Seth
5b28002001
Major changes from this session: Training: - 0.6.0 training running: 9B on steel141 3090 Ti, 27B on rented H100 NVL - 7,256 merged training examples (up from 3,183) - New training data: failure modes (85), midloop messaging (27), prompt injection defense (29), personality (32), gold from quarantine bank (232), new tool examples (30), claude's own experience (10) - All training data RCON-validated at 100% pass rate - Bake-off: gemma3:27b 66%, qwen3.5:27b 61%, translategemma:27b 56% Oracle Bot (Mind's Eye): - Invisible spectator bot (mineflayer) streams world state via WebSocket - HTML5 Canvas frontend at mind.mortdec.ai - Real-time tool trace visualization with expandable entries - Streaming model tokens during inference - Gateway integration: fire-and-forget POST /trace on every tool call Reinforcement Learning: - Gymnasium environment wrapping mineflayer bot (minecraft_env.py) - PPO training via Stable Baselines3 (10K param policy network) - Behavioral cloning pretraining (97.5% accuracy on expert policy) - Infinite training loop with auto-restart and checkpoint resume - Bot learns combat, survival, navigation from raw experience Bot Army: - 8-soldier marching formation with autonomous combat - Combat bots using mineflayer-pvp, pathfinder, armor-manager - Multilingual prayer bots via translategemma:27b (18 languages) - Frame-based AI architecture: LLM planner + reactive micro-scripts Infrastructure: - Fixed mattpc.sethpc.xyz billing gateway (API key + player list parser) - Billing gateway now tracks all LAN traffic (LAN auto-auth) - Gateway fallback for empty god-mode responses - Updated mortdec.ai landing page Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Mortdecai
A 9B parameter language model fine-tuned for Minecraft server operations. Translates natural language to commands, controls an AI God character, manages plugins, writes mcfunction scripts, and learns from its mistakes.
Base model: Qwen3.5-9B | Current version: 0.5.0 | Quantization: Q4_K_M (5.6GB)
Training Progress
| Version | Base Model | Training Examples | Loss | Key Addition |
|---|---|---|---|---|
| 0.1.0 | Qwen3-8B | 500 | 2.10 | Seed data only |
| 0.2.0 | Qwen3-8B | 1,200 | 1.45 | +entities, +mobs |
| 0.3.0 | Qwen3-8B | 2,100 | 0.82 | +error correction |
| 0.4.0 | Qwen3.5-9B | 3,175 | 0.35 | +tool-calling, base model upgrade |
| 0.5.0 | Qwen3.5-9B | 4,358 | 0.16 | +plugins, +memory, +scripts |
Bake-off: 0.5.0 vs 0.4.0
| Category | 0.4.0 | 0.5.0 | Change |
|---|---|---|---|
| Enchantments | 20% | 67% | +47% |
| EssentialsX | 0% | 60% | +60% |
| Effects | 0% | 25% | +25% |
| Basic commands | 75% | 75% | — |
| Teleport | 100% | 100% | — |
| Overall | 45.2% | 46.8% | +1.6% |
Architecture
17 tools across 5 categories:
| Category | Tools |
|---|---|
| Execution | rcon.execute |
| Knowledge | minecraft.wiki_lookup, plugin.docs_lookup, minecraft.changelog_lookup, paper.docs_lookup |
| World | world.player_info, world.server_state, world.nearby_entities |
| Memory | memory.read, memory.write |
| Scripts | script.write, script.validate, script.execute, script.read, script.list, script.delete, script.schedule |
Plugins: FastAsyncWorldEdit, WorldGuard, CoreProtect, EssentialsX, Vault, LuckPerms
Training Data
~20,000+ examples from:
- Hand-curated seed data (3,196)
- Tool-calling sequences with 17 tools (1,430)
- IGLU build dataset — Microsoft Research (4,656)
- RCON-validated plugin examples (104)
- Exploration self-play with wiki grounding (150)
- Self-play across 3 GPUs (2,900+)
- Live server audit from wolf bots + real players (8,000+)
Infrastructure
| GPU | Role |
|---|---|
| RTX 3090 Ti (24GB) | Training + self-play |
| RTX 2080 Ti (11GB) | Exploration self-play |
| Quadro RTX 4000 (8GB) | Production inference — 3 MC servers |
| GTX 1660 Super (6GB) | Prompt generation |
GPU Scheduler: gpu.sethpc.xyz — preset-based job scheduler with live monitoring
Links
- Play:
minecraft.mortdec.ai - Model card: MODEL_CARD.md
- Domain: mortdec.ai