"If these tools are so good, where's all the magic software?" I see more bugs now, in software that was rock solid before. The uncomfortable question: when does that hit something load-bearing? Has it already?
Nicolas Dular
@nicolasdular.com
Staff Engineer @gitlab.com Web: https://nicolasdular.com/ Podcast https://walkingdevs.com/
📭 New Blog Post 📬 I wrote a tiny piece about how to store your SSH keys more securely. In times of distribution attacks like Shai-Hulud, it's easy to install malware on your computer that extracts your private SSH keys. Don't let them. peterullrich.com/store-ssh-ke...
Store your SSH keys in a Password Manager
Your SSH keys are sitting in ~/.ssh as plain text exactly where malware looks first. Move them to a password manager instead.
peterullrich.com
I've decided to leave GitLab. What a journey! nicolasdular.com/blog/2026/06...
Leaving GitLab
Leaving GitLab - Blog post by Nicolas Dular
nicolasdular.com
I like how @mariozechner.at and @mitsuhiko.at are at the forefront of using AI & at the same time not bullshitting 🙌 Worth listening.
OpenClaw was built on top of Pi, a minimalist, self-modifying agent. I sat down with Pi's creator, Mario Zechner, and longtime Pi user Armin Ronacher to talk on the "why" of Pi, their grounded takes on building with AI, why we should probably slow down, as an industry, and more. (cont'd)
"Everyone has a plan until code meets production" - Mike Tyson
Why does every o11y platform has a look and feel like it was made in the early 2010 era 🙈
Data migration at work: 3-6 month project. Data migration at sideproject: "AI Code LGTM. Push!"
Third episode of our podcast is out. We talk about how we end up working more with agents, and what it does to our brain. And of course, complain about the missing sun in Vienna ☀. walkingdevs.com/002-agent-fa...
002: Agent Fatigue - Walking Devs
In this episode, we cover the impact of AI on the developer community, work-life balance and AI burnout, concerns about job security, exhaustion and context switching in development, and touch on the ...
walkingdevs.com
Anyone experience with Rustler/NIF not free-ing up memory ? Is there any way around it or is the solution just to use Ports? #ElixirLang
An official VSCode extension for Expert LS has been released. You can be a 30th person to download it. marketplace.visualstudio.com/items?itemNa... #elixir
Expert LSP - Visual Studio Marketplace
Extension for Visual Studio Code - Elixir language support for Visual Studio Code
marketplace.visualstudio.com
Developer UX is superb compared to infra. Why can't we have nice things there?
I am bullish on Elixir elixir-lang.org/blog/2026/01...
Type inference of all constructs and the next 15 months
Today we celebrate 15 years since Elixir’s first commit! To mark the occasion, we are glad to announce the first release candidate for Elixir v1.20, which performs type inference of all language const...
elixir-lang.org
While we like to complain about the Frontend ecosystem constantly changing. At least they have one language. Backend has the problem of language diversity.
When the multiple you made on $NVDA is only the second best investment. #RAMprices
The product of my dreams: lutefish.com But even under best conditions, the physical latency barrier is just too high :( I guess I'll wait for quantum entanglement.
Play Music Together, Miles Apart
Lutefish makes it dead simple to rehearse with real musicians in real time—without leaving your space or lowering your standards. It's super low-latency hardware pairs with an online platform for fast...
lutefish.com
I spent a good chunk of 2025 building side projects no one asked for. And I loved it! 🦥 nicolasdular.com/blog/2025/11...
Side Projects in 2025
Side Projects in 2025 - Blog post by Nicolas Dular
nicolasdular.com
The job of a programmer is to manage exactly two things: uncertainty and complexity. Code is sometimes an asset and sometimes a liability, but itself has no meaning outside the context of those two factors.
So far I ignored all these clickbait "prompt engineering" guides. Works grat. Telling the LLM what to do is all you need.
Months of work went into this and I finally saw it live in production. From queries that used to take 15+ seconds (and timed out) to just ~10-30ms 🚀 More details in the design doc: handbook.gitlab.com/handbook/eng...
Traversal IDs on issues
Business Objectives To further roll out work items, we need to tackle the scaling problems that are inherent to our hierarchy-based approach to querying data. Without addressing these performance prob...
handbook.gitlab.com
Now that Strava has launched a similar feature, I'm ready to show you the project I've been working on this year: easyhard.carbonative.com It will help you to obey the 80/20 rule in running: 80% of your runs should be easy, 20% hard.
Easy Hard | Login
Track your running intensity and master the 80/20 rule. Train smarter with data-driven insights into your easy vs. hard efforts.
easyhard.carbonative.com
I am building a small application with #ElixirLang and Codex. Process is pretty much: - Let Codex run with my prompt - Look at generated code - Delete 50% of the code It's okay-ish at writing tests, which is nice. Especially since IMO Elixir tests are too verbose compared to RSpec.
Looked more into @ash-hq.org now. A few thoughts - Not completely sold on the declarative style. - It's fast to prototype stuff. - LLMs are doing surprisingly ok. - Having types end to end is a game changer.
I looked into Node/Typescript ecosystem for a Fullstack framework last week. Quality on the backend is just too far off from what I see in Rails/Phoenix. My biggest requirement is: full typing between backend to frontend. Now looking into @ash-hq.org 👀
I am looking into the NodeJS ecosystem. A bit shocked about the state of ORMs. Maybe it's what I needed to appreciate ActiveRecord 😂
After being a drummer all my life, it still blows my mind that I can take my Trombone WITH ME. Disassemble. Put it in the case. Go Just like that 🤯
What are good open source Elixir/Phoenix codebases to look into? I already looked into @changelog.com and like that they're cutting out context completely.
"What do you hate about ActiveRecord?" That interview would definitely go over time.
With LLMs making leetcode and other whiteboard problems trivial, software hiring IMO should shift towards establishing taste, which LLMs absolutely do not have. "What do you hate about ActiveRecord? What would you change about Rails if you could?" Review this PR, etc.