> For the complete documentation index, see [llms.txt](https://gameland-1.gitbook.io/galeon/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://gameland-1.gitbook.io/galeon/galeon-brain.md). # Galeon Brain ## ### Overview Galeon Brain is a **DeFi Cognition Engine** — it understands crypto market dynamics and outputs structured market intelligence for any trading agent, DApp, or wallet to consume. Brain answers one question: **What stage is this token in, where is it going, and why?** It is not a trading bot. It does not execute trades. It **thinks**. *** ### How It Works Brain combines two layers: * **Rules Engine** — Code handles all repeatable analysis: statistics, pattern detection, stage classification, confidence scoring. Instant, deterministic, always available. * **Thinking Engine (LLM)** — Handles what code cannot: reasoning about *why* patterns exist, diagnosing problems, discovering non-linear signal combinations, judging whether a trading decision makes sense. Code does the work. LLM does the thinking. *** ### Token Lifecycle Brain classifies tokens into seven lifecycle stages: ``` Discovery → Early Breakout → Acceleration → Exhaustion → Distribution → Decline → Rug Danger ``` For each token, Brain predicts: * Which stage it is in now * Which stage comes next (with probability) * Price direction across three timeframes (1h / 4h / 24h) * Risk signals and confidence level *** ### Learning System Brain learns from real trading data — not from assumptions or historical backtests. **DataMiner** continuously analyzes hundreds of real trades to: * Discover where win rates change (breakpoints) * Find multi-factor combinations that outperform or underperform * Automatically generate trading rules from data * Detect systematic problem patterns Example discovery: DataMiner found that funding rate 0.03-0.05% with LONG direction has 80% win rate, while above 0.05% it drops to 35%. This breakpoint was discovered from data, not hardcoded. Rules are updated automatically as new trades come in. Brain today is different from Brain yesterday. **Thinker (LLM)** receives DataMiner's statistical findings and reasons about them: * Why does this pattern exist? * Under what conditions would it fail? * What cross-dimensional relationships might statistics miss? *** ### Cognition Output Brain produces a unified market cognition for any token: ```json { "token": "SKYAIUSDT", "cognition": { "token_stage": "early_breakout", "market_state": "consolidation", "confidence": 0.68 }, "predictions": { "1h": { "direction": "up", "change_pct": 1.2 }, "4h": { "direction": "up", "change_pct": 3.0 }, "24h": { "direction": "sideways", "change_pct": 0.5 } }, "scenarios": [ { "path": "acceleration", "prob": 0.55 }, { "path": "exhaustion", "prob": 0.30 }, { "path": "distribution", "prob": 0.15 } ], "risk_signals": ["funding_rising_watch"], "context": { "regime": "bull", "sector_momentum": "positive" } } ``` *** ### Design Principles **Data Over Assumptions** — Rules come from real trade outcomes, not intuition. Every threshold is discovered, not hardcoded. **Self-Improving** — Every trade outcome feeds back: DataMiner re-analyzes, rules update, parameters auto-adjust. Continuously, measurably getting better. **Fail-Safe** — LLM unavailable? Rules engine handles everything. Data feed interrupted? Historical patterns provide guidance. The system degrades gracefully, never stops. **Separation of Concerns** — Repeatable computation is code. Reasoning and judgment is LLM. Neither does the other's job. *** ### Use Cases | Consumer | How They Use Brain | | ---------------------- | ------------------------------------------------------------------ | | **Trading Agents** | Get structured cognition to decide entries, exits, position sizing | | **DApps / Wallets** | Display token stage, risk level, and market outlook to users | | **AI Agent Platforms** | Use Brain as the market understanding layer for autonomous agents | | **Individual Traders** | Query "is this token safe?" via API or interface | *** ### Current Status Brain has been running in production analyzing DeFi markets: * 4,600,000+ cognition cycles completed * 850,000+ real trades analyzed by DataMiner * 660 auto-generated rules currently active * 270 statistical breakpoints discovered * 1,080,000+ verified experiences in memory * Multi-timeframe predictions (1h / 4h / 24h) * Market regime detection (bull / bear / transition) *** ### Roadmap **Current** — Self-learning cognition engine with data-driven rules, multi-timeframe predictions, and LLM reasoning. **Next** — DeFi Cognition API for external integrations. Real-time trade judgment for agent supervision. Cross-token sector intelligence. Domain-specific fine-tuned model from accumulated experience. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://gameland-1.gitbook.io/galeon/galeon-brain.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.