Archive
Every article published on Tech & IA, newest first, grouped by year.
2026
Measuring a website: Core Web Vitals without the jargon
Three measures summarise what a visitor feels: how fast the main content appears, how responsive the page is to interaction, how stable the layout is. What they mean, how to measure them, and where to start.
Small models, big effects: AI that runs on your own machine
A model with a few billion parameters fits in the memory of an ordinary laptop once quantised. What it does well, what it does badly, and how to try it without installing anything exotic.
A hardened container image, point by point
Non-root user, read-only filesystem, SBOM, signature: what each measure really prevents, in what order to apply them, and what they do not replace.
GitOps explained: ArgoCD, desired state in Git, drift corrected
In a GitOps approach, Git describes the desired state and a controller inside the cluster enforces it continuously. At the scale of a small k3s cluster, here is what changes: deployments, rollbacks and drift.
Prompt injection: the flaw agents have to learn to live with
A model does not firmly tell an instruction from a piece of data: it is all text. As long as that holds, an agent reading outside content can be hijacked. You cannot remove the flaw, you can bound what it can cause.
Open models or closed APIs: how to choose
Host a model yourself or pay for an API: the right answer depends on five criteria you can examine one by one. Real cost, privacy, latency, licence and dependency, without dogma.
RAG: making a model answer from your own documents
A model has not read your documents. RAG means finding the right passages before you ask it the question. Chunking, embeddings, hybrid search, citations: where quality is won and where it is lost.
What an AI agent really is
The word “agent” is used to sell everything. Underneath sits a fairly simple loop: a model picks an action, a program runs it, a result comes back. What matters is the limits you put around it.