# Retrospectiva #1 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 box, what’s the point? I couldn’t disagree more. Creativity, personality, and *good taste* have never been so important. If anything, I’m planning to write more. Winter has arrived in Denmark, and the sun is setting at around 4PM. It’s what we call the [hygge](https://www.oed.com/dictionary/hygge_n) period. It’s dark, cold, and rainy most of the time - and so we make the most out of staying inside. Using ----- The endless chase for the perfect browser continues. I’ve been a long time Firefox user, until things started breaking, and I moved to Chrome. I hate Chrome. Don’t get me wrong - Chrome works 99% of the time - but I like taking Google products with a grain of salt. [Helium](https://helium.computer/) has been a breath of fresh air. The guts are the same as Chrome, but it’s de-googlified and completely [open-source](https://github.com/imputnet/helium). One Chrome feature I rely on is the built-in webpage translation. Chrome’s Google Translate’s integration does this better than anything else. But [this little alternative](https://github.com/FilipePS/Traduzir-paginas-web) has worked great for me. It’s a fruitful period for tech! In the AI-tool space there’s [Claude Code](https://www.claude.com/product/claude-code), [Codex](https://developers.openai.com/codex/cli/), [Cursor](https://cursor.com/blog/2-0), [Cline](https://cline.bot/), [Copilot](https://github.com/features/copilot/cli), and probably 54 others. We are the ones that get to benefit from every company fighting to get us to use their tools. When I’m asked about what I’m using, my answer has consistently been: “It changes every week”. Recently I’ve gone back to what I love: [Neovim](https://neovim.io/), [tmux](https://github.com/tmux/tmux/wiki), and a couple of CLI based agents. To pull everything together [Sidekick](https://github.com/folke/sidekick.nvim) has been great. Reading ------- For most of 2025, I’ve switched between [digital](https://duarteocarmo.com/blog/goodbye-kindle-i-dont-think-ill-miss-you) and physical books. But living in a small Danish apartment means I don’t have a bedside light when reading at night. Vitto got me this little [20 USD light](https://www.igritin.com/products/gritin-9-led-rechargeable-book-light-for-reading-in-bed) and I’ve been back on the physical book train. Two highlights. The first is [Software Engineering after the Vibe Shift](https://www.lulu.com/shop/sean-goedecke/software-engineering-after-the-vibe-shift/paperback/product-jew4ver.html?q=Sean+Goedecke&page=1&pageSize=4) by Sean Goedecke. It’s a short but insightful book. Even though I don’t consider myself a Software Engineer (I’m more on the Machine Learning side of things) a considerable amount of ideas hit home: > «*The best way to anticipate problems is to deploy early. In general, a helpful question to ask is **can I ship this right now?** Not this week, not today: right this second. If not, what would have to change for me to be able to ship something?»* I’m finishing up [Machine Learning System Design](https://www.manning.com/books/machine-learning-system-design) by Valerii Babushkin and Arseny Kravchenko. It’s a great read so far, with lots of interesting fireside stories about building machine learning systems at scale. It’s funny how a simple binary classification can get complicated when you are building at scale. I need to get back to reading some fiction. Listening --------- Vitto asked me to move off Spotify with her. This is not a political blog - she has her reasons. So I took the plunge. After all, it’s fairly easy to move [my playlists around](https://support.apple.com/en-us/118249). Overall: mixed bag. I could get around all the Google Home limitations by running [Airconnect](https://github.com/philippe44/AirConnect) on my Raspberry Pi, and I understand the focus of Apple Music vs. Spotify is much about quality vs. quantity. But I’m not sure I’m sold. The Apple Music app on the Mac is buggy, and the playlist diversity is just not there. I want “Chill Italian Rap” - Spotify has at least 45 playlist results - how come Apple has none? And don’t get me started [on this](https://discussions.apple.com/thread/255774369?sortBy=rank). For podcasts, I have two recommendations. [Karpathy's interview with Dwarkesh](https://www.dwarkesh.com/p/andrej-karpathy) is a must-listen. I found myself smiling regularly throughout this one. Another great listen was [Armin Ronacher’s interview](https://newsletter.pragmaticengineer.com/p/python-go-rust-typescript-and-ai) on the Pragmatic Engineer. Watching -------- [This video from Hank Green about the state of the AI industry](https://www.youtube.com/watch?v=Q0TpWitfxPk) is a very interesting watch (and commentary on [this article](https://archive.vn/V0dWR)). Sure, he’s funny and entertaining, [but I also agree on a lot](https://www.publico.pt/2025/10/08/opiniao/opiniao/ia-bolha-utilidade-2150129). But not everything is about ML and AI - and since it's winter - we also took some time to get some new blankets and fire up the old projector. HBO's [Trial of Karen Read](https://www.rottentomatoes.com/tv/a_body_in_the_snow_the_trial_of_karen_read/s01) wasn't the most *intellectual* watch but it was certainly entertaining. Recommend that one for the “binge in a day” category.