July was busy — like always — but we made sure to take some time and touch some grass. Or mountains in our case. With the high temperatures across Europe we went to the Trentino region and hiked around the Dolomites for five days. Cool temperatures and beautiful scenery. Vitto's idea of …
For the past year, I've been very curious about the intersection of Large Language Models (LLMs) and the European Portuguese language. I've built evaluations , classifiers , and eventually the largest pretraining dataset in European Portuguese . Now, that's all fine and dandy, but it raises the obvious question: Can we pretrain an …
This month's Retrospectiva comes a little later than usual. The whole house got sick battling the flu the past couple weeks. Mixed with a small heatwave here in Copenhagen, means summer hasn't blessed us like we expected. At least not yet. Now that I'm feeling a bit better, I finally …
In January this year, OpenClaw went viral. You probably heard about it. My guess is some of you even use it - or at least tried to. I certainly did. I remember it clearly. I installed it, and very quickly noticed I was now spending time debugging a cron job created …
If there's one thing Retrospectiva does for me - other than giving me an excuse to write for you of course - it's making me realise how fast time moves. It's June already - and May was – well – hectic to say the least. Data Makers Fest was awesome - and I absolutely love Porto …
About 6 months ago I wrote a small rant on how open source TTS models still sucked . 6 months later, I'm happy to report that isn't the case anymore. January this year, Qwen , the famous Chinese AI lab, released Qwen3-TTS , an open-weights series of TTS models. The release included 2 …
Educational score over time vs. document count for Bagaco v2 A couple of months ago I released Bagaço - a pretraining dataset for European Portuguese. The idea was simple: take the FineWeb 2 dataset, limit it to web pages that look like they came from Portugal, and classify them into categories …
My Hermes agent – Saramago – giving me coaching feedback and looking for the ideal flights for an upcoming trip. Another month bites the dust. We've been back in Denmark for the past couple of weeks. I often joke that these are the only months in the year where Copenhagen is actually …
Update July 2026: The AMÁLIA models have now been released! You can find the weights/datasets on Hugging Face . Some parts of this blog post might be out of date. So beware! In December 2024, the Portuguese government announced AMÁLIA: a 5.5 Million Euro investment on a large-scale LLM …
Well, now would you look at that. Article 100 on this website. Ten years of writing on this small corner of the web. I'm writing this month's newsletter from my favourite place: the airplane. Happy Easter if you celebrate. With clients spread across time zones, I didn't really get the …
As you get older, life becomes complicated. Not in a bad way. There’s just more going on. We don’t all want to run marathons. Maybe you want to run a parkrun . Maybe you want to gradually increase your volume. Maybe you don't want to run at all. Whatever …
Given some text in Portuguese, how easy is it to determine if it's from Brazil or Portugal 1 ? For native speakers, this is pretty easy – it's almost a feeling. But for machines: not so much. This might seem like a useless problem at first. But in the age of language …
Happy February! Just like that, we are close to wrapping up the first quarter of 2026. After quite some time roaming around, we finally flew back home to Copenhagen. It's cold, windy, and grey, but it's also calm, organized, and cozy. Most of all: it's home . Another good thing about …
Let's say your goal is to train a Large Language Model only on European Portuguese. Where do you start? What datasets are out there? What websites are being scraped for the large black box? Bagaço - named after the popular Portuguese moonshine - is a small step in that direction. In June …
Well—that was quick. Just like that, the first month of the year is gone. Winter blues would normally peak around this time. But not this year. We're spending time with family in my favorite place on Earth, somewhere along the Adriatic coast of Italy, in the Marche region. It's …
Let's face it. It might just be the year of agents. If you work in tech and your workflow hasn't changed in the last year or so - you're probably doing something wrong. For those of you who have. It's fun. We're building more than ever before! But there's a dark …
I get up feeling sleepy. I lace up my running shoes and head out the door. I fire up my favourite podcast. "This show is brought to you by..." I hate ads. I understand the attention economy. Companies are capitalizing more and more on everyone's time. I have nothing against …
And just like that, it's the end of the year again. We did not expect to make it to the South for Christmas this year. Getting a passport for a newborn is a painful, bureaucracy-filled process, especially for a Portuguese-Italian baby born in Denmark. But somehow—magically—the Italian embassy …
Benchmarking calorie prediction for Taralli There's something very funny about the current Machine Learning and AI landscape. If you're in the field you probably heard about it. "Vibes" they call it. When someone wants to test something out, they conduct a "vibe test". I call bullshit. How are you supposed …
Big news in the state of Denmark. And no, nothing's rotten . Allegra just came into the world. As she takes a nap, I take the opportunity to write November's Retrospectiva update. The most relevant thing this month is probably the release of my recent book about DeepSeek . It's nice to …
Back in May this year, my longtime friend Alex reached out and asked me if I wanted to collaborate with him on a book about DeepSeek. I would love to tell you the story of how I thought long and hard before getting back to him. I didn't. I just …
or at least the open versions of it. I have this very stupid rule. A couple of years ago I decided to turn this blog into a podcast . At the time, I decided to make up a stupid rule: whatever model I use to clone my voice and generate article …
Welcome to Retrospectiva. Retrospectiva is a monthly update about what I’ve been up to. In the age of LLMs, I’ve heard many argue that it’s hard (and useless) to write anything at all anymore. When anyone can prompt a model and get some text from the magic …
Why do this? ML and AI are moving at an incredible pace. The amount of research coming out vastly surpasses anyone's ability to interiorize it. By interiorize I mean study, experiment, or even just test it out. However, the importance of learning has never been greater. Even if we don't …
The world moves fast, faster every day. For those who work with technology - and even those who don't - it's hard to keep up with the news. But I've always enjoyed staying up to date with what is happening, and my main tool to get it done hasn't changed in years …
It's that time of year again. As usual, we took a couple of weeks off and came south: a bit of Portugal, a bit of Italy, a lot of friends and family. My family has a long-running joke that I hate people and love my computer. That's not (entirely) true …
A couple of weeks ago in Lisbon, I went to a friend's birthday dinner. In front of …
..but they can also be a lot of fun. If you work in tech, I'd say there's a 98% chance you've heard about it. MCPs are the future of agents, MCPs will be everywhere, MCPs are the future. The Model Context Protocol, first introduced by Anthropic is blowing up. For …
Last month, I built an app called Taralli . It was fun to close the loop and get it out there. Still - there was an elephant in the room: "...[calorie tracking] Accuracy isn’t great, and it makes some pretty basic mistakes. I’ve got plans to fix that..." And things …
For years, I never really cared about what I ate, how much I ate, or when I ate it. But sometime late last year, I finally decided to listen to Vitto. I started noticing that what I eat actually had an impact on how much I ran, how well I …
Even though it never really replaced physical books for me, I've been a big Kindle user for many years. The ability to take hundreds of books with me anywhere is priceless. Vitto even has a running joke that I tend to lose them on the plane. I love my Kindle …
Last year I made an effort to speak less and learn more. However, I still had the opportunity to present at a couple of conferences. One of them was PyCon Wroclaw . The main goal was to talk about a couple of interesting paradigms I've come across while using LLMs to …
Last week over lunch, Pedro was telling me how he uses Perplexity to improve the way he searches around the web. He mentioned we're entering a phase where we have at least 20 'AI' tools to boost our productivity - and the tough thing now is choosing which ones to use …
In the period of the generative AI boom, everything is happening all at once. Every day a new model comes out, every week a billion dollar company teases a great new advancement. More often than not, they'll show us some fancy version of this: We all love a misterious chart …
Notebook download - nbsanity preview Some context We had a database of 50M strings and I couldn't wait to embed them all. Embeddings for recommender systems were, for a long time , the HOLY grail I longed for. I still remember, long before the RAG rage, having conversations with Pedro about the …
I'm flying over the Atlantic just between France and Spain for my yearly Christmas trip back to Lisbon. For some reason, I just remembered the blog post I always wanted to write. This one is not about tech, but it's about something I'm equally passionate about: The Marathon. Funnily - I'm …
For the past year , this website's podcast companion has been running on a text-to-speech model called XTTS-v2 . It's not horrible . And those who have heard my voice before might notice some similarities. But it's far from a pleasant listenning experience. But the world of text-to-speech (TTS) has been gradually moving …
For the past years Vitto and I have used Tricount to track our shared expenses. The app is actually pretty good, but there’s one small thing that annoys me quite a bit. Even though we spoke English to each other for the first month, we’ve since spoken a …
For a short summer break, Vitto challenged me to visit Iceland with her. As someone that is used to going south for the summer - this sounded stupid at first, but I was incredibly surprised. For 6 days, we were pretty much in another planet. Here are some notes and tips …
Tanda Runner is a web app that shows me the things I care most about when preparing my next marathon. I've also added a running coach/agent designed to give me actionable feedback about my training. Some of that feedback is probably hallucinatory - I'll get to it in a bit …
I'm lucky enough to work with some pretty talented folks. During a recent offsite, one of them gave a completely improvised talk. He had just gotten back home from holidays - so naturally, they didn't really have time to prepare anything. But what they said resonated. It was about improvisation . How …
I'm a sucker for side projects. There's nothing quite like building something to learn about it. Over the last 5+ years I've accumulated a little over 15 small web apps and websites. Almost all of them are hosted on a small Hetzner server and deployed using a mix of ssh …
There I was. Ready to click the big blue purchase button. 350 Euro , for a keyboard (!) Asking Vitto multiple times: is this really worth it? Should I go for it? I mean, that's a whole lot of dinners. After much hesitation, I did it. All black , with blank key caps …
Conversas de café . Literally translated, means "coffee shop conversations". An upcoming election, the weather next week, a corruption scandal, a new policy. Just some examples of Conversas de café . As an expat living in Denmark for the past 7 years (without speaking the language), that's perhaps one of the things …
Step-by-step instructions to do it yourself If you ever read this blog, you probably already know by now. For the past couple of months, I've been trying to create my very own Large Language Model. OpenAI's models have had a significant impact on my productivity. But something still bothers me …
ChatGPT is now part of our daily lives. A quick question, an extra input, some quick feedback, I always reach for it. AGI or not AGI, I certainly can't deny the impact it has had on our lives. It's pretty incredible! But that small voice just won't go away. "Where …