THOUGHTS

My old tablet was getting painfully slow. Before buying a replacement, I gave LineageOS another try, this time with help from a ChatGPT Work agent.

How much information does JEV need to make a useful decision, and when is a custom classifier worth training? An exploration of context, labels, and evaluation.

This experiment started with an uncomfortable question: if models, APIs, permissions, and interfaces describe the same business, why does none of them understand the whole system?

The agents could already code, test, and review. What they lacked was organization. Once roles, handoffs, evidence, and blockers persisted beyond individual sessions, my daily became a short review of the decisions that actually needed attention.

While building a coordination system for agents, I found a broader question: what if agent systems borrowed proven roles, governance, and institutional memory instead of inventing an organization from scratch?

Linux already ran on almost anything. The hard part was handling the exceptions. Omarchy, agents, and skills are making those exceptions cheaper to diagnose.

From Code to Legislation · Part 3 of 3

A technical design for transforming legislative Markdown into a structured, versionable representation: contextual lexing, an Abstract Law Tree, projection to Akoma Ntoso, and consolidation operations.

From Code to Legislation · Part 2 of 3

The parser could already build an Abstract Law Tree. The real challenge came next: partial amendments, identity, temporality, and consolidation. Research changed once the problem found its vocabulary.

From Code to Legislation · Part 1 of 3

A compiler course and a small Spanish-language educational programming language seemed unrelated to law. Years later, they provided the mental model needed to start interpreting legislation.

Vibe coding is everywhere lately—mostly as a meme or criticism aimed at people trying to build software without really knowing how to code. But… what happens when it's used by an experienced engineer solving a real problem?

As a back-end developer I live in the terminal. This post covers the setup that finally gave me a fast, distraction-free environment — Vim, tmux, zsh, wezterm, and a handful of tools that made it all click.

Building Codexivo · Part 2 of 2

Part two of my programming language series. Tokens are the smallest meaningful units in source code — I cover what they are, how they're classified, and how I defined the full token set for Codexivo.

Over a year of using GitHub Copilot has shown me it's far more than a code autocomplete tool. In this post, I share the most useful ways I've found to leverage it beyond just writing code — from docs to testing to translations.

Building Codexivo · Part 1 of 2

I decided to build a simple programming language in Spanish to make learning to code more accessible. This post covers why I started, what goals I set, and how I approached the architecture of an interpreter from scratch.

A practical walkthrough for building microservices in NestJS using gRPC and RabbitMQ — from defining protobuf services to wiring up the client and message queue in your application.