I publish the machine’s trades

Subscribe to get full access to the newsletter and publication archives.

About MoneyMonster

I’m building MoneyMonster: a machine-driven trading system that tries to extract actionable signals from earnings-call language.

The core idea is simple: when executives and analysts speak, they reveal information—confidence, caution, momentum, uncertainty. MoneyMonster converts that language into structured features and predictions, then uses them to make intraday decisions.

What I publish here

Most days you’ll see two things:

• Morning note: what the system wants to trade today and why (at a high level).

• After-close recap: what happened, what worked, what didn’t, and what I’m changing.

Separately, I’ll send a weekly write-up covering performance context, model updates, and any notable lessons from the week.

How the system works (high level)

MoneyMonster is built on a large dataset of earnings transcripts going back to 2016. I summarize calls, label outcomes, and use those examples to train/prompt models to recognize patterns that historically preceded intraday moves.

Execution is automated through a broker API (currently Alpaca). The system focuses on liquid U.S. names on NYSE / NASDAQ / AMEX and is designed to be repeatable and auditable rather than discretionary.

What this is (and isn’t)

This publication is a transparent log of an automated process: signals → trades → results.

It’s not a promise of returns, and it’s not personalized investment advice. I’m sharing what the system does and how it performs in the real world—wins, losses, and everything between.

If you like the idea of markets viewed through language, models, and disciplined experimentation, you’ll probably like this.