Andrés Frei
Buenos AiresFounder of NeuralMind● programming since 1995

I create products and test them before I believe in them.

I wrote my first program at 15. Today I design and build products with AI, contribute to open source, and keep doing what I did in 1995: wondering how things work.

01

About

the short version

I like to understand things. Technology, ideas, why one thing works and another doesn't.

I started on a computer I was given, running DOS and Windows 3.0. Since then I've built systems that weren't allowed to fail, spent some time away from the industry, and never stopped programming. Now I've founded NeuralMind to create our own products with AI.

I read philosophy because it teaches me to ask better questions, and I build products because it is the best way I know to put them to the test.

  • Technologythe tool
  • Knowledgethe raw material
  • Critical thinkingthe filter
  • Philosophythe questions
Avatar of Andrés Frei: an illustrated portrait in the style of Nietzsche, with a mustache and glasses with colored lenses
02

Creating

three fronts, one method

What I'm creating today.

02.1My products

what I start and build
founder● active

NeuralMind

Software that thinks. Products that get used. We design and build our own products and bring AI into how companies actually run. Each product starts as a hypothesis about a concrete problem.

02.2In production

AI on systems that must not fail
technical lead · since 2023● in production

Modernize without rewriting.

hypothesis
You don't replace a legacy system: you understand it. AI can explain code nobody wants to touch.
test
I lead a platform of agents that read, document, and explain a COBOL system in daily use.
result
The team asks and the system answers. The code stays in production, now with something that explains it.
~3,000COBOL programs0rewrites
no names: it's a company's system, not a product of my own

02.3In the community

open source
creator · v3.5.0 · MITpublished

Playbook

The method I build everything with: the human decides, the agent executes, and every decision stays in git. Plugins for Claude Code and Codex CLI.

claude plugin marketplace add andresfrei/playbook
playbook.neuralmind.lat ↗
designs and builds its new interface (in progress) · Rust · GTK4in progress

ZapTide

Native WhatsApp for Linux. I design and build its new interface.

github.com/luminusOS/zaptide ↗
reports and discusses improvements to the skills and review flowactive

gentle-ai

Sets up the coding agents you already use. I report and discuss improvements to the skills and review flow.

github.com/Gentleman-Programming/gentle-ai ↗
03

Log

1995 → today

Programming since 1995.

2026

I start NeuralMind and publish Playbook.

I move from keeping systems running to creating my own products, and I open-source the method.

learned → the method is worth more if others can use it
2023

I use AI to modernize about 3,000 COBOL programs still in production.

Agents that read, document, and explain code nobody wanted to touch. I'm still in charge of that work.

learned → AI helps when someone decides what to ask it
2019

I take over IT at a company.

Infrastructure, security, and the migration of its ERP.

learned → the critical part isn't the technology; it's continuity
2009–2017

A while away from the industry. Never away from code.

I kept programming for fun, learning and experimenting.

learned → I program because I like it, not just because it's my job
2004

My first system in production: payroll.

More than 20 years later, it still works.

learned → simple things last
1995

My first program, at 15.

On a computer I was given, running DOS and Windows 3.0.

learned → curiosity is the best language
04

How I think

until proven otherwise

What I stand by. And what I discarded.

  1. The agent executes. The person decides.An agent proposes, builds, and shows the evidence. Choosing what to do is still up to a person.
    held
  2. If it isn't written down, it isn't decided.A conversation gets forgotten. A repository doesn't.
    held
  3. Simple first.The simplest solution that solves today's problem and can grow tomorrow.
    held
05

Contact

let's discuss a hypothesis

exp-000 · this site

This site is also an experiment.

Hypothesis: a personal brand can be defined like a product, with every decision written down before any code. Result: 65 decisions in a day, a dozen styles discarded and one chosen.

● in progress