If you follow technology and development news, you’ve already noticed that Artificial Intelligence (AI) dominates the discussions. Far beyond the launch of new tools and workflows, there is a critical debate about the practical impacts of the mass adoption of these technologies in software engineering.

Observing this scenario, a fascinating — and challenging — trend began to emerge.

The impact of AI and the avalanche of code

Over the past year, Amazon faced several outages in its services, affecting both e-commerce and AWS infrastructure. The company is one of the biggest promoters of using Artificial Intelligence tools by its developers. All indications are that these blackouts were a direct consequence of the increasing use of AI in code generation, to the point of requiring a limitation on its direct incorporation into production. Today, a senior engineer needs to rigorously evaluate the material before its implementation.

Recently, I bought a second pen for my iPad with the goal of having an immediate backup: if one’s battery runs out during use, just grab the other while the first one charges. The new choice was the ESR 6C006 Geo Digital Pencil, a model with excellent build quality, USB-C charging, and native integration with Apple’s Find My network.

Everything seemed perfect, until I ran into a frustrating detail: this pen does not automatically shut off after a period of inactivity (auto shutoff or sleep mode). If you don’t remember to turn it off manually by holding the top button for two seconds, it will continue to drain the battery.

On May 6, 2021, the Supreme Federal Court declared the sole paragraph of article 40 of the Industrial Property Law (Law 9,279/1996) unconstitutional. The decision ended a mechanism that, in practice, automatically extended pharmaceutical patents beyond the 20 years provided for in international agreements — and paved the way for the entry of generics into dozens of markets.

What the Overturned Provision Stated

The sole paragraph of article 40 guaranteed a minimum term of 10 years of validity from the date of the invention patent’s grant (7 years for utility models). On the surface, it seemed like just protection for the patent holder. In practice, it was an automatic extension mechanism: since the INPI took, on average, more than a decade to examine applications, a patent filed in 1997 and granted in 2011 would be valid until 2021 — not until 2017, as stipulated by the 20-year rule from the filing date.

Evolutionary biologist Richard Dawkins recently wrote an article titled “When Dawkins met Claude”, where he describes his experience after two days of intense conversations with the artificial intelligence Claude on various topics.

Throughout the text, Dawkins gives examples of interactions he had with the machine. However, what really sparked discussion about the article was his main provocation: From what point will we consider something to be conscious?

The Imitation Game: The Turing Test in the Age of AI

To substantiate the need for this question, Dawkins takes a historical journey that goes back to Alan Turing and his proposition of the “Imitation Game” (today known as the Turing Test). In the 1950s, the possibility of a machine convincing a human being that it is conscious seemed almost nil. However, here we are debating exactly that.

I got the idea for this text after learning from the channel Fireship about the latest exploit in the Linux kernel that was related to a 2017 commit. And last week, I was discussing with friends who work with data science why I prefer to stay away from Linux today.

Alright, but who am I to talk about this topic? I’ve been using macOS since mid-2007, without dual-booting with Windows until mid-2010. After that period, I primarily used Windows to run Microsoft Flight Simulator X, until I eventually returned to exclusively using macOS until mid-2018.

Until recently, I had no intention of creating a newsletter, mainly due to the additional maintenance work. But, after exploring a few alternatives, I chose Beehiiv for its ease of use and excellent scalability. However, the challenge arises when you already have pre-existing web infrastructure and want to achieve seamless integration, without relying on standardized iframes.

Beehiiv is a fantastic platform for newsletter management and also allows you to create your own website through them. However, if you already have an established website, you will need to make modifications to your DNS records and source code to implement this integration. Depending on your project’s architecture, this can become somewhat complex.

On December 16, 2025, the 4th Panel of the Superior Court of Justice concluded the most anticipated judicial discussion in the Brazilian pharmaceutical sector. Unanimously, the STJ denied the patent term adjustment for semaglutide patents (REsp 2.240.025/DF), ending Novo Nordisk’s attempt to extend Ozempic’s protection by up to 12 years beyond the anticipated expiration date.

The patent expired on March 20, 2026. But the debate didn’t end — it migrated to Congress.

In January 2026, the INPI discontinued the “backlogometer” — the public dashboard that monitored the reduction of the backlog of pending patent applications — with the justification that the project achieved its objective. For the general universe of applications, the statement is technically correct. For pharmaceutical patents, the reality is quite different.

What was the Backlog Combat Plan

Launched in July 2019 by INPI/PR Resolutions No. 240 and 241, the plan started with a backlog of 149,912 pending applications filed until 2016. The central strategy was to use prior art searches already conducted by foreign offices (USPTO, EPO) to issue preliminary requirements (dispatch 6.21), forcing the applicant to respond within 90 days or have the application dismissed.

For years, I was trapped in a cycle of ambitious beginnings and disappointing endings. My GitHub repository had become a digital graveyard of half-completed projects—enthusiastic starts abandoned at the first sign of a significant challenge. But over the past few months, something remarkable has happened: I’m actually finishing what I start, and the catalyst wasn’t a new productivity system or time management technique. It was AI.

A Little Background

Before diving into my productivity transformation, it helps to understand where I’m coming from. I’m a chemist by training, working in a government position while also collaborating on research projects at the local university. I don’t have a formal IT or software development background, but I’ve been increasingly drawn to data science projects that intersect with my professional work.

Over the past few months, I’ve been on a journey to find the perfect code editor that balances excellent developer experience with the productivity boost of AI assistance. This exploration came after I had already spent almost two years using VS Code, which I adopted after ditching Emacs and Neovim in search of a more modern editing experience. My latest search led me through several modern editors, primarily those in the VSCodium family, with some interesting detours along the way. Here’s my experience and what I’ve learned.

I’ve been using GitButler for about two weeks now, and it’s already transformed how I interact with Git. As someone who regularly juggles multiple feature branches and context switches between tasks, I was looking for a tool that could simplify my Git workflow. GitButler caught my attention with its promise of virtual branches and simplified collaboration features.

My interest was further piqued after watching presentations by Scott Chacon, one of GitButler’s creators and also a co-founder of GitHub. His deep knowledge of Git is evident in talks like this one, and it shows in how GitButler has managed to create an almost frictionless Git experience.

Anthropic recently published a fascinating paper with an unusual title: “On the Biology of a Large Language Model.” But what does biology have to do with AI? As it turns out, quite a lot. The researchers are essentially performing a digital dissection of Claude’s “brain” to understand how it thinks—and finding surprising parallels to biological systems along the way.

In this post, I’ll break down this complex research into digestible parts, explaining how scientists are starting to understand what’s happening inside these AI systems when they answer our questions or write poetry.

The Brazilian National Institute of Industrial Property (INPI) is embarking on a significant transformation of its patent services. A recent presentation detailed plans to replace the current web search system with a comprehensive, unified platform designed to streamline the entire patent process. This initiative promises to enhance efficiency, reduce user errors, and improve the overall experience for patent applicants and stakeholders.

Current Challenges in the Brazilian Patent System

Currently, the INPI’s patent system relies on several disconnected systems. These fragmented systems handle various stages, including application submission, GRU generation (payment slips), process consultation, filing submissions and requirement compliance, and internal examination stages (formality, technical aspects, nullity, appeals, etc.). This lack of integration often leads to duplicated effort and user confusion due to unfamiliar interfaces, complex instructions, and technical jargon in official communications.

We’ve all been there: headphones on, music pumping, fingers flying across the keyboard, lost in the “flow” with your favorite AI agent. This, my friends, is vibe coding. It’s when you’re in the zone, seemingly effortlessly producing code.

It’s the idea that you can create great software simply by immersing yourself in the feeling of coding, trusting your intuition and riding the wave of inspiration. But lately, I’ve realized that for me, at least, the vibe is…off. It’s like I’m watching an AI agent code for me, occasionally chiming in with a suggestion or a correction, but often just observing the code unfold without fully grasping the intricacies of what’s happening can lead to wasting my time later.

So, I was watching this video the other day, and it got me thinking… you know, about how everything’s changing with AI. It seems like the old ways of doing things in software are just not cutting it anymore.

Remember how companies used to try and, like, trap you in their ecosystem? Make it impossible to leave? That’s software lock-in, and honestly, I think AI is making that a thing of the past. I could be wrong, but it doesn’t make sense anymore to pay for yearly plans anymore for most of the things that are appearing in AI. I mean, every week we have something new in the pipeline.

This is a quick fix to a problem I experienced with Open WebUI, which was that custom models were not showing in the models list. There was nothing in the documentation or in the issues pointing to a solution. I decided to open a new issue 4 days ago and today the user @dcolley gave the missing information and it worked.

For some reason, custom models are not working when you use Direct Connections. In my case, I use OpenRouter to use LLMs, so I thought the way to setup the API was from the Direct Connections menu. It turns out that you can setup the OpenRouter API from the Connections page, specifically in the OpenAI API section. And this is the solution, replicate your configuration for the API inside the OpenAI section.

A really simple script for those that, for some odd reason, decided to run the Open WebUI server from inside a LXC.

The problem

I really like the interface in Proxmox and i prefer it than dealing with Portainer for some services that are more accessed. This is the case for Open WebUI, which I use daily.

The problem I faced was the update process of the service, which I knew little about. So I need to understand how the service works under the hood. In Proxmox we can use the Proxmox VE Community Scripts which are really handy, but they don’t have much documentation on how the scripts work. Because of this, the user has to go to the service documentation and learn how it works. For Open WebUI, the documentation is directed to the installetions using Docker.

The problem

This week I have been searching for solutions to use domains with my applications running in containers in my MacBook Pro server. Yes, I have converted my MacBook Pro into a server, at least for now. The first solution was to use a Pi-Hole as a local DNS and it works fine, but the problem is that I have to specify ports to get into the applications and it gets annonying. The solution seemed simple enough, create a Nginx Proxy Manager (NPM) container, use it to point to the containers in the local network and this should solve the issue. Well, that’s what I thought.

The last months have been very exciting to me. The most exciting and difficult part by far is having my newborn son, whom is almost 3 months old. The other exciting part is that I finally understood that if I really like technology I should really invest in it.

Investing more in my formation in technology right now could come as odd for most people because I already have a Masters in chemistry and a specialization in intellectual property and also happen to work in the field that I have specialized in. The main problem for me is not that I don’t like what I do but that sensation that I feel about this not being scalable enough, mainly because of the difficulty that I encountered after first leaving the university and because after so much hard work on specializing not having enough return on the investment.