Toolsllm-anthropic 0.27 adds support for Anthropic Python SDK 1.0's httpx2 migrationThe llm-anthropic plugin has been updated to version 0.27 to ensure compatibility with Anthropic's Python SDK 1.0.0, which migrates its HTTP client from httpx to httpx2; OpenAI made an equivalent change in their v3.0.0 release shortly before.1 min read
ToolsHands-free development with Codex voice mode: controlling browser and apps without a keyboardCraig Hewitt, founder of podcast hosting platform Castos, demonstrated how he uses the Codex application’s voice mode to run parts of his development workflow without touching a keyboard on the Zero to Agent in 30 Minutes podcast.3 min read
ToolsNVIDIA Scale‑In: BlueField‑4 és DOCA a nagysebességű, biztonságos agentikus AI infrastruktúráhozNVIDIA bemutatta a Scale‑In hálózati infrastruktúrát, amely a BlueField‑4 DPU, a DOCA szoftverkeret és a Spectrum‑X Ethernet együttműködésére építve célzott gyorsítást ad az agentikus AI gyárak észleléséhez, adathozzáféréséhez és biztonságához.5 min read
ToolsNVIDIA Vera CPU designed for balanced performance in agent-based AI fleetsNVIDIA positions the Vera CPU as a single, balanced design point for “agentic” AI workloads, arguing that these workloads combine long sequential reasoning chains with intermittent parallel fan‑outs.4 min read
ToolsGroq 3 LPX boosts low‑latency, long‑context interactive inference on the Vera Rubin platformNVIDIA’s Groq 3 LPX accelerator, paired with the Vera Rubin NVL72 platform, targets high‑interactivity agentic AI workloads with long context windows.5 min read
ToolsHawkeye: minimal supervision enables hardware-aware GPU kernel generationResearchers from Harvard, Stanford, Together AI, and Caltech released Hawkeye, an open-source framework that helps autonomous agents generate hardware-optimized GPU kernels using a small set of curated unit tests.3 min read
ToolsAnthropic shifts Claude from solo assistant to team-facing Slack agentAnthropic updated its Claude Tag Slack agent to read whole-channel context and act proactively across conversations, a change the company says makes the agent about 30% better at deciding when to intervene.7 min read
ToolsEnterprise AI Reliability Requires a Shared Knowledge Platform, Not Per-Application ContextCurrent enterprise AI practices build application-specific context (chunks, embeddings, indexes) from the same corporate sources, producing inconsistent, duplicated and hard-to-update knowledge.7 min read
ToolsExtension philosophy around 𝚏𝚡 built on open protocolsThe 𝚏𝚡 project team builds extensibility on open protocols, highlighting the use of MCP (modelcontextprotocol.io), Skills (agentskills.io) and Plugins (agent-plugins.org).1 min read
ToolsLinkdaze launches family-focused smart calendar tablet with AI meal-planning featureLinkdaze released a touchscreen calendar tablet aimed at organizing entire households rather than individual users.2 min read
ToolsDeveloper Builds Mini Personal, Minimalist Browser with fxA developer used a tool called fx to create a personal browser named Mini for sharing and testing in a simple, stateless environment.1 min read
Toolsllm 0.33 released: OpenAI Python 3.x support, httpx2 migration and embedding key handling expandedThe llm package reached version 0.33, upgrading to OpenAI's Python 3.x client and replacing the HTTP dependency from httpx to httpx2.2 min read
