New research from Microsoft resolves the Kannan–Lovász–Simonovits (KLS) conjecture, a 31-year-old grand challenge in mathematics with implications for high-dimensional computing.
NVIDIA's Kumo Relational model reportedly outperforms supervised machine learning on relational data, notably achieving this without requiring any prior training.
OpenAI, PerturbAI, and Scripps Research leveraged thousands of distributed AI agents to identify promising research leads within millions of brain cell data points.
OpenAI details the development of ChatGPT's ad stack, explaining how LLMs are used to understand conversational intent, ensure ad separation, and deliver relevant ads at scale.
GitHub details its massive migration of repositories to Azure, explaining how per-reference quorums and object storage are scaling Git for agent-driven workloads.
Google's Agents CLI Extension System enables unified deployment and evaluation workflows for AI agents, supporting frameworks like LangChain and custom stacks, with policy enforcement capabilities.
Claude introduces "Duet," a new AI offering available as a research preview by waitlist for US Pro and Max subscribers.
OpenAI and Google Cloud partner to bring frontier models and Codex workflows to Gemini’s Model Garden, accessible via Google Cloud Marketplace.
OpenAI's "Patch the Planet" initiative coordinates global responses to critical open-source vulnerabilities, leveraging AI agents to uncover and address threats.
Simon Willison details a critical PHP security release, the leak of RCE vulnerabilities discovered by AI, and the breakdown in the coordinated disclosure process.
Research indicates that specializing the vocabulary of Small Language Models (SLMs) can lead to a dramatic 77% reduction in inference costs for real-world applications.
Claude Security now offers continuous repository scanning, finding tracking, and context learning, available in public beta for Enterprise customers.
Granite Speech 5.0 is released, promising enhanced accuracy and speed for transcription, translation, and attribution tasks in speech AI.
An open-source mobile RTS game, Battle Drones, receives a delightful French fork called "Panache," reskinning it to steampunk 1900s France with new models and an animated intro.
This pnpm release introduces flexible workspace configuration delegation, tarball integrity pinning, and zstd compression support, alongside performance boosts.
This alpha release of pnpr, a registry server, introduces a new storage layout, ecosystem-specific URL prefixes, and enhanced artifact management.
This week's CodePen community highlights feature creative projects including generative SVG playgrounds, pure CSS dither dissolves, and intricate JavaScript mathematical art.
This article provides a system-level profiling guide for TorchNeuron, focusing on its performance characteristics when deployed on AWS Trainium hardware.
