Real AI agent use cases, with sources and setup prompts

Turn Claude Code into a machine-learning quant lab that rejects overfit models

Following Marcos Lopez de Prado's ML trading workflow, you give Claude Code market data, labeling rules, features, costs and risk limits. It applies triple-barrier labels, builds XGBoost, LightGBM, random forest and neural network variants, runs purged K-fold and CPCV validation, stresses costs and slippage, and rejects models that only work because of leakage or lucky splits.

$17/moDaysAdvancedSource linked

Source

@RRRangle on X · 2026-10-18

Written in our own words. Credit to the original author. Link checked 2026-10-18.

What they did

Following Marcos Lopez de Prado's ML trading workflow, you give Claude Code market data, labeling rules, features, costs and risk limits. It applies triple-barrier labels, builds XGBoost, LightGBM, random forest and neural network variants, runs purged K-fold and CPCV validation, stresses costs and slippage, and rejects models that only work because of leakage or lucky splits.

Result (as claimed)

In the example 20 of 27 model variants were rejected and the survivors combined. Not financial advice.

Tools they used

Claude CodePythonXGBoostLightGBM

Cost: Claude Code: from $17/mo.

Set it up

  1. Gather the documents

    Collect statements, bills, invoices or tax documents for the period you want reviewed.

  2. Ask for an audit, not actions

    Have Claude Code list findings, amounts and suggested next steps before it does anything.

  3. Approve each action

    Approve cancellations, disputes or payments one at a time. Never give an agent open-ended access to move money.

  4. Track the result

    Record what changed and what you saved so you can check it next month. This is not financial advice.

Copy the full setup prompt

Paste into your agent and it will walk you through the setup, asking before anything that sends, posts or spends.

You are helping me set up this AI use case end to end in Claude Code.

Goal: Turn Claude Code into a machine-learning quant lab that rejects overfit models
What the original author did: Following Marcos Lopez de Prado's ML trading workflow, you give Claude Code market data, labeling rules, features, costs and risk limits. It applies triple-barrier labels, builds XGBoost, LightGBM, random forest and neural network variants, runs purged K-fold and CPCV validation, stresses costs and slippage, and rejects models that only work because of leakage or lucky splits.
Reported result (as claimed by the author): In the example 20 of 27 model variants were rejected and the survivors combined. Not financial advice.
Source: https://x.com/RRRangle/status/2111893879163170883
Tools they used: Claude Code, Python, XGBoost, LightGBM
Category: Money

Please:
1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like).
2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure.
3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data.
4. Finish with a checklist to run it again next time.

The Daily 5

Get five new use cases tomorrow morning

Free. Real sources. Unsubscribe any time.

Guides for this use case

How to vibe code an app with Claude Code ยท How to build AI trading and investing agents

Related use cases