Topics on Mohamed Khelifa Zaghdidi

Self-hosted inference

Mon, 01 Jan 0001 00:00:00 +0000

What a model costs, what limits it, and what it reveals when you run it yourself on hardware you bought.

This series started in February 2026. The argument fits in one line: most of the decisions you make about an AI system are taken before the model, and verified after it.

Three things keep coming back. The real ceiling is almost never the one you expect. A measurement with no stated scope spreads faster than it can be corrected. And exhaustiveness is not something similarity search can deliver, however good it gets.

Data

Mon, 01 Jan 0001 00:00:00 +0000

What has to happen to a corpus before you hand it to anything.

Most of what identifies a person does not sit in a column. It sits in free text — a comment, a note, a written exchange — which is exactly what gets sent to a model. A clean schema tells you nothing about that.

The argument of this series is one of ordering. Governance applied after the model has already seen the data is documentation, not control. Whatever has entered the model cannot be recalled: logged prompt, built index, cached context, every stage keeps a copy.

Managing agents

Mon, 01 Jan 0001 00:00:00 +0000

{{/* Pas d’accroche ici, contrairement aux deux autres pages de sujet : sa paire francaise n’en a pas non plus. Le texte fourni ouvre sur une phrase de prose, pas sur une accroche. */}} I manage thirty engineers. And on a personal project of my own, outside my professional work, several positions are held by agents.

This series is about what running an organisation becomes when the positions have no memory, no career, and no pride. What carries over and what does not. What amortises and what gets paid in full every single time.