Descriptions:
This tutorial from the All About AI channel walks through a complete workflow for using OpenAI’s Codex 5.5 as an agentic assistant to build, backtest, and validate algorithmic trading strategies on HyperLiquid, a decentralized perpetuals exchange. The creator demonstrates what he calls a “trading pod” approach — building isolated, autonomous strategies that can be stacked for diversified exposure — using Bitcoin as the test asset.
The workflow moves through several structured phases: prompting Codex to propose three candidate strategies (a volatility-targeted trend breakout, an intraday mean reversion, and a funding premium carry), generating best-practice backtesting frameworks for each, collecting historical HyperLiquid data, and running the backtests. The most instructive segment is the robustness check: when the trend breakout strategy shows initially promising results, Codex is directed to run Monte Carlo and walk-forward optimization targeting a Sharpe ratio of 1.2 — and the strategy fails, scoring only 0.4, with a clear recommendation not to trade it live.
The video is notable for its methodological honesty and its illustration of how an AI agent can enforce trading best practices, including overfitting detection, rather than simply validating whatever the user hopes to find. The prompt patterns used — specifying Sharpe targets, requesting Monte Carlo stress tests, flagging forward-looking data bias — translate well beyond trading into any domain where Codex is used to evaluate quantitative outputs.
📺 Source: All About AI · Published June 22, 2026
🏷️ Format: Tutorial Demo
