Édition sans rédaction — Bourse, Dev : l'IA n'était pas disponible ce matin. Voici les titres bruts des sources, tels quels ; la rédaction reprendra dès que possible.
Quand l'obligation d'utiliser l'IA au travail dérive : le témoignage d'un développeur
Un retour d'expérience publié sur le blog de Simon Willison décrit les effets pervers d'une automatisation intégrale du code imposée par le management dans une grande entreprise. Entre perte totale de contrôle technique et journées de 13 heures à appuyer sur Entrée, l'usage aveugle de Claude Code montre ses limites.
Fil « Témoignage sur l'imposition de Claude Code et l'automatisation du développement en entreprise »
Nouveau plugin llm-keys-ui pour sécuriser la gestion des clés API à distance
Simon Willison annonce la sortie de llm-keys-ui 0.1, une extension pour l'outil CLI LLM. Elle permet de configurer des clés d'accès API sur une machine distante via une interface web, sans devoir les copier-coller dans l'interface de discussion d'un agent de code.
simonwillison.net
Donald Trump propose la création d'une AI Force et d'un haut responsable dédié
Alors que les appels au ralentissement de l'intelligence artificielle se multiplient, Donald Trump annonce vouloir créer une « AI Force ». L'initiative s'accompagnerait de la nomination d'un haut responsable pour soutenir la croissance du secteur technologique.
theverge.com
Jensen Huang affirme qu'il y a zéro risque d'une fin du monde liée à l'IA
Le PDG de Nvidia, Jensen Huang, a fermement balayé les craintes d'un risque existentiel lié à l'intelligence artificielle lors d'une interview télévisée. Il s'oppose à toute nouvelle régulation du secteur, estimant que les appels au ralentissement lancés par OpenAI et Anthropic ne reposent sur aucune base scientifique.
theverge.com
Pourquoi l'humain armé d'IA reste la pire menace pour les réseaux d'énergie
Même si les scénarios d'agents autonomes devenus incontrôlables occupent l'actualité, les experts en cybersécurité rappellent une réalité plus immédiate. La menace principale pour les infrastructures énergétiques vient d'attaquants humains dont les capacités sont démultipliées par les grands modèles de langage. Face à des réseaux industriels anciens et difficiles à corriger, la vitesse d'attaque dépasse désormais celle des défenses.
theverge.com
En bref
- Huntington Cut Its Outlook Right as the Fed Hiked. First Crack in Regional-Bank Margins? — Yahoo Financefinance.yahoo.com
- Firmus Seeks to Raise $5 Billion in Landmark Australian IPO — Data center operator Firmus Grid Ltd. is looking to raise at least $5 billion in an initial public offering, according to terms of the deal seen by Bloomberg, kicking off one of the biggest-ever listings in Australia.bloomberg.com
- Apple (AAPL)’s New CEO Bets Big on a $1,999 Foldable iPhone — Yahoo Financefinance.yahoo.com
- Analog Devices (ADI) Bets $1.35 Billion on Chips that Let Machines Think for Themselves — Yahoo Financefinance.yahoo.com
- Prysmian, Rio Tinto partner to bring low-carbon aluminum to Amazon data center — Investing.cominvesting.com
- American Eagle (AEO) Holds its Forecast Steady, But Investors Sell Anyway — Yahoo Financefinance.yahoo.com
- Vietnam joins FTSE Russell emerging market benchmark — Investing.cominvesting.com
- Paramount, California AG discuss concessions including $1.5 billion investment, WSJ reports — Investing.cominvesting.com
- Wheat Jumps After Positive Signals From Initial US-China Talks — Wheat and corn futures rose after initial US-China trade talks were described as positive by both sides, raising hopes for the leaders’ summit later this week.bloomberg.com
- Trump ‘open’ to meeting Iranian President Pezeshkian- Fox reporter — Investing.cominvesting.com
- Kurt Campbell: US-China AI Talks Won't Limit Beijing — Former Deputy Secretary of State Kurt Campbell says talks between Washington and Beijing on security and AI will struggle to yield binding limits due to mutual distrust. He adds substantive progress may have to wait for a potential meeting between President Trump and President Xi at APEC or the G20.bloomberg.com
- India’s FX Reserves Dip, Snapping the Longest Rising Streak in Over Five Years — RBI interventions in the currency market could put further pressure on reserves.bloomberg.com
- Canada optimistic about concluding trade talks with India by year-end, minister says — Investing.cominvesting.com
- China’s Selective Crop Buying Tests US Trade Truce Before Summit — China is over halfway to meeting its pledge to buy 25 million tons of US soybeans this year, but progress on a separate commitment to spend billions of dollars more on American crops has largely stalled.bloomberg.com
- Should You Buy Goldman Sachs Stock After Its 26% Run-Up From Its 52-Week Low? — Yahoo Financefinance.yahoo.com
- Bot-free self-hosted analytics with GoatCounter on NixOS — Commentsvincent.bernat.ch
- What happened to the Snowden archive — Article URL: https://libroot.org/posts/what-happened-to-the-snowden-archive Comments URL: https://news.ycombinator.com/item?id=49780820 Points: 164 # Comments: 71libroot.org
- Your AI agent failed. The model might not be the problem. — As AI agents move into production, the path between a request and its result is becoming less predictable. An agent The post Your AI agent failed. The model might not be the problem. appeared first on The New Stack.thenewstack.io
- JEP targeted to JDK 28: 401: Value Objects (Preview) — The following JEP is targeted to JDK 28: 401: Value Objects (Preview)inside.java
- v16.4.0-canary.37 — Misc Changes Preserve closed-parameter restrictions in client route prediction: #98889 [test] Fix deployment tests that relied on implicit startup: #98935 Turbopack: Add support for specifying additional roots: #98003 Turbopack: Add symlinks and additional roots to NFT metadata: #98469 Credits Hugegithub.com
- Copilot code review: An improved review experience — Copilot code review now gives you a clearer view of how a review changes over time, more intelligently auto-resolves its own suggestions, and generates useful commit messages when you accept… The post Copilot code review: An improved review experience appeared first on The GitHub Blog.github.blog
- Spend Management expands to Enterprise Flexible Commitment plans — Enterprise teams on Flexible Commitment plans can now use , already , at no additional cost.Spend Managementavailable on Pro You can set a budget at any time in .Spend Management settings Set a budget per billing cycle, and when your team's metered usage approaches or crosses it, Spend Management cavercel.com
- Saving another 100TB of RAM with math (and Rust) — Cloudflare's global network is immense but not limitless. As we look for small ways to trim our resource usage, we sometimes get lucky and we can cut significantly more. Here’s how we reduced one of our Pingora-based service's RAM usage with statistics.blog.cloudflare.com
- Leave the Class Path in the Rearview Mirror — Netflix Tech Blognetflixtechblog.com
- What's been going on in w64devkit the past year — Commentsnullprogram.com
- Google's Open Agentic Orchestrator — Article URL: https://agentexecutor.io Comments URL: https://news.ycombinator.com/item?id=49780797 Points: 196 # Comments: 85agentexecutor.io
- One engineer shipped 2,000 PRs a month to production. Verification is the key. — Lauren Tan, an engineer on the Grok team at SpaceXAI, who previously worked at Cursor and Meta, recently published a The post One engineer shipped 2,000 PRs a month to production. Verification is the key. appeared first on The New Stack.thenewstack.io
- v16.4.0-canary.36 — Misc Changes Persist resurrected tasks as new tasks: #98729 Rspack: Remove bincode dependency from rspack by way of turbo-unix-path: #98855 Pin tasks across restore waiter handoff: #98728 test: honor skipStart for Vercel deployments: #98776 Use internal prefix for Turbopack shutdown env: #98747 testgithub.com
- Manage the code coverage ruleset condition with the REST API — You can now use the generally available REST API to manage the Restrict code coverage repository ruleset option, in addition to the existing UI support. This ruleset lets you enforce… The post Manage the code coverage ruleset condition with the REST API appeared first on The GitHub Blog.github.blog
- WebMCP support now available in mcp-handler — now has experimental support for , the proposed web standard for exposing tools to in-browser agents. Add a single script tag to your site, and your existing MCP tools become available there too.mcp-handlerWebMCP Opt tools in by adding them to the object:experimental_webMcp Then load the script fromvercel.com
Sans rédaction IA · titres bruts des sources