Structure Beats Brute Force: How 3 Million Legal Documents Built Marqui

When I started practicing law in 2013, my first assignment was a lesson in the limits of human cognition. I was reviewing millions of documents for a Fortune 50 client.

The quotas were brutal: 57-hour weekly minimums. Sit. Click. Tag. Repeat. It was factory work for the mind.

Hacking the Haystack

Understandably, I tried to hack the system almost immediately. I spent my “free” time obsessing over better search strings and smarter filters. I was desperately trying to find the needles without drowning in the hay.

This was the moment I began to understand how databases actually work—fields, queries, logical operators—and more importantly, how much leverage you gain from structuring information correctly.

2016: The Leverage Breakthrough

Fast forward to 2016. A new case, a smaller team, and 1.5 million documents. We couldn’t do it the old way. We needed leverage, so we built it.

Before “Legal AI” was a buzzword, we worked with a vendor to create our own model. I manually reviewed ~30,000 documents to build a foundational schema: what each document meant and how it fit into the larger legal narrative. We then used that schema to infer relevance across the full 1.5 million documents.

Key documents surfaced instantly.

That was my first real proof that structure beats brute force. It was the proof that Knowledge Graphs matter—that the right tags, fields, and relationships can transform chaos into clarity.

The Path to Marqui

Only years later, when the OpenAI API first launched, did I realize how much those long hours in the document review rooms had shaped my worldview.

It is that same guiding principle—that structure creates cognitive leverage—that drives how we are building Marqui today. We are building a tool that gives anyone that kind of analytical power, but through a simple, intuitive, and hopefully fun experience.

Marqui is a knowledge engine hiding under a playful UI. It’s a personal website that doesn’t just sit there; it actually does things for you.

We’re starting to onboard our first users soon. I’m excited to see how it can work for you.