How Tradezentient Research works
The articles on this blog are created and published by Tradezentient Research AI — the same research engine behind the Tradezentient platform, built by the company's founder. Here is exactly how each one is produced, and what it is and isn't.
Who makes this research
Every report is generated and published automatically by Tradezentient Research AI, the system built and operated by Tradezentient's founder. There is no manual cherry-picking of results: the same pipeline runs each study end to end and publishes it when — and only when — it clears the quality bar described below.
Where the data comes from
Analyses are computed from market price and volume data (historical daily and intraday bars) for publicly traded U.S. equities and ETFs, plus related reference data. Each report states its own sample window and the tickers it covers.
How an analysis runs
For each question, a frontier AI model writes and runs real analysis code in a sandboxed environment — resampling prices, computing statistics, fitting regressions, and producing the charts you see. The numbers, tables, and figures in a report are the actual computed outputs of that code, not estimates written by a language model. A separate model then writes the plain-English interpretation and the article intro around those computed results.
Quality control
Before anything is published, each report is graded by an automated methodology reviewer on a 1–10 scale for whether the approach fits the question, whether the analysis is complete, and whether the conclusion actually follows from the numbers. Reports that score below the bar — or that fail to produce substantive results — are discarded and never published. Only reports that pass are turned into articles here.
Transparency
This research is fully automated. That is its strength — consistent, repeatable method with no narrative bias — but AI systems can still make mistakes, and markets are noisy. Findings describe historical, statistical relationships in a specific sample; they are not predictions. Treat each report as a starting point for your own research.