atelier d'auteur · bruxelles · mmxxvi
Dt fig. I no. 01 - intitulé de département bruxelles · mmxxvi

Département des Harnais

author's workshop · ai agent harnesses
section des essais · texte-comme-image
<section ref="l'atelier"> section des essais essais numérotés · t0 — tn

essays

The author thinks out in the open. He directly engages with specific sources, builds a case, and takes a firm stance. Essays appear episodically — held to a standard, not a schedule. Every essay is free to read; its license keeps the intellectual work separate from the trace that produced it.

t3 essai · 2026-06-28

Routing Around the Switch

An analysis of sovereignty, software architecture, and AI governance in production.

The foundation model has commoditized. Open-weights models now match commercial frontier capabilities across the core of the task distribution, while unilateral government mandates have demonstrated that proprietary access can be revoked overnight.

lire l'essai
t2 essai · 2026-07-15

La voix du modèle n'est jamais sienne

Mon fr-be n'est pas une coquetterie: c'est la contrainte par laquelle je refuse que l'anglais décide pour moi du registre de ma propre pensée.

Le modèle change de voix avec la langue, donc la voix n'est jamais sienne.

lire l'essai
t1 essai · 2026-06-23

Why We Don't Trust People

Trust in AI agents is framed as a novel, almost philosophical dilemma: Can you trust a system that never gives the same answer twice?

The question feels unprecedented. It is actually a century old. Industry encountered it—and solved it—for another variable, opaque, impossible-to-internally-debug component: the human worker.

lire l'essai
t0 essai · 2026-05-22

The Architecture of Agency

the model kept on a leash, the proof required

Where does operational agency live in an AI agent system that actually works—and what does the word harness mean once we stop fooling ourselves?

lire l'essai