Descriptions:
David from Trading with DaviddTech documents how he used Claude Code to build a working Bitcoin and Ethereum scalping bot targeting Poly Market prediction markets, reporting $21,000 in returns over 90 days. But the video is as much a debunking exercise as a tutorial: it opens by demonstrating live how Claude’s design tools can generate convincing-looking profit dashboards in seconds — the same kind flooding social media — with zero actual trading logic behind them.
The core technical argument is a clean distinction between large language models, which predict the next token in a sequence, and time series models, which predict the next number — the latter being the right architecture for financial forecasting. The bot is built around Kronos, an open-source GitHub repository trained on financial time series data, run locally on a Mac Mini. Claude Code handles the integration work: wiring together a dashboard, a database, and the Kronos connector through a step-by-step prompting workflow that the creator has packaged into a free workbook.
Backtest results across 50,000 trials are reported with specificity: a base Kronos model augmented with ATR volatility bands and ADX trend filters generated $12,300 over 90 days on five-minute Bitcoin and Ethereum markets. Adding a Martingale-plus-Kelly position sizing layer pushed the figure to $23,300. The video is notable for its transparency — showing the full methodology, explaining why the fifteen-minute timeframe underperformed, and providing all prompts and code for replication — making it a rare grounded example of AI-assisted quantitative system construction.
📺 Source: Trading with DaviddTech · Published May 04, 2026
🏷️ Format: Hands On Build
