In June 2022 I pressed publish on the first of these notes. I was nervous about my English, unsure if anyone would read, and not sure I had anything to say that was not already said better by someone else. Four years later, here we are. This one is a retrospective of the notes themselves, the lessons that four years of writing in public taught me about technology, business, and a little bit about myself.

A small confession before I start. I recently used AI to go back over all of these posts. The goal was never to make them grammatically perfect, my accent is still in there and I like it that way. The goal was to make my thinking clearer for the people who read. Same ideas, same voice, just a little easier to follow. It felt right to admit that in a retrospective, on a site that spent four years writing about exactly these tools.

The hype waves come, and they go

When I started writing, Copilot was the new exciting thing and people argued if autocomplete on steroids would change anything. Then ChatGPT arrived and the argument changed shape. Then GPT-4, then RAG everywhere, then the vibe coding phase, then agents doing real work in real teams. Four years, maybe five or six distinct waves, each one announced as the end of programming as we know it.

My own tools changed almost as fast. I paid for Cursor and used it for a good while, then spent a long stretch on Augment Code, and right now, as I write this, my main tool is Claude Code on the Max 20x plan. By the time you read this it has probably changed again . The tool I use is the least stable thing about my work. The way I think about the work barely moved.

Writing through all of them gave me a perspective I would not have gotten otherwise. These notes are a record of my own reactions, and reading old ones is humbling. Some worries aged badly. Some excitement aged badly too. The waves were all real, I want to be fair, each one changed something. But none of them changed what I will say in the next paragraph.

The framing that survived every single wave was the business value framing. Every time I wrote a post asking “what does this cost, what does this save, what does this risk,” the post aged well. Every time I drifted toward “is this technology good or bad” in the abstract, the post aged like milk. Technology questions expire. Value questions do not. If I could send one sentence back to the 2022 version of me, it would be that one.

My little framework

Over the years, without planning it, the posts converged into a simple set of questions. I use them at work every week now, for any technical decision, any tool, any architecture, any hype.

  • What does it cost? Not just the price. The learning time, the migration, the attention it eats.
  • What does it save? Time, money, errors. And can I measure the saving or am I just hoping.
  • What does it risk? What breaks if this fails, who gets hurt, how do we walk it back.
  • Who owns it? Every system needs a human who answers for it. No owner, no adoption.

That is the whole framework. Four questions a smart twelve year old could ask. I watched these four questions kill bad projects, save good ones, and cut through more hype than any deep technical analysis I ever did. Writers like Martin Fowler taught me long ago that the hard part of software is rarely the software, and four years of writing these notes turned that lesson from a quote into a habit.

Writing in a second language

A personal lesson now. When I started, writing in English was work. I translated in my head from Portuguese, then cleaned the result, then doubted everything. Some posts took a whole weekend, and I still found embarrassing mistakes a month later.

It got easier. Not perfect, you can still hear my accent in my sentences, and I stopped fighting that. Somewhere in year two I realized the accent was not a bug. Readers wrote to me saying the simple direct style was why they read. I was trying to hide the very thing that made the writing mine. So if you are an immigrant thinking about writing in English, take this as your sign. Your simple sentences are not a weakness. Clarity has no accent.

Writing also made me better at my job in a way I did not expect. The discipline of explaining an idea to strangers is the same discipline of writing a good spec or a good design doc. These days, when agents turn clear writing directly into working software, that discipline literally became my profession. Writing here was training for a job that did not exist when I started it.

The posts that mattered were the personal ones

Here is the lesson that surprised me most. I sometimes spent days on careful technical posts, full of diagrams and trade offs. They did fine. But the posts that traveled, the ones that brought emails from strangers, were the personal ones. The post admitting that my identity was never in the code. The posts where I was a person first and a developer second.

I think I understand why now. Technical information is everywhere, and lately a machine can produce it on demand, decent quality, instant. What a machine cannot produce is one specific human saying “this happened to me, here is what it felt like, here is what I did.” Experience is the only content with no substitute. That changes what is worth writing, and honestly, what is worth reading.

Thank you

Four years, a few dozen notes, one career move along the way, one industry transformation lived in real time. Writing these notes gave me clearer thinking, some friendships that started in my inbox, and a written proof that my opinions can change, which I now consider a feature.

To everyone who read, shared a post, or sent a kind or angry email, thank you. Truly. Writing into the void is hard, and you made it not a void. Special thanks to the readers who corrected my English gently and my arguments firmly. That is the right order.

The notes continue. There is a lot to think about in the years coming, and thinking in public is still the best way I know to think at all. See you in the next one.

Pax et bonum.