About

Hi, I’m Fabio! 👋

I’m a software engineer based in Brazil. I’ve been hacking on developer environment tooling since 2013 - first to make my own setup less painful, then as open source, and most recently as the kind of environment a coding agent can work inside.

What I’m Working On

Most of my recent writing these days are about AI and one series about Modernizing my Terminal-Based Development Environment, which walks through how my dev stack looks like these days in the modern agentic world. The standalone posts orbit the same neighborhood - things like giving agents a mailbox so they can work async, or arguing that agent commits deserve better than Co-Authored-By. A lot of my recent posts mention AI and I’m getting tired of the discourse like many of you, the work itself is still interesting though.

I also recently split off a /dev/random lane for chaotic, lower-bar content: scattered notes, link dumps, and AI-collaborated pieces I wouldn’t publish on /blog. More on the why in Embracing Chaos in the Age of AI.

Believe it or not, my day job is (and has been) Ruby on Rails for like 15 years with a 1-2 year shift towards Golang / Kubernetes and back along the way.

How I Got Here

I started writing code professionally around 2007 and worked through PHP, .NET, some Go, and (yes) some VB6 before settling into Rails. The dev-env tooling started as a side project around 2013 and has come back in waves ever since: vagrant-lxc, ventriloquist, and devstep were all different shots at the same target - git clone && cd into a project and start working without thinking about infrastructure. Those are archived now. My current take is crib, a small runner on top of the devcontainer spec - same problem, three iterations later, this time built on a spec someone else maintains.

The other thing I keep getting pulled back to is using tech for civic stuff - back in 2017 I used serverless to help Operação Serenata de Amor flag suspicious congressional reimbursements, and during the pandemic I built covid19br.pub to monitor official COVID-19 publications in Brazil. Haven’t touched that space since, but honestly, with how much easier it’s gotten to hack on things with LLMs in the loop, I probably should.

I started this blog in 2013, then “life happened” and I took an 8-year break before coming back in 2025. In case you find a broken link on the internet to this site, in 2022 I rebuilt it on Hugo and some content was deleted as part of that transition.

🤖 AI Use

I use AI as a coding and writing assistant, and I want to be transparent about how.

For code, it’s a pair programmer - sometimes driving while I review, sometimes the other way around. I wrote about that evolution in Skip the ‘No AI Days’ and an earlier post on resistance. Architectural decisions and quality standards are mine.

For writing on /blog, AI is a copywriter and editor, not a ghostwriter. I brain-dump, it helps me structure, I refine. The ideas and opinions are always mine. I’m still learning the craft - some slop has made it through and I fix it as I go. AI amplifies, it doesn’t replace.

/dev/random plays by different rules. It’s where chat summaries, AI-generated digests, and other AI-collaborated pieces go - with a 🤖 banner declaring the model and role, and a “clickwall” you have to accept before the AI portion of a post is revealed. Pure-human chaos in that section gets a ⚠️ HUMAN DISCLAIMER banner instead. I unpack the segmentation and the polish-as-signal thinking in Embracing Chaos in the Age of AI.

I also adopted dweekly’s AI Content Disclosure proposal for machine-readable signals: every AI-tagged post emits <meta name="ai-disclosure"> at the page level and ai-disclosure="ai-generated" on the gated sections, with model + provider attribution.

Get in Touch

Questions, collaboration, or just want to nerd out about tech in general - reach out.

Thanks for reading! 🚀