Research
Structured, semi-automated R&D with Claude Pro
Claude Pro is the reasoning partner in our research loop — not an autopilot. Every stage pairs a Claude-assisted step with a human-owned decision, so the student stays the author of the work and the demo account stays the final judge.
Hypothesis
Claude Pro
Claude helps sharpen the claim into something falsifiable and ties it to the math.
Student
The student owns the question and the prior — Claude refines, never originates.
Data & features
Claude Pro
Claude drafts data plumbing and feature transforms; flags leakage and look-ahead.
Student
The student verifies the data is clean, in-sample, and honestly out-of-sample.
Strategy & code
Claude Pro
Claude drafts the StrategyQuantX strategy and the surrounding analysis scripts.
Student
The student reads every line, tests it, and understands why it works before running it.
Backtest & robustness
Claude Pro
Claude proposes robustness checks and interprets the results in plain math.
Student
The student decides whether the edge is real or an artefact of the test.
Review
Claude Pro
Claude summarises the run, the risks, and the open questions for the next pass.
Student
The student signs off — or sends it back. Nothing graduates without a human yes.
Iterate
Claude Pro
Claude carries the running notebook of hypotheses tried and rejected.
Student
The student keeps the discipline: demo first, live later, always.
The guardrails
- Claude drafts; the student reviews and owns every line that runs.
- No strategy is deployed from a prompt — it must pass the demo pipeline.
- The research notebook is reproducible: data, code, backtest, reasoning.
- Live capital is the last step, sized and capped, never the first.
Semi-automated, not automatic
Automation speeds the loop; judgement keeps it honest. Claude Pro compresses the time from idea to tested hypothesis, and the student spends that saved time on the part that matters: deciding whether the edge is real.