Blog

It's minimal, but I'm posting things.

Articles

  1. Retour d'expérience de la VAE pour un Bachelor ...
  2. The simplest morning routine which gives me the...
  3. My Taskwarrior Workflow
  4. Tiling managers for KDE Plasma on Ubuntu
  5. [Vim] You should use commit hashes in your vim-...

Code snippets

  1. Combining the asdf version manager with the Jav...
  2. Automatically snoozing all incoming emails unti...
  3. Why Google Antigravity beats Claude's subscript...
  4. LLM auto-complete plugin for Vim for OpenAI, Mi...
  5. [KDE] CTRL+T Shortcut: Focus or Launch Konsole ...
  6. [Vim] Proper code folding for Python files
  7. [Vim] How to compile Vim
  8. Researching the best small VPS for under 5 euro...
  9. [Vim] Fuzzy finding a string within a project w...
  10. [Vim] Prank your coworker by forcing them to us...
  11. A monadic Result type using generics in Python
  12. Interacting with Bitwarden in your terminal
  13. [Vim] Disabling typehint diagnostics for the `p...
  14. Free git repository with actual large file stor...
  15. Script utilitaire pour générer automatiquement ...
  16. [Vim] Navigating between Tmux panes using Vim m...
  17. Autogenerating an OpenAPI json file using Maven

Thoughts

mynoise.net rivals brain.fm and is completely free :)

brain.fm fixes my need for some sort of background noise when I want to work. Reduces the time to find the perfect fit to near zero, and in that process avoids countless potential for distractions.

It's good to use AI models which use something called "input/output caching". Cheaper models without caching may cost you more than an expensive model with caching, once you hit a certain amount of requests.

I find it incredible how LLMs default to object-oriented anti-patterns again and again... isn't object oriented development the most popular paradigm? Why don't they use strong design patterns?

Rebinding your CAPS LOCK key to CTRL is an insane upgrade in your keyboard ergonomics.

LLM agents benefit from iterative improvements. Use an LLM to build a harness for automatic agent enhancement. https://github.com/microsoft/agent-lightning

I imagine some people are now declaring 10x devs as "the norm", and are aiming for 100x dev status /s

Developing user interfaces has never been easier ; it used to be so tedious ; thanks AI : )

One of the greatest investments in your productivity is learning (and customizing shortcuts) for a terminal de-multiplexer such as TMUX.

An excellent OpenCode plugin which gives coding agents persistent memory; pretty much plug and play https://github.com/tickernelz/opencode-mem