OpenAI Ships Keyless Auth and Secure MCP Tunnels for Enterprise Teams
OpenAI ships private MCP tunnels, keyless IAM auth, and expanded Admin API controls — making enterprise-grade AI deployments finally practical
20 stories
OpenAI ships private MCP tunnels, keyless IAM auth, and expanded Admin API controls — making enterprise-grade AI deployments finally practical
Sesame's iOS app brings four lifelike voice agents to 39 countries, powered by a speech model that searches the web while it talks
IBM and Artificial Analysis launch ITBench-AA, a Kubernetes SRE benchmark where every frontier model fails more than half the time — and more turns make things worse.
Pika adds in-app Community and Support Chats to its AI Agent platform, turning solo creative sessions into a shared, multiplayer experience
Gemini Spark and Daily Brief bring always-on autonomous agents to the Gemini app, running 24/7 on Google Cloud even when your devices are off
LlamaIndex rewrote LiteParse in Rust, delivering up to 100x faster document parsing that now runs natively in Python, Node, Rust, and the browser via WASM
TokenSpeed's open-source inference engine hits 580 tokens/sec on a 397B MoE model, using kernel fusion, pointer tricks, and CPU-GPU overlap to keep Blackwell GPUs fully saturated.
Perplexity open-sources a Rust Unigram tokenizer that cuts CPU latency 5x over HuggingFace, reaching 63µs at 514 tokens with zero heap allocations
Cognition closes a $1B Series D at a $26B valuation as Devin writes 89% of its own code and revenue rockets from $37M to $492M in 12 months
Krea opens its first in-house foundation image model to developers, with tunable creativity, style transfer, and moodboard support via REST API

Callstack's open-source CLI gives AI coding agents a live feedback loop on iOS, Android, and TV apps through accessibility snapshots instead of screenshots.
Sakana AI's DiffusionBlocks trains neural networks one block at a time, cutting GPU memory proportionally while matching end-to-end performance across five architectures
ElevenLabs launches a Speech Engine Skill that adds real-time voice to any existing LLM-backed agent in a single command, with zero architecture changes.
Lovable now spins up parallel read-only subagents to research, explore, and QA your codebase — making large builds faster and often cheaper
Hermes Agent now ships a curated, one-click MCP catalog — letting you install reviewed integrations like GitHub, Linear, and n8n without touching a config file.
Runway's new MCP server lets Claude, ChatGPT, and Cursor generate production-ready images and video without leaving your workflow
Warp's new Cloud Handoff lets you close your laptop mid-task and have your AI agent seamlessly continue running in the cloud, with full context preserved.
StepFun open-sources a 198B MoE vision-language model built for agentic coding, search, and multimodal workflows — with a Blackwell-native FP4 variant for maximum throughput
PaddleOCR-VL-1.6 hits 96.33% on OmniDocBench — a 0.9B model that beats models 260x its size on document parsing

Rowboat is a free, open-source, local-first AI coworker that builds a persistent knowledge graph from your email, meetings, and notes — and acts on it.