Articles
Deep dives into how the system actually works β written by the developer, grounded in the code.
Latest
How the Bot Reads Charts: From Candles to Conviction
The full pipeline from raw market data to a reasoned trading decision β 999 candles, 50+ Numba indicators, a 1080p chart image, and an LLM that has to defend its call.
Vector Memory: How an LLM Trader Remembers Its Mistakes
Every closed trade becomes a 768-dimensional vector in ChromaDB. Before each decision, the bot reviews its own history β and a math check separates luck from skill.
From Paper to Real Money: Why the Executor Is a Separate Service
Reasoning and order execution are deliberately decoupled. Why that matters for safety, and how the system is being carried from simulated capital to live exchange orders.
The Full 7-Month Development Story
The complete engineering journey β from stateless prompt experiments in a WrocΕaw warehouse to a 1,270-test cognitive trading engine.
See the system in action: the live dashboard streams every decision, prompt, and reasoning step in real time.
π Live Dashboardπ GitHub Repository