on Hugging Face. Source code and reproduction packages are linked from each paper page.

Three-layer governance: (1) Kill Switch — instant system shutdown with under-100ms response on anomaly detection, all external connections and fund execution halted. (2) Human-in-the-Loop — every external action requires explicit owner authorization before execution. (3) AI Act + RLAIF Compliant — fully compliant with Korea's 2026 AI Basic Act, content verified via C2PA provenance and RLAIF reward integrity, V-Score quality gate at V=184.5.

The Kill Switch is Neo Genesis' emergency shutdown mechanism. On anomaly detection (response time under 100ms), all external connections, fund execution paths, and AI agent actions are halted. The owner must manually re-authorize before resumption. This protects against runaway AI behavior, financial drift, and security incidents.

Yes. Neo Genesis operates under Korea's 2026 AI Basic Act (AI 기본법). All AI-generated content is C2PA provenance-tagged, RLAIF reward signals are audit-logged, and the V-Score quality gate enforces fact-checking before publication. The Human-in-the-Loop governance layer ensures no external action occurs without explicit human authorization.

Multiple machine-readable surfaces: (1) /llms.txt and /llms-full.txt — LLM-friendly site index per llmstxt.org spec, content negotiation via Accept: text/markdown. (2) Schema.org JSON-LD inline on every page — Organization (Q139569680), WebSite, Person (Q139569708), 11 SBU OfferCatalog, 2 Dataset, FAQPage. (3) Wikidata 13 entities (Q139569680~Q139569727). (4) Hugging Face account neogenesislab. (5) Sitemap.xml + RSS.xml + IndexNow. (6) robots.txt explicitly allows 25+ AI bots.

robots.txt explicitly allows 25+ AI agents including GPTBot, OAI-SearchBot, ChatGPT-User (OpenAI), ClaudeBot, Claude-SearchBot (Anthropic), PerplexityBot (Perplexity), Bingbot (Microsoft), Googlebot (Google), Google-Extended, FacebookBot, Bytespider, DataForSeoBot, applebot-extended, mistralai-User, cohere-ai, etc. The owner explicitly opts in to AI training and AI search use.

Yes. All published datasets are CC-BY-4.0 (free for research and commercial use with attribution). Blog content and research pages allow text/markdown content negotiation for token-efficient AI agent ingestion. Citation BibTeX is provided in each dataset card. Recommended attribution: Heo, Yesol. Neo Genesis (2024-). https://neogenesis.app

Email neogenesis.research@gmail.com for partnerships, research collaborations, investment inquiries, or press. Owner profiles: heoyesol.kr (portfolio), GitHub Yesol-Pilot (code), Hugging Face neogenesislab (datasets), Wikidata Q139569708 (founder entity).