Posts on Ineiti's thoughts

Projects

Fri, 04 Sep 2026 13:37:01 +0200

Now that LLMs start to write more and more code, we can finally concentrate on what matters: using our code, advertising it, networking with others… Here are some of my most recent projects (well, going back to 2023) I’m working on - sometimes with LLMs, sometimes without.

Ineiti’s Projects

How do LLMs answer your question?

Mon, 17 Nov 2025 19:28:00 +0100

The dream of an interaction with our computers through natural language is very old, and many different systems have been tried. A breakthrough has been reached in 2022 with Large Language Models, based on Transformers¹, which are able to interact through natural language. Nowadays we have a conversation with our computers in a way which was not possible just five years ago.

But how does this work? Where do Large Language Models (LLMs) get their information from? And why do they sometimes create answers which sound correct, but are completely wrong? Let’s simplify an interaction with a service like https://chat.deepseek.com with the following components:

Traefik 101

Tue, 07 Jan 2025 00:00:00 +0000

Many of us are running traefik. If you’re like me, you copy/pasted some docker-compose.yaml from a webpage, added some labels to your other docker-compose.yaml files, and things went up and running. Sooner or later you want to have something more fancy, and you need to understand what you actually did. RTFM and all that, you look at https://doc.traefik.io, and … it’s a mess. Honestly, I’m lost in this documentation.

In this post I’ll walk you through the differences between static and dynamic configurations, I’ll show what other configurations than http exist, which providers you need, and how entrypoints, routers, and services play together if you mix manual and automatic configurations.