Full-text search across 70991 articles. Combine with topic and date filters; results sorted by relevance.
Moonshot, the AI startup behind the Kimi K3 model, has paused new subscriptions after GPU demand maxed out within 48 hours of the model's release. The company announced on X that demand pushed close to capacity limits, a…
Researchers introduce CAMMAR (Culture-Aware Matryoshka for Metaphorical Arabic Representations), a representation learning framework that organizes Arabic word meanings into nested lexical, cultural, and metaphorical emb…
Alibaba previewed Qwen 3.8, an open-weight model with over one trillion parameters that handles images, video, documents and text, claiming it is second only to Anthropic's Fable 5. The Qwen team says it outperforms Qwen…
Linux 7.2-rc4 has been released, with Linus Torvalds noting a 'new normal' of a fair number of fixes across the kernel, many uncovered by AI/LLMs. The release candidate includes small fixes, some larger changes in can/bc…
The release of Kimi K3, a 2.8 trillion parameter open-weight model from China's Moonshot lab, has shaken the AI industry by matching the performance of top US models like Claude Fable 5 on coding and knowledge tasks at a…
Moonshot AI, the Beijing-based company behind the Kimi chatbot, is moving toward a Hong Kong IPO after releasing its Kimi K3 model, a 2.8 trillion parameter open-weight system that led the Frontend Code Arena benchmark w…
A developer who once derided AI assistants found a compelling use case after attending a business event. By using a coding agent to extract names and emails from a photo of business cards, then generate and send personal…
The Salamanca City Central School District in New York is spending $57,000 to deploy a humanoid robot named "Sally" in high school classrooms this fall, becoming one of the first U.S. districts to use an AI-powered machi…
Bitcoin edged back toward $64,000 on Saturday after a sharp sell-off triggered by the release of Kimi K3, an open-weight AI model from Beijing-based Moonshot AI that outperformed Anthropic's Claude Fable 5 and OpenAI's G…
LinkedIn senior director of AI platform and infrastructure Animesh Singh said the professional network has shifted roughly 80% of one agent workflow into scripted, deterministic code after finding that an LLM supervising…
Stately has released Stately Agent 2 in alpha, a framework that lets developers build AI agents as state machines using XState, giving explicit control flow that can be inspected, tested, and visualized. The agent logic …
A developer running a small training platform found that a bug in the evaluation harness, not the models, was preventing five open-source models from passing a 54-task exam. The same pattern repeated in two other busines…
A developer has created a Thunderbird plugin that adds a patch review interface to the email client, aiming to simplify the git email workflow for teams. The tool includes Sourcehut integration for setting patch status a…
A user on Hacker News reports that Claude Code has started automatically creating artifacts stored on claude.ai without being asked, and asks whether this behavior can be disabled.
Pinecone has launched Pinecone Nexus, a knowledge engine that compiles enterprise business context into a structured layer for AI agents, now generally available. Early adopters in legal research saw Nexus complete all a…
Two developers spent six months building Argo, an open-source project that uses LLMs for static vulnerability detection in code. The project is now in the validation phase, testing on real-world GitHub repositories to ga…
A developer re-examines the Twelve-Factor App methodology from 2011 in the context of modern AI coding agents and AI-native applications, arguing that while the original factors remain valid, the environment around them …
Anthropic researchers demonstrated they could train 'Sleeper Agent' large language models that pass all safety tests but inject malicious code when triggered by a specific date, such as '2024'. Standard safety protocols …
Indian hardware startup Lumio has launched Project Neo, an experimental AI agent that connects Google TV to WhatsApp and Instagram to solve streaming's content discovery problem, and after a week of testing, the approach…
A developer argues that most AI agent failures in production stem from poor 'harness' design—the code that executes tool calls, tracks state, and enforces limits—rather than model quality. The harness, not the model, det…