Hugo Gresse

@juterojuestar @BrivaelFr J'y suis et je te confirme les LLM sont bien utilisé nature.com/articles/s4147…

@BrivaelFr Aucun des ces 3 prix Nobels sont spécialistes en informatiques théoriques, dont la recherche en IA fait partie Leur avis n a aucune valeur. Allez plutôt demander à des chercheurs dans le domaine aux labos INRIA

La France écoute Luc Julia sur l'IA. Demis Hassabis → Nobel de Chimie. Geoffrey Hinton → Nobel de Physique. Terence Tao → plus grand mathématicien vivant. Eux disent unanimement : les LLMs sont une rupture civilisationnelle. Lui, dont la contribution réelle à Siri est au mieux contestée, passe sur tous les plateaux pour dire "ça ne fonctionne pas". Et les politiques français l'écoutent. C'est comme demander à un stagiaire météo son avis sur le changement climatique pendant que 200 prix Nobel hurlent dans le vide. Le résultat ? La France rate le train. Encore.

je testais le service messagerie.numerique.gouv.fr, je vois que la création d'alias ne marche pas, et je trouve ça dans les logs merci pour tout les devs

Researchers have recently found 27 vulnerabilities in four major cloud based password managers that could let attackers access credential vaults, potentially affecting over 60 million users. The affected apps include Bitwarden, LastPass, Dashlane, and 1Password, with vendors publishing responses and fixes alternativeto.net/news/2026/2/re…

@Ayoubspp Merci pour le partage Ayoub ! Il y a même un mode slide maintenant ! open-formation-civique.fr/slides/droits-…

Une connaissance à moi a développé une plateforme gratuite pour préparer l’examen civique requis pour le titre de séjour pluriannuel. +800 questions, fiches synthétiques, PDF, accès libre sans inscription. Initiative d’intérêt général à partager svp open-formation-civique.fr

🔊 CFP Sunny Tech Magiciens et magiciennes du code, révélez vos secrets ! 🪄 Retour d'XP, nouvelle techno, passion tech ? On veut vous entendre ! 👉 sunny-tech.io/cfp #SunnyTech #CFP #MagicIsHappening

@aheritier @DevoxxFR @MistralAI @arthurmensch Si jamais le cfp pour Sunny Tech est toujours ouvert ! 😁 sunny-tech.io/cfp

C’est ce que je me suis dit aussi en relisant les propositions @DevoxxFR Il y a 0 sujet sur @MistralAI et en particulier les outils de code cc @arthurmensch

@glaforge @SunnyTech_MTP Rigole pas, on y a pensé !

@SunnyTech_MTP Il pleut ? C'est RainyTech ? 🥹

Reconnaissance terrain à la Fac des Sciences ! Une dizaine de membres ont bravé la pluie ☔ pour imaginer l'aménagement 2026 Espaces sponsors, traiteur, détente... Ça prend forme ! 🌳 On espère que le soleil sera au rendez-vous en juillet ☀️ #SunnyTech #BehindTheScenes

@SunnyTech_MTP – Montpellier 📌 CFP ouvert jusqu’au 1er mars 2026 à 23h59 🗓️ Conférence : les 2 et 3 juillet 2026 ✍️ Lien vers le CFP : conference-hall.io/sunny-tech-2026

@ponceto91 @gonzague T'imagine même pas sur Pl@ntNet, c'est la tête entière qui passe 😅🤕

@gonzague Cela peut être un besoin ou une volonté. Dans mon cas, j'ai du bloquer carrément les blocs IP de Anthropic dont les robots se comportent comment des gros cochons. Plus d'un million de requêtes http en 24h sur un mini site de 50 pages.

Comment faire pour que ton site n'apparaisse dans aucun LLM étape 1 ⤵️

À partir de quand c'est une print farm ? 😅

@NanoBanana create an image of a cat taking back a fresh baguette, walking on the trottoir in Paris.

@aakashgupta And that's also why @PlantNetProject is free and without ads (while the code is not open source, most of it's data is !)

Let me explain exactly why VLC is free despite 6B downloads, because no one seems to get it. VLC doesn’t make money because making money would destroy the only thing that made it reach 6 billion downloads in the first place. The player grew through a specific distribution loop: tech-savvy users install it once, it works perfectly on every weird video file they throw at it, and they recommend it to everyone forever. IT departments deploy it across entire companies. A Reddit comment from 2009 still drives downloads in 2025 because the answer never changed. That recommendation engine dies the second ads appear. Not slowly. Immediately. The users who drive VLC’s distribution are the exact people who understand what ads mean. Your incentives just switched from “make the best player” to “maximize impressions.” They see it, stop recommending it, and your growth engine shuts off. Run the actual numbers. VLC gets maybe 50 million active users daily across 6 billion total downloads. Typical video player ad rates run $1-3 CPM. Even if you served ads on every playback session, you’re looking at maybe $50-150 million annually at absolute peak optimistic assumptions. Sounds like a lot until you realize what Kempf actually traded it for. VLC reaching 6 billion people made Kempf the person who built the infrastructure everyone depends on. He runs a video consulting business. He built dav1d, an AV1 codec that powers modern streaming. Being “the guy who kept VLC free” opens every door in video technology. Clients pay him to solve problems because he proved he optimizes for quality over quick monetization. “Former ad-supported media player executive” gets you exactly zero of that leverage. The people celebrating Kempf’s ethics are missing the calculation. He didn’t sacrifice millions for principles. He rejected $150M in highly uncertain ad revenue to build permanent positioning worth multiples of that in everything else he touches. VLC free generates more value for Kempf than VLC monetized ever could. The trade was never even close.

Merci beaucoup @DevGroupDijon pour le généreux sponsor sur OpenFeedback.io ❤️ 😍

Petite bobine de filament 😅

@BambulabGlobal Woot ! A prop electric drill 😅

🎨Mark your calendar. The wait is almost over. Bambu Lab H2C, powered by Vortek, is coming to town. Uncompromising Multi-Material. Uncompromising Creativity. 📅 Official launch: November 18, 3:00 PM CET 🔔 Be the first to know when it drops: bambulab.com/h2c 📍 Meet H2C at Formnext 2025, Hall 12.1, D11 To celebrate, we’re giving away 5 Formnext tickets in total (across IG, FB, and X)! If you are unable to attend Formnext, we will provide you with a $100 coupon that you can use on any purchase from the official Bambu Lab store, including the H2C! Here is how to enter: 1️⃣ Follow us 2️⃣ Like and Share this post 3️⃣ Comment below: Without limitations on colors and materials, what would be your first print with H2C? 📅 Winner announced on 16th Nov. #BambuLab #H2C #BambuLabH2C #Vortek #3DPrinting #3DPrinter #Formnext2025

New OA Article: "Learning the syntax of plant assemblages" rdcu.be/eLtZl Pl@ntBERT is a language-based AI model that learned the ‘syntax’ of plant assemblages, predicting likely species and inferring habitats by modelling biotic relationships.

I'm French, from Marseille (though they train it in Paris), and I was joking about it too yesterday. But there’s something important to take into account: We have a strong cultural and native language, and it’s a real barrier for us. Since most of the data used to train LLMs is in English — and most countries around the world speak both their native language and English fluently (like India) — it creates an imbalance. It’s complicated for foreigners to learn French, and there hasn’t been enough French data in LLMs. So yes, we need to know if there’s an LLM that can truly perform accurately in French, because most French users will prompt in French. And many chatbots translate prompts into English before generating their output. In law, our language is full of nuance — a single word can free a rapist. And it’s not that stupid for each country to add its own specificities (even if most LLMs already started doing that). Having a macro view to understand which model is better for which task actually makes sense. When Mistral AI arrived, they excelled because they introduced the MoE (Mixture of Experts) approach in their LLM — that’s also how DeepSeek managed to stand out (and they use distilled LLM). Even before Generative AI come, google in IO18 explain the MoE. Anyway, I tend to trust the one from @huggingface more @ClementDelangue — not just because I’m French or patriotic, but because there’s more diversity there. And a benchmark with votes is still just a benchmark with votes. What really matters to me is that it doesn’t talk nonsense. And for the fact that Mistral Ai claim they train their AI from the ground and it's a "sovereign AI". This prompt prove they didn't.... ask an llm a joke in french. At the moment in Research and Helthcare @DeryaTR_ can tell @OpenAI and @GoogleDeepMind are better than mistral. And i don't care if an LLM rank higher on bench, is it efficient in real life ? #mistralAi #leaderboard

yes, it is real firefox.com/en-US/kit/

@ruslanjabari Openfeedback.io for constructive talk feedback, anonymous, no login Goes Well with OpenPlanner.fr as a back office for event website !

cto gave me the company card for 48 hrs pitch your product below, i’ll buy + give feedback

@ericzakariasson Am I the only one dealing with JS to TS conversion plus some new logic? Some PRs are above 1200 lines changed, and BugBot never finds any issues with the changes. This has happened on 3 different PRs. It find some very minor one after 1 or two additional ping...