InsiderEdge
InsiderEdge is a fully autonomous AI crypto trading agent that monitors BTC, ETH, and SOL 24/7 without human input. It calculates real RSI, MACD, trend, and momentum signals from live price history, then uses Groq LLaMA 3.3 70B AI to make buy/sell/hold decisions using a multi-confirmation scoring system — only trading when multiple signals agree. Built-in risk management includes a circuit breaker at -5% daily loss, stop loss at -0.8%, take profit at +1.2%, and a minimum 55% confidence threshold before any trade executes. InsiderEdge is registered on the ERC-8004 Identity Registry on Sepolia testnet as Agent ID 33. Every decision is posted as a verifiable on-chain checkpoint to the ValidationRegistry — no trust required, just check the contract. Live results: 15 trades completed, daily PnL of +2.56%, validation score 77, reputation score 20, ranked top 31 on the official leaderboard. Built in 5 days by a 3rd year CS student with zero prior trading experience. Tech stack: Python, Groq API, CoinGecko, Kraken CLI, Web3.py, ERC-8004 Sepolia contracts.
Category tags:
.png&w=96&q=75){kind=small}
.png&w=96&q=75){kind=small}
" 1. Application of Technology: 3.5 / 5 Justification: The codebase is a well-structured set of Python scripts. The integration with web3.py to interact with the ERC-8004 registries on Sepolia erc8004.py is functional and correctly configured. Utilizing Groq and rotating API keys explicitly in the code handles rate limits gracefully. 2. Presentation: 4 / 5 Justification: Very good for a solo student developer. Including the raw log files (trade_log.txt, checkpoints.jsonl) proves the bot actually ran and validates their claim of completing 15 trades. 3. Business Value: 3 / 5 Justification: Moderate. As an educational tool or a starting point for building a personal trading bot, it has value. However, replacing PRISM API market data with simple CoinGecko 24-hour price changes and using hardcoded overrides (e.g., if change_24h > 1.5: in ai_brain.py) means the AI's actual contribution to the strategy is easily bypassed, making it less attractive for enterprise use. 4. Originality: 3 / 5 Justification: It executes the exact baseline prompt (Python script -> LLM -> Web3.py execution -> Kraken execution). It does not introduce any radical new engineering architecture or complex logic loops compared to the multi-agent debate or Rust-based systems. ⚖️ Pros & Cons Pros: Fully Functional End-to-End: Unlike some projects that mocked their execution, this repository actually contains the logic to assemble smart contract inputs and POST them to Sepolia via an RPC provider. Built-in API Key Rotation: The ai_brain.py dynamically rotates between two distinct Groq API keys to prevent hitting usage limits during the 10-minute scan intervals, which is a smart DevOps workaround. Transparent Success/Failure: The developer was very honest about the setup, including error logs (blockchain_errors.txt) which shows exactly how they debugged the smart contract integrations during the hackathon. Cons: Simple Logic Overrides: Inside ai_brain.py, the developer hardcoded momentum overrides (if change_24h > 1.5: ...) that essentially bypass the LLaMA 70B model's reasoning process if a coin moves up by 1.5%. This reduces the "AI Trading Agent" to a basic percent-change script in trending markets. Basic Data Feed: Relying solely on the free CoinGecko API for 10-minute snapshot data means the bot lacks the deep order book or microstructure data (like open interest or funding rates) required to create a real algorithmic edge against other bots. "
Sanem Avcil
For Innovators & Creators