Skip to content

From hypothesis to astrategy that holds up.

Describe an idea in plain language. The loop turns it into code, tests it twice, and brings the results back for the next revision.

  1. 01

    Start with a hypothesis.

    Write what you want to test in plain language: a market observation, a pattern, a question.

  2. 02

    The model writes the code.

    Your sentence becomes a Python strategy that runs on the platform’s own engine.

  3. 03

    Test it on history.

    The strategy runs in an isolated sandbox against our historical data.

  4. 04

    Then on the live market.

    It trades with simulated orders on real-time data, so it meets conditions history cannot show.

  5. 05

    Send the results back.

    The model reads the results and proposes revisions. You decide what to keep.

  6. 06

    Connect only what holds up.

    A strategy that survives both tests can be connected to live trading.

  1. 01Idea
  2. 02Code
  3. 03Backtest
  4. 04Paper trading
  5. 05Refine
  6. 06Live

Idea

Do themes that lead on day one still lead on day three?

strategy.py

Backtest · isolated sandbox · historical data

Real-time data · simulated orders

Before

Revised

Connected to live trading

> hypothesis received

Illustrative log, not a real run.

Beyond the strategy loop.

  • AI stock research

    Pick a stock and keep asking. The model reads quotes, minute and daily bars, backtests and factors, and writes down the data each answer rests on.

  • Factor mining

    The model drafts factor formulas from our own data panel. Each is scored on history and explained in a plain sentence before the formula is shown.

  • Checked afterwards

    Stock briefs are compared with the next one, three and five trading days; daily market reviews number every piece of evidence and are reconciled the next day.

What stays with the trader.

  • You decide what to test.

    The model drafts and revises. The hypothesis, and the call on what to keep, are yours.

  • Tested before trusted.

    A strategy reaches live trading only after a backtest and a paper-trading run.

  • Evidence, not advice.

    We publish no stock picks and no investment advice. The output is evidence for research.

Bring a hypothesis.

Research services are planned within 6–12 months. Tell us what you would want to test.

Request early access