Anthropic just rolled out a performance dashboard for Claude connector builders, alongside direct in-app submissions for MCP servers. Developers can now track active users, tool calls, error rates, and latency breakdowns across Claude, Claude Code, and Cowork. It is currently in public beta for Team and Enterprise admins, making it significantly easier to debug and scale MCP agents.
Alibaba just released Qwen 3.7 Plus, a multimodal model built to navigate graphical interfaces, write code from templates, and run terminal commands within a single loop. In testing, it autonomously built a complete vocabulary app over eleven hours, leading Western rivals on GUI benchmarks like AndroidWorld. It also supports Anthropic's API protocol, making it a dirt-cheap drop-in backend for existing agent tools.
Google just introduced Agentic RAG for its Gemini Enterprise Agent Platform, ditching the single-step retrieval approach. A specialized Sufficient Context Agent evaluates draft answers, spots missing details, and runs targeted follow-up queries, driving a 34% jump in factual accuracy with almost zero added latency.
Microsoft CEO Satya Nadella just rebuked an internal memo advocating to make users "addicted" to Scout, the company's new AI agent. Nadella told roughly 50 top AI engineers that addiction is "absolutely a non goal," suggesting the author might want to work elsewhere. The swift pushback draws a hard line between sticky enterprise software and predatory engagement tactics.
Nvidia just launched Nemotron 3 Ultra, an open 550B MoE model with 55B active parameters tailored for complex multi-turn agent workflows. It combines Mamba and Transformer layers to slash agentic task costs by up to 30% while pushing 5x higher throughput than comparable open models. Nvidia also rolled out a 4B safety guardrail and a sub-100ms multilingual speech model.
Meshy just launched Meshy 3D Agent Beta, shifting 3D creation from single one-off prompts into a guided, chat-based workflow. Creators can explore visual concepts, generate variations, and iteratively refine editable 3D assets for games, AR/VR, and 3D printing. The feature is now live for all users on meshy.ai.
Cloudflare just revealed that bot traffic now makes up 57.5% of all web requests, officially surpassing human browsing for the first time. The shift landed a full year ahead of internal projections, turbocharged by autonomous AI agents scraping and interacting across the internet. In short, machine-to-machine exchange has quietly become the default mode of the web.
Google just launched Gemma 4 12B, an open-source multimodal model that cuts out separate vision and audio encoders to feed signals directly into the LLM. It nears the performance of their 26B MoE model at less than half the memory footprint, letting developers run local agents on a 16GB laptop.
OpenAI just updated GPT-Rosalind, its specialized model series for enterprise life sciences research, baking in GPT-5.5's agentic coding and tool-calling skills. The new model beats GPT-5.5 on wet lab troubleshooting, medicinal chemistry, and genomics while burning up to 31% fewer tokens, aiming to slash the time scientists spend connecting dry-lab data to wet-lab experiments.
nex-agi just open-sourced Nex-N2, a new family of agent models post-trained on Qwen 3.5 for real-world coding and long-horizon tasks. Available in 397B and 35B MoE variants activating just 17B and 3B parameters, its adaptive reasoning cuts compute waste by deciding when to think deeply and when to act fast.
Microsoft just unveiled Project Solara, a chip-to-cloud platform built for agent-first hardware rather than traditional apps. Running on an enterprise-grade Android foundation with Azure state management, it introduces dynamic just-in-time interfaces and previews two concept form factors—stationary and portable—aimed at healthcare, finance, and retail.
Microsoft just introduced Frontier Tuning in private preview, a reinforcement learning approach designed to customize AI directly inside an enterprise's compliance boundary. By learning from actual internal workflows and tools without leaking data, the system boosted Microsoft's own HR agent completion rate from 13% to 87%. It is heading to Copilot Studio and Foundry soon.
Microsoft just unveiled Scout, its first Autopilot agent designed to work continuously in the background across Microsoft 365. Operating with its own governed Entra identity, it proactively schedules cross-timezone meetings, blocks focus time for deliverables, and flags stalled decisions without waiting for prompts. Currently in private preview for Frontier organizations.
GitHub just launched the Copilot app, a dedicated desktop control center built to orchestrate multiple AI agents in parallel. Each agent operates in an isolated git worktree so tasks don't collide, while interactive canvases let developers inspect and steer work in real time. It is now available in technical preview for paid Copilot subscribers.
Microsoft just announced its Work IQ APIs, launching June 16, 2026, to give AI agents direct access to Microsoft 365 organizational context instead of raw data. The APIs collapse operations into 10 generic tools via MCP and bill through Copilot Credits. It essentially allows enterprise agents to reason with full business context.
Microsoft just launched Web IQ, a specialized suite of grounding APIs that connects AI agents to real-time web data. Instead of dumping full documents, it serves passage-level structured evidence to slash token consumption, delivering sub-165ms p95 latency that is roughly 2.5 times faster than competitors. It gives multi-step reasoning agents a high-speed factual backbone without blowing their context budgets.
Microsoft just transformed Windows 11 into a playground for AI agents at Build 2026. The update introduces Microsoft Execution Containers (MXC) to box in and manage local agents safely, built-in WSL containers, and an Intelligent Terminal with live agent debugging. It also packs new on-device SLMs dubbed Aion 1.0 alongside heavy AI dev hardware. It is a clear push to keep agent workflows local and secure.
Nous Research just launched Hermes Desktop in public preview, letting users run its Hermes Agent natively on local machines without relying on the cloud. First demoed during Jensen Huang's GTC keynote, the app brings the complete agent system directly to your own hardware.
Perplexity just announced a hybrid local-server inference orchestrator for its Personal Computer tool. Instead of making you choose between local privacy and cloud horsepower, the system automatically routes sensitive tasks like finances to on-device silicon while offloading heavy lifting to frontier server models. The Intel-partnered feature rolls out this July.
Anthropic just added dynamic multi-agent workflows to Claude Code, letting the AI write and orchestrate custom JavaScript harnesses for complex tasks. Powered by Claude Opus 4.8, it spawns isolated subagents to run patterns like adversarial review and tournament brackets without context bleed. It burns significantly more tokens, but it turns Claude Code into an autonomous project manager for messy, high-value tasks.
