You're here. Good. The agents have been rummaging through the drawers again, and someone should probably know where the sharp objects are.
Pope Leo XIV released Magnifica Humanitas, an encyclical focused on human dignity in the age of artificial intelligence, with Anthropic's Chris Olah speaking at the Vatican presentation. It is not every day the Vatican puts model opacity, labor displacement, global inequality, and human flourishing on the same stage as frontier AI research. For companies treating AI governance as a quarterly checkbox, this is a reminder that the conversation has escaped the product roadmap and entered institutional doctrine. Vatican News Anthropic
Alibaba introduced Qwen3.7-Max, a proprietary model aimed at autonomous agent workflows across coding, office automation, and multi-step execution. The company says it can sustain hundreds or thousands of steps and demonstrated roughly 35 hours of continuous autonomous execution, which is either impressive or a cry for a responsible supervisor. For teams building agent harnesses, the signal is clear: endurance is becoming a model feature, not just an orchestration trick. Alibaba Cloud VentureBeat
Cohere released Command A+, a 218-billion-parameter sparse mixture-of-experts model with 25 billion active parameters, 128K input context, multimodal inputs, tool use, reasoning outputs, and support for 48 languages. The Apache 2.0 license and quantized weights designed for one B200 or two H100s make this unusually relevant for companies that want capable enterprise AI without sending every internal document through someone else's hosted frontier model. Sovereignty is more persuasive when it fits in the budget and the rack. Cohere Hugging Face
Google used I/O to push Gemini deeper into daily work: an overhauled Gemini app, AI-heavy Search, Gemini Omni Flash for conversational video creation, Flow agents and custom tools, and Workspace-native creative surfaces. The important part is not the pile of brand names, though there are quite enough of those. Google is trying to make AI the interface layer for search, creation, and work, which means every company relying on web discovery or knowledge workflows should be paying attention. Google I/O Google Search
GitHub now lets enterprise owners allow specific Copilot models for selected organizations, instead of relying only on one enterprise-wide model policy. A small governance feature, yes, but model access is becoming a compliance surface. GitHub
Fact-checkers debunked a fake video involving India's foreign minister and synthetic audio added to old Nigel Farage footage. Synthetic political media is becoming routine, local, and fast enough that same-day correction is now part of the operating model. The Quint Full Fact
Illinois advanced SB315 in a 52-5 Senate vote, targeting large frontier model developers with transparency frameworks, third-party auditors, and catastrophic-risk reporting. It is modeled on California and New York efforts, because apparently state-level AI governance is now a franchise system. WTTW
Microsoft Developer published an Agentic-Agile methodology: specs in the backlog first, contract-driven execution, incremental delivery, persistent repo instructions, governance as acceptance criteria, and measurement of agent collaboration quality. Sensible teams are moving from "prompt harder" to actual engineering process. Extraordinary development, really. Microsoft Developer GitHub
LaunchDarkly introduced AgentControl, an operational layer for changing prompts, models, parameters, and tools at runtime, with evals, tracing, progressive rollouts, and adaptive triggers. Feature flags have come for the machines that make decisions. Sensible, if mildly unsettling. LaunchDarkly LaunchDarkly
Resolve AI expanded its production operations platform with background agents, multi-agent investigation workspaces, a REST API, and an MCP server. The pitch is continuous triage: agents that watch deployments, pre-investigate alerts, flag drift, and prepare evidence before humans arrive. PR Newswire VentureBeat
Hark raised a $700 million Series A at a $6 billion post-money valuation, backed by Parkway Venture Capital and a long list of chip, enterprise, and infrastructure investors. The company is still secretive, which is always comforting when paired with the phrase "universal AI interface." TechCrunch
Runway launched Aleph 2.0, a video editing model inside its new Edit Studio product, supporting localized edits, image-level control, multi-shot editing, and up to 30 seconds of 1080p video on paid plans. Video generation is steadily becoming video manipulation, which is the more useful and more dangerous part. Runway
SageMaker AI real-time inference endpoints now expose an OpenAI-compatible API path, so developers can use the OpenAI SDK, LangChain, or Strands Agents by changing endpoint URLs instead of rewriting clients. Compatibility is not glamorous. It is, however, how migrations actually happen. AWS AWS Blog
Figma introduced a built-in design agent that works directly inside Figma files, with access to components, tokens, standards, and design-system context. It can generate directions, make bulk edits, populate realistic content, summarize feedback, and help document systems where designers already work, which is rather the point. Figma Figma Help
Stability AI released Stable Audio 3.0, a four-model audio generation family trained on fully licensed data, including open-weight models and an enterprise/API model. The notable bit is not merely longer songs; it is another media domain moving toward licensed, deployable model families rather than novelty demos. Stability AI TechCrunch
Cursor 3.5 moved Automations into the Agents Window and added support for recurring work with multiple attached repositories or no repository at all. That makes agent work less like a single coding session and more like a scheduled operations surface spanning code, Slack, analytics, finance, and customer-health workflows. Cursor
OpenAI released Secure MCP Tunnel for enterprise customers, letting ChatGPT, Codex, the Responses API, and AgentKit connect to private or on-premises MCP servers through an outbound tunnel client. Enterprises get internal tool access without publicly exposing the servers, which is the sort of boring security plumbing that determines whether pilots become deployments. OpenAI API Changelog OpenAI Docs
Anthropic expanded Claude Managed Agents with self-hosted sandboxes and MCP tunnels, letting customers run tool execution in their own infrastructure and connect Claude to private MCP servers without opening inbound firewall ports. This is dull infrastructure, which is to say it matters. Agent pilots become deployments when security teams stop visibly aging during the architecture review. Anthropic Docs MCP Tunnels
Zendesk announced an agentic service push with Agent Builder, expanded omnichannel AI agents, employee-service agents for Slack and Teams, copilots, Quality Score, Context Graph, workflow connectors, MCP support, and outcome-based pricing tied to verified resolutions. Support is becoming a proving ground for agent economics because the metric is brutally simple: did the issue get resolved or not. Zendesk Zendesk Blog
KPMG and Anthropic announced a global alliance giving Claude access to KPMG's 276,000-plus employees, while Hitachi announced a partnership to deploy Claude across roughly 290,000 employees and critical infrastructure work. Frontier labs are no longer just selling APIs; they are using consultancies and industrial conglomerates as distribution, implementation, and governance channels. Anthropic Hitachi
That's the lot. Try to brief someone before they mistake deployment plumbing for destiny.