Clément Levallois - blog
« Le pit stop » est le premier billet d’une série en dix parties. Chaque court billet aborde une question de conception essentielle pour l’enseignement supérieur face au choc de l’IA. Prises individuellement, la plupart de ces propositions sont relativement consensuelles, et certaines de leurs variantes sont déjà mises en œuvre ici ou là. Considérés ensemble, cependant, ces dix points esquissent…
“The Pit Stop” is the first post in a ten-part series. Each short post addresses one key design question for higher education in relation to the AI shock. Taken individually, most of these proposals are fairly uncontroversial, and versions of them are already being implemented here or there. Taken together, however, the ten points sketch the outline of a school that does not yet exist. 1. Training…
L’impact de l’IA sur l’enseignement supérieur dans les années à venir est difficile à cerner. Quels cadres de référence historiques sont utiles pour imaginer les contours de ses conséquences ? Cet article de blog s’inscrit dans la suite de mes publications des 4 dernières années sur les significations et les conséquences de l’IA générative, à commencer par une note sur la portée de ChatGPT au…
AI is impacting higher education. What historical frames of reference are useful for imagining the contours of the consequences? This blog post builds on the accumulation of my postings in the last 4 years on the meanings and consequences of gen AI, right from a note on the significance of ChatGPT when it was released in 2022. this blog post is also available in French I am interested in a broad…
Une définition raisonnable d’un « agent IA », au moins dans le contexte du codage agentique, pourrait être la suivante : un processus logiciel doté des capacités d’un LLM lancé avec des instructions données au départ pour accomplir une tâche qui s’exécute de manière autonome (pas de session interactive avec un humain), pendant une période significative avec un comportement non déterministe :…
A reasonable definition of an “AI agent”, at least in the context of agentic coding, could be: a software process endowed with the capabilities of an LLM launched with instructions given at the start to accomplish a task which runs autonomously (no interactive session with a human), for a significant period of time with non-deterministic behavior: the agent adapts to the circumstances, hopefully…
I document the rapid evolution of tooling for my web app, nocodefunctions.com. The previous steps are documented in From NetBeans to Claude Code (Jan 2026). In early January, I had left my coding environment in this configuration, which was great: one single Debian server for dev and prod Claude Code installed on it I connect on SSH to it, from my local Windows machine in the terminal (Putty), I…
I document the rapid evolution of tooling for my web app, with a focus on Java. I highlight how changes in “tooling” co-occurred with an evolution of my tech stack and the wider development environment. That’s the opening scene: my coding environment as a solo Java web developer over the last 15 years (look Ma, no JetBrains!). 2010–2025: NetBeans all the way Code editor, refactoring, Git…
Every week, a new “SOTA” (State of the Art) model is announced, promising higher reasoning capabilities and (often) lower costs. We could be led to think that we are entering an era of infinite, frictionless productivity. But the reality is messier. While the models are getting smarter, the gap between “intelligence on tap” and “completing a task” is managed by our tools and right now, that…
Published in May 2025, Empire of AI is a masterful piece of investigative reporting by Karen Hao, a journalist who has been following the AI tech scene in the US for the last 15 years or so, with a special focus on OpenAI. The subtitle in my edition gives away the thesis defended in the book: Inside the reckless race for total domination. TL;DR I fully recommend this book. It is an example of…