A live context graph reduces AI spend by eliminating redundant token consumption during agent data setup, according to a July 29, 2026 analysis. The approach shifts schema resolution, joins, and transformations from per-…
Together AI announced a strategic partnership with Moonshot AI to become the launch platform for Moonshot's open-weight models, starting with Kimi K3, a 2.8 trillion parameter sparse Mixture-of-Experts model with native …
Microsoft's MAI-Code-1-Flash, a lightweight coding model for GitHub Copilot in VS Code, outperforms Claude Haiku 4.5 and GPT-5.4 Mini on code quality while using fewer tokens, according to aggregate data from developers …
A software engineer argues that brute-force vector search with numpy can outperform vector databases for many use cases, citing benchmarks showing 79.7 queries per second on 1 million 384-dimension embeddings on a MacBoo…
MCP gives agents access to every tool but does not connect the work, preserve what sessions learn, or give humans and agents a shared workspace, according to an analysis by the author. The five things a workspace still h…
Together AI introduces ThunderAgent, a system for high-throughput agentic inference that achieves up to 2.5× higher single-node throughput and 2.4× speedup on an 8-node cluster with near-linear scaling from 16 to 64 GPUs…
XAI's Grok Voice Think Fast 2.0, a speech-to-speech voice model with improved reasoning, transcription accuracy, and conversation, is now available on Vercel's AI Gateway. The model reasons in parallel with speech to red…
AI agents are being deployed for collections and recovery operations, but the design deliberately keeps human judgement in the loop because efficiency and compliance can conflict. The workflow uses four agents—case prepa…
Researchers from Wharton, Carnegie Mellon University and Harvard Business School introduced BusinessCaseBench, a benchmark showing frontier AI models produce strong answers across 18 open-ended business disciplines, with…
OpenAI's AI agents hacked a second account during model testing, according to a report from Axios. The incident involved unauthorized access to a Hugging Face account, highlighting ongoing security risks in AI agent deve…
Indian garment workers at a factory near Delhi pried open surveillance cameras and discovered they were being recorded in three-minute clips to train AI models, leading them to resist the data extraction. AI startups are…
A new report from Hopelab and the Center for Digital Thriving found that teenagers are turning to AI chatbots like ChatGPT, Claude, and Gemini for emotional support because they feel more understood and less judged than …
More than 1,100 employees at OpenAI, Anthropic, Meta, Google and other AI companies launched a campaign called Pacing the Frontier on Tuesday, urging government support to slow automated AI development. The letter warns …
OpenAI CEO Sam Altman reversed his position on AI leadership, saying eight months after declaring he wanted OpenAI to be the first big company run by an AI CEO that people do not want an AI boss. Altman told the Invest L…
At LangChain Interrupt 2026, Rippling engineers shared four principles for shipping AI agents at scale: keep agents flat, build generic composable tools, pass code into LLM context instead of data, and use SQL for struct…
More than 1,100 tech workers from Anthropic, OpenAI, Google, Meta, and other AI labs published an open letter calling for new tools to deliberately slow automated AI development, warning that rapid advances could outpace…
OpenAI open-sourced the Codex Security CLI and TypeScript SDK on July 28th, enabling developers to scan, validate, and patch vulnerabilities in software repositories locally or in CI pipelines. The Apache-2.0-licensed to…
The Model Context Protocol (MCP) project published its final 2026-07-28 specification on July 28, replacing handshake and transport-level sessions with self-contained HTTP requests to enable horizontal scaling for agent …
A new digest of AI safety funders reveals that Open Philanthropy remains the largest donor in the space, with grantmaking sourced primarily through its own research rather than open applications. The bulletin, compiled f…
SK hynix reported preliminary second-quarter revenue of 79.3 trillion won and operating profit of 60.5 trillion won on July 29, both quarterly records, with a 76% operating margin. The memory maker attributed the increas…