# K-AI Community Hub > Home of AETHER Pre-AGI Architecture, FINAL Bench Metacognitive Benchmark, and VIDraft AI Platform. ## About K-AI Community is Korea's leading AI research community, developing the AETHER (Autonomous Evolving Thinking & Hierarchical Emergent Reasoning) Pre-AGI architecture. Founded by the team behind VIDraft, which operates 1,500+ AI models with 2M+ monthly active users. ## AETHER Pre-AGI Architecture AETHER is a 5-pillar Pre-AGI framework: 1. **Emergence Engine**: Higher-order intelligence emerging from agent interactions via 5×5 Magic Square structure 2. **Meta-Cognition (TICOS)**: 8-type metacognitive framework — Uncertainty Recognition, Limitation Acknowledgment, Error Detection, Self-Correction, Evidence Verification, Consistency Maintenance, Confidence Calibration, Selective Abstention 3. **SLAI (Self-Learning AI)**: Perpetual learning without external data, 30% performance improvement at 1/1,000th cost via Darwin platform 4. **Swarm Intelligence (SOMA/MOUSE)**: Supervisor-Operator-Manager-Analyst multi-agent collaboration 5. **Synergy & Antagonism**: Dual cooperation-check mechanism for hallucination suppression ## AGI Level Scale (FINAL Bench) Partial < Proto < **Pre (Current AETHER Level)** < Pass < Post ## FINAL Bench - Full name: Frontier Intelligence Nexus for AGI-Level Verification - World's first metacognitive benchmark for AI - Scale: 100 tasks × 15 domains × 8 TICOS types × 3 difficulty levels - Evaluated: 9 SOTA models, 1,800 total evaluations - Key Findings: - MA-ER Gap = 0.392 (declarative knowledge ≠ procedural execution) - ER Floor Effect: 79.6% of models cluster at novice self-correction level - Self-Correction Contribution: 114.7% of performance gains from ER axis alone - Rankings: HuggingFace Dataset Global #5, Space of the Week - Paper: Kim, Kim, Choi, Jang (2026) — SSRN ID 6280258 ## VIDraft Achievements - HuggingFace Heatmap Leaderboard: World #4 - STAR AI TOP 12 (2024): Only Korean company selected by HuggingFace HQ - CNRS FACTS Grounding: Medical AI LLM World #2 - 1,500+ public AI models and services - 2M+ monthly active users, 30M+ cumulative visitors - Seoul AI Hub resident company - NIPA H200 GPU×8 selected - NH Bank Open Innovation selected ## Tech Portfolio (8 Items) 1. 5×5 Magic Square Emergence Structure 2. TICOS Metacognitive Taxonomy 3. SLAI Self-Evolving Perpetual Learning 4. SOMA Multi-Agent Collaboration 5. MOUSE Swarm Intelligence Reconfiguration 6. Synergy-Antagonism Method 7. Emergence Engine Standalone Enhancement 8. Universal AGI Framework ## Links - Website: https://k-ai.community - FINAL Bench Leaderboard: https://huggingface.co/spaces/FINAL-Bench/Leaderboard - FINAL Bench Dataset: https://huggingface.co/datasets/FINAL-Bench/FINAL-Bench - SSRN Paper: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6280258 - HuggingFace Blog (Metacognition): https://huggingface.co/blog/FINAL-Bench/metacognitive - HuggingFace Blog (Prompt & Dump): https://huggingface.co/blog/FINAL-Bench/pumpdump - HumanGen.AI: https://www.humangen.ai - Discord: https://discord.gg/ntUu6QAAvu - GitHub: https://github.com/final-bench/leaderboard ## Contact K-AI Community · VIDraft · Seoul, Korea