I have been building an automated trading system called HumbleExplorer. It is not a DEX or a lending protocol itself. It is a bot that executes through its own smart contracts, using an existing money market as leverage infrastructure.
The system monitors price action in real time and executes trades through upgradeable smart contracts. Positions are over-collateralized, which means the bot can go long or short with borrowed capital rather than requiring full spot liquidity for every trade.
The architecture separates concerns across multiple layers. The Solidity contracts use a UUPS upgradeable proxy pattern for the core execution logic. The TypeScript bot handles price monitoring, gas estimation, and transaction submission. The Rust key-manager is the secure signing infrastructure (no plaintext keys anywhere). And the ML signal layer is a small language model trained on market data for directional bias.
Using a money market as the base layer instead of perpetuals or spot DEXs gives us a few advantages. Built-in liquidation mechanics provide safety margins. Interest accrues continuously rather than via funding rates. And multi-collateral support lets us dynamically shift between volatile assets and stablecoins depending on market conditions.
The contracts have been live with real capital for about a month now. The bot runs autonomously — I get alerts, not tickets. I am not sharing profit figures yet because one month does not prove anything and I do not want to turn this into a marketing page before the strategy has survived a real drawdown.
So far the system has behaved as expected. Gas costs are within the model. Health factor margins held during a volatility spike three weeks ago. The key-manager has not required a single manual unseal. These are the metrics I care about more than PnL right now — infrastructure resilience over alpha.
This project will stay closed-source. I may offer HumbleExplorer as a managed service down the road, so the code will remain private.