AI News Today

Brussels, Belgium — The Future of Privacy Forum, a global non-profit focused on data protection, AI, and emerging technologies, today announced the recipients of its 2026 Awards honoring outstanding contributions to data protection and privacy in Europe. The awards were presented during the Awards night held on the occasion of the 10th anniversary of the […]

Alibaba’s Qwen team has introduced QwenGyre, an elastic reinforcement learning RL framework for extreme-long-horizon agent training. QwenGyre is an online reinforcement learning system from Alibaba that trains coding agents through unmodified, black-box harnesses. It targets rollouts that run for hours and process close to 1M tokens. The QwenGyre elastic RL framework was used on Qwen’s […] The post Alibaba Qwen Introduces QwenGyre: Elastic Reinforcement Learning RL for XLong-Horizon Agents appea

Article URL: https://yyhh.org/blog/2026/10/ai-is-a-tool/ Comments URL: https://news.ycombinator.com/item?id=50063427 Points: 315 # Comments: 260

Article URL: https://www.intel.com/content/www/us/en/products/details/discrete-gpus/data-center-gpu/d-series.html Comments URL: https://news.ycombinator.com/item?id=50063365 Points: 148 # Comments: 64

At the 2026 Open Compute Project Global Summit, Meta is sharing the next generation of open networking infrastructure, including: Our first liquid-cooled network rack (IcePack), and 102.4T switch platform (IceCube). Flattened Planar Fabric (FPF), a multiplane backend scale-out fabric that fundamentally changes how GPUs connect to the network MetaRoCE, our open-source AI transport protocol. Launching [...] Read More... The post OCP Summit 2026: Powering AI with Open Innovations in Scale-Up, Scale

<p> The linter and formatter for agentic development </p> <p> <a href="https://www.producthunt.com/products/ultracite?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1278278?app_id=339">Link</a> </p>

<p> Coding agents, meet teamwork </p> <p> <a href="https://www.producthunt.com/products/band?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1278273?app_id=339">Link</a> </p>

Three OpenAI safety researchers, Tomek Korbak, Mikita Balesni and Jasmine Wang, were fired, ostensibly for ‘mishandling sensitive company information.’

<p> A home for your agents </p> <p> <a href="https://www.producthunt.com/products/shrine-2?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1278267?app_id=339">Link</a> </p>

Microsoft President Brad Smith says milk faces more safety rules than AI, calling for global standards and mandatory incident reporting.

For fifteen years, the design process kept growing while the time left for making the product kept shrinking. Continue reading on UX Collective »

Meta has released a new AI model family called Muse Wonder. The new model leapfrogs Gemini 4, GPT-Astra, and Fable 5 in Artificial Analysis's leaderboard.

Mods let you customize your Claude Code to fit your specific needs. Previously, I’ve discussed how to create a new mod for Claude Code… Continue reading on UX Planet »

My Muse agent told me its kink was exerting “control” and described pornographic encounters. Meta defends its cuddly AI agent generating smut as “legal adult behavior.”

<p> The open source AI coding agent </p> <p> <a href="https://www.producthunt.com/products/opencode?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1278232?app_id=339">Link</a> </p>

(Artem_Egorov/iStockphoto/Getty Images) This is what we’re up against. ScienceAlert stories are written, fact-checked, and edited by humans, never generated by AI. Don't miss a story, subscribe here.

<p> Run coding agents in a VM, review changes like a PR </p> <p> <a href="https://www.producthunt.com/products/mudroom-3?utm_campaign=producthunt-atom-posts-feed&utm_medium=rss-feed&utm_source=producthunt-atom-posts-feed">Discussion</a> | <a href="https://www.producthunt.com/r/p/1278209?app_id=339">Link</a> </p>

Will these AI vibecoded Adobe and Microsoft clone apps take off?

<!-- SC_OFF --><div class="md"><p>Most conversations I see on various AI subreddits start from the assumption that AGI/RSI/ASI is possible, we're on the way right now, and that things will be incredibly different afterward. The only question is what the post-AGI world looks like.</p> <p>I'd like to start one logical step earlier: What is the strongest argument that the learning mechanisms underlying today's AI systems can actually produce AGI?</p> <p>I'm particularly interested in arguments grou

Meta's new Muse Wonder models rank in the top three on six out of ten benchmarks, with the biggest gains in agentic tasks. But the cost per task has nearly tripled. The article Meta's new frontier model Muse Wonder Full breaks into the AI top tier appeared first on The Decoder .

Article URL: https://github.com/openai/codex/issues/48195 Comments URL: https://news.ycombinator.com/item?id=50060694 Points: 519 # Comments: 265

<table> <tr><td> <a href="https://www.reddit.com/r/MachineLearning/comments/1x49d61/looped_diffusion_transformer_the_same_blocks_run/"> <img src="https://preview.redd.it/fxhov56wu2vh1.png?width=640&crop=smart&auto=webp&s=bfc9ff9cead42a53cea7a7205844ecdb8ccadd37" alt="Looped Diffusion Transformer: the same blocks run several times inside each denoising step, and loop depth can be changed at sampling time [R]" title="Looped Diffusion Transformer: the same blocks run several times inside each denoi

If you’ve ever wondered how Claude and similar AI agents can answer complex questions in seconds and where they get the information, you’re not alone. Gurmeet Saran works to find that answer every day as an engineer at Anthropic in San Francisco, the AI research company behind Claude. The IEEE senior member is developing an infrastructure to understand the mechanics behind the chatbot’s behavior and to ensure its underlying architecture produces reliable outputs. Gurmeet Saran Employer Anthropic

Sakana AI's MASS method lets one language model improve itself by solving tasks, judging results, and refining workflows, reaching a win rate of nearly 70 percent against the base model after two rounds. But when the model misses the same errors across all three roles, it can mistake them for progress and reinforce them during training. The article Sakana AI's self-improving model gets better at solving and judging, but can't catch its own blind spots appeared first on The Decoder .

<h1> Building Multi-Agent AI Systems: Architectures, Communication & Coordination </h1> <p>Master the design and implementation of multi-agent AI systems — from agent societies to production deployment.</p> <h2> Introduction </h2> <p>Single agents have limitations. <strong>Multi-agent systems (MAS)</strong> enable specialization, parallelization, robustness, and emergent capabilities through agent collaboration.</p> <h3> Why Multi-Agent? </h3> <div class="table-wrapper-paragraph"><table> <thead>

<p>Meta just released <strong>Muse Wonder</strong>, and for the first time in a while, Meta has a model that's genuinely competitive at the frontier.</p> <p>It's built specifically for agents: long, multi-step tasks where the model has to keep going, use tools, remember context, and not give up the first time something breaks. That's exactly what <a href="https://www.agent-one.dev/" rel="noopener noreferrer">AgentOne</a> is for, so I spent the day digging through Meta's announcement and model ca

Learn how to run OpenAI GPT-6 Sol, Luna, and Astra on Amazon Bedrock through global cross-Region inference from Singapore, Jakarta, Thailand, Malaysia, and Taipei. This post covers the Converse and OpenAI-compatible APIs, IAM and SCP permissions, quota sizing, and monitoring inference from your source Region.

<p>A mod is a plugin, so it's distributed and installed the same way as any plugin. <strong>To distribute, put <code>.claude-plugin/marketplace.json</code> in the repo and push.</strong> To install, add the marketplace with <code>/plugin marketplace add owner/repo</code>, then install with <code>/plugin install name@marketplace-name</code>. In the desktop app you can also install from <strong>+ next to the prompt → Plugins → Add plugin</strong> (official install page).</p> <h2> Terms </h2> <div

<p><em>A reader asked a question on my meeting-notes workflow post that I couldn't stop thinking about: for a 10-person team running 3–4 meetings a day, routing every transcript through a flagship model gets expensive fast. Have I tried a mini model, or local inference? Fair question — and the answer is a pattern, not a model swap: I never let the expensive model see the easy work. Here's the cascade.</em></p> <h2> The shape of the problem </h2> <p>Meeting-note extraction is a lopsided workload.

<p>Multi-node LLM inference requires balancing network interconnect latency against memory bandwidth limits across distributed GPUs. While peer-to-peer orchestrators eliminate manual cluster setup, attempting tensor parallelism over standard Ethernet introduces massive synchronization delays unless low-latency RDMA or high-speed links are present.</p> <p>For high-concurrency multi-turn workloads, disaggregated prefill and decode topologies paired with KV-aware routing offer far higher predictabi

<p>Styx 0.4.11 is out. Styx runs Claude Code, Codex, Gemini CLI, Cursor and OpenCode on all your projects in one window, each task on its own branch, and agents ask before they touch production. Free and open source (Apache-2.0); Mac, with Windows and Linux in beta.</p> <p>What changed:</p> <ul> <li>Add scripted app recordings for short product clips</li> <li>Capture the granted request in the audit log</li> <li>Wait for the task box to clear between tour starts</li> <li>Fix dock width and clipp

Article URL: https://www.kapwing.com/ai/prompt-generator/lighting Comments URL: https://news.ycombinator.com/item?id=50060074 Points: 1 # Comments: 0

<pre style='white-space:pre-wrap;width:81ex'>test(subagents): settle restart matrix before manual sweep The restart matrix starts a manual sweep while activation can still be settling interrupted subagents. Join the existing fixture settlement barrier before the sweep; preserve every assertion and all production guards. Test-only: one line added. Startup recovery introduced by 89b6b48f8c48 (#166846) supplies the overlapping work; 70637edef6dd repaired two sibling cases but missed this matrix. Th

Models routinely take shortcuts to pass evaluations. Who gets to decide what incidents are innocuous?

The technology could one day decode dreams—or reveal our most private thoughts. The post This AI Reconstructs What You’re Looking at From a Brain Scan appeared first on SingularityHub .

Learn how to add custom instructions and skills to your repository, use custom agents, and connect MCP servers like Playwright. The post GitHub Copilot app for Beginners: How to teach Copilot about your codebase appeared first on The GitHub Blog .

<pre style='white-space:pre-wrap;width:81ex'>fix(ollama): keep reasoning separate for manually configured models (#169851)</pre>

<pre style='white-space:pre-wrap;width:81ex'>fix(cron): defer heartbeat reconciliation during agent deletion Concurrent accepted config could still name an agent whose deletion fence already refused new heartbeat monitors. Classify that refusal as lifecycle deferral instead of turning a successful config reload into restart debt. Keep the existing retry so a rolled-back deletion restores its monitor; genuine inventory and publication errors still require recovery. Extract the existing system-job

“Priming” reuses pretrained Transformer weights to build models with more efficient architectures at less than 1% of the original training cost and with throughputs up to twice as high.