This article is adapted from a LinkedIn post published on March 24, 2026. It reflects how I was working at that time, and I've deliberately kept it in its original context.
I get asked the same question often:
"What's the best LLM?"
For a while, I was looking for that answer too.
Today, I think the question is the wrong one.
I'm no longer looking for the best model. I'm looking for the best combination of models.
We've always worked with several tools
In software development, nobody looks for the best language, the best framework, or the best database.
We pick the right tool for the context.
Why would generative AI be any different?
Every model has its own strengths, limits, and personality. Using them together is, in the end, a fairly natural evolution of the way we already work.
Why a constellation?
I could have called it a toolbox.
But a constellation captures what I actually feel better.
Each star has its own identity.
Yet it isn't any single star that creates the value.
It's the shape they draw together.
My AI tools work exactly the same way.
The stars of my constellation
ChatGPT
My working companion.
I use it to think out loud, structure ideas, prepare articles, explore new directions, or make a topic more accessible.
The dialogue feels fluid and natural.
Claude
When a subject gets complex.
Architecture, code review, large documentation, long reasoning — it's usually the one I turn to when I need depth.
GitHub Copilot
The one that takes action.
It speeds up writing code, refactoring, tests, and repetitive tasks.
It doesn't replace the developer.
It reduces friction.
Mistral
I use it more out of technical curiosity.
Its European approach and fast pace of evolution are worth following.
It reminds me that several visions of generative AI exist.
What this approach changes
Since I stopped comparing models, I spend less time choosing a tool.
I spend more time building a coherent working environment.
Each AI steps in where it brings the most value.
The result isn't just a time saving.
It's a smoother way of working.
A reflection meant to evolve
Rereading this article today, I realize it marks an important step.
At the time, I was mostly discovering the complementarity between several LLMs.
My later experiments would take me toward other subjects: MCP servers, AI agents, and the orchestration of tools able to collaborate with one another.
But this idea remains, to me, foundational.
I'm no longer looking for the best AI.
I'm looking to build the best ecosystem around my work.