{"repo":"adam-s/alphadidactic","free":true,"listed":false,"github":"https://github.com/adam-s/alphadidactic","clone":"git clone https://github.com/adam-s/alphadidactic.git","description":"An iteration research agent: searches academic research, applies it to time series data, and probes it to find novel discoveries.","language":"Python","stars":72,"topics":["backtesting","quant","stock-price-prediction","timescaledb"],"license":null,"category":"trading","readme_excerpt":"Alphadidactic An iteration research agent: searches academic research, applies it to time series data, and probes it to find novel discoveries. Claude Code instructions—not hand-written strategies—build, verify, and optimize quantitative experiments through a gated pipeline. Each step must pass before the next begins: literature search, dataset construction, strategy implementation, independent verification, and parameter optimization. The Two Rules 1. All data used to produce a signal must exist before the position it triggers is entered or exited. Future data is not allowed to enter the system. 2. All optimization and parameter tuning use only training data. Out-of-sample test data that was never trained on is the real measure of whether a signal generalizes. Everything else is up to Claude. The Loop Three agents—experiment, reviewer, adversary—cycle on every experiment. Shared infrastructure makes entire bug classes impossible by construction rather than relying on instructions the agent might skip. Every iteration that exposes a gap in the instructions gets patched immediately—the instruction set evolves through failure, not design. The Hard Problems Bugs in quantitative research don't crash—they produce beautiful equity curves. A same-day return pairing or a dropped timezone conversion inflates Sharpe by 1–2 points and passes every naive check. The most dangerous variant is a self-referential audit: we built an honesty audit around a strategy with a test Sharpe of 3.9, a","default_branch":null,"files":null,"tree":[],"storefront":"/r/adam-s","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/adam-s/alphadidactic/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}