iiinova — where your digital self keeps evolving

Reposted by Antoine Michéa

CloudNativePG @cloudnativepg.bsky.social · 12/08/2025

🚀 We've released CloudNativePG 1.27.0 and we recommend upgrading to benefit from the latest features, enhancements, and long-term stability: cloudnative-pg.io/releases/clo... Enjoy: ✨ Dynamic loading of PostgreSQL extensions 🔁 Logical decoding slot synchronization 🔒 Primary Isolation Check

cloudnative-pg.io

CloudNativePG 1.27.0 Released!

The CloudNativePG community is excited to announce version 1.27.0, featuring powerful new capabilities for PostgreSQL on Kubernetes.

064

Reposted by Antoine Michéa

Adina Yakup @adinayakup.bsky.social · 31/07/2025

Qwen team did it again!🔥 They just released Qwen3-Coder-30B-A3B-Instruct on the hub huggingface.co/Qwen/Qwen3-C... ✨ Apache 2.0 ✨30B total / 3.3B active (128 experts, 8 top-k) ✨ Native 256K context, extendable to 1M via Yarn ✨ Built for Agentic Coding

huggingface.co

Qwen/Qwen3-Coder-30B-A3B-Instruct · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

032

Reposted by Antoine Michéa

Martin Stefanko @xstefank.bsky.social · 16/06/2025

I'm super happy that I can share that the FREE FULL VERSION of Quarkus in Action continues to be available until January 16, 2026!!! Download and share! developers.redhat.com/e-books/quar... Thanks to Red Hat developer!!!

developers.redhat.com

Quarkus in Action | Red Hat Developer

Download Quarkus in Action, a practical guide to building resilient and scalable, cloud-native, enterprise Java applications using the Quarkus framework

03319

Reposted by Antoine Michéa

CloudNativePG @cloudnativepg.bsky.social · 23/05/2025

CloudNativePG 1.26.0 and 1.25.2 released 🎉 New in CloudNativePG 1.26.0: 🔹 Declarative offline in-place major upgrades of PostgreSQL 🔹 Enhanced startup and readiness probes for replicas 🔹 Declarative Management of Extensions and Schemas

222

Reposted by Antoine Michéa

Devoxx France @devoxx.fr · 12/05/2025

Les vidéos de Devoxx France 2025 seront toutes publiées demain matin. Vous pouvez voir et revoir 230 vidéos dont : - 114 conférences de 45 mn - 12 Deep-dive de 3 heures - 6 Keynotes - 52 Lunch-talks de 15 mn - 46 Tools-in-Action de 45mn Total 157,50 heures (soit 6.5 jours non stop) #DevoxxFR

711042

Antoine Michéa @michea.net · 08/04/2025

Je viens de découvrir Apache ECharts, une bibliothèque JS opensource incroyablement puissante pour créer des visualisations de données interactives et personnalisables. Avec plus de 20 types de graphiques, une gestion efficace des grandes quantités de données,à essayer rapidement! echarts.apache.org

000

Reposted by Antoine Michéa

Thomas Wolf @thomwolf.bsky.social · 31/03/2025

Wow - open-source LLMs for code are so back Just one week after the latest DeepSeek release, here is the All-Hands team dropping a 32B model with matching performance on software engineering tasks benchmarks (SWE-Bench). 20 times smaller than the DeepSeek model --

77717

Reposted by Antoine Michéa

BeeAI @beeaiagents.bsky.social · 14/03/2025

💥 Introducing BeeAI Alpha!💥 The open source platform to discover, run, and compose AI agents from ANY framework. Try BeeAI today to: 🌐 Run agents from any framework 🔎 Search an agent catalog to find and integrate agents ⚙️ Connect specialized agents to create workflows & more! Try now at beeai.dev

0198

Reposted by Antoine Michéa

Bryan Smith @bryansmith.io · 11/03/2025

Anthropic's CEO, Dario Amodei "in 3 to 6 months AI is writing 90% of the code ... in 12 months we may be in a world where AI is essentially writing all of the code" 🤯 3 months ago I wouldn't have believed this. But in the past few weeks I've found Cursor+Claude 3.7 is writing 90% of my code. You?

221

Reposted by Antoine Michéa

Tim Hockin @thock.in · 14/02/2025

I think we all know that YAML defines all of these as boolean false: no, No, NO, false, False, FALSE But did you know that these are all integers in YAML: 42, 42_, _42, _4_2_, __________4__________2__________ ... and they all mean the same thing?

6197

Antoine Michéa @michea.net · 30/01/2025

Plutôt rapide le nouveau modèle Mistral Small 3 sur ollama avec un m4

000

Antoine Michéa @michea.net · 30/01/2025

🇫🇷 Mistral AI lance aujourd'hui Mistral Small 3, modèle de 24B de paramètres sous licence Apache 2.0. Annoncé comme plus rapide et efficace que ses concurrents, il rivalise avec des modèles 3x plus grands 81% de précision sur MMLU 150 tokens/s en latence Ok avec RTX 4090 ou Mac 32GB #IA #OpenSource

000

Reposted by Antoine Michéa

Cédric Levieux @farlistener.bsky.social · 09/12/2024

📅9/24🎄 #IAAdventCalendar Ok nous avons un modèle avec 8 myards de paramètres. Mais il pèse combien ? On va commencer par lourd. Et puis on va parler quantification. Un modèle entraîné est habituellement taggé en FP16. FP est pour Floating Point ou en français nombre à virgule flottante.

131

Antoine Michéa @michea.net · 04/12/2024

Les prédications du CEO d'Hugging Face relatives à l'IA pour 2025 et le bilan de celles réalisées pour 2024.

Post LinkedIn de Clem Delangue, Co-fondateur & CEO de Hugging Face, partagé il y a 1 jour, présentant six prédictions pour l'IA en 2025 et un bilan des prédictions 2024. Les prédictions 2025 incluent : première manifestation majeure contre l'IA, baisse de capitalisation d'une grande entreprise, 100 000 robots IA pré-commandés, leadership chinois dans l'IA, avancées en biologie/chimie, et 15M de développeurs sur Hugging Face. Le bilan 2024 montre plusieurs prédictions réalisées dont la faillite d'entreprises surévaluées, les progrès des LLMs open source, et l'IA générative dans les médias.

010

Reposted by Antoine Michéa

Sung Kim @sungkim.bsky.social · 04/12/2024

QwQ is a code architect, not an editor Used alone without being paired with an editor, QwQ was unable to comply with even the simplest editing format. It was not able to reliably edit source code files. As a result, solo score on the benchmark was quite underwhelming. aider.chat/2024/12/03/q...

aider.chat

QwQ is a code architect, not an editor

QwQ is reasoning model like o1, and needs to be used as an architect with another model as editor.

082

Reposted by Antoine Michéa

please-open.it @please-open-it.bsky.social · 02/12/2024

In Keycloak, you MUST take a look and uncheck "full scope allowed" checkbox if you use roles. blog.please-open.it/full-scope-a...

101