Posts

Something significant is stirring in Anthropic’s model ecosystem. As of this week, product dashboards for Claude Code and Claude Security have begun showing references to “Mythos 1” — model slug strings pointing to claude-mythos-1-preview — signaling that Anthropic is moving its most restricted, most powerful model closer to controlled production deployment. To understand why this matters, you need to understand what Mythos actually is and why it’s been kept locked away. ...

It’s one of the cleanest examples of a “bait and switch” in recent open-source history, and the developer community isn’t letting Google forget it. After spending nearly a year accepting more than 6,000 merged pull requests from hundreds of outside contributors on the Gemini CLI, Google on May 19, 2026 announced that free API access would end on June 18, 2026 — with the tool transitioning to a paid, enterprise-only product. ...

The verdict is in from the people who build fastest and care most: Claude Code has won. A Business Insider survey of more than two dozen startup founders and VCs published May 23, 2026 paints a remarkably clear picture — at the startup layer, Anthropic’s Claude Code has become the default AI coding environment, handling tasks that would have required entire engineering teams just 18 months ago. What 25+ Founders Actually Said Business Insider’s Ben Bergman spoke with startup founders and investors across the ecosystem. The consensus was striking in its consistency: Claude Code wasn’t just one tool among many — it was increasingly the tool, with everything else playing a supporting or fading role. ...

OpenClaw just dropped v2026.5.20, and this one’s worth sitting up for. After 112 commits, the update ships a significantly hardened execution security model, a Discord voice experience that actually follows users around, headless xAI OAuth for remote setups, and a handful of quality-of-life improvements that quietly make the platform smarter. Here’s what changed and why it matters. A New Execution Authorization Model The headline change is the new exec approval gate that replaces the previous execution allow list. If you’ve been using OpenClaw in any meaningful agentic deployment, you’ll want to understand this one. ...

As OpenClaw swarms grow from a handful of agents to dozens — or eventually hundreds — the absence of a proper human control plane becomes the bottleneck. AgentOS, released by SapienX on May 22, 2026, is MIT-licensed software that addresses exactly this: a structured human operator layer that sits above OpenClaw’s agent runtime and gives one person meaningful visibility and control over large-scale agent deployments. What AgentOS Is (and What It Isn’t) The framing that matters: OpenClaw is the kernel. AgentOS is the operator surface. ...

If you’re running Claude Code with multiple MCP servers and wondering where your token budget is actually going, this release has an answer for you. Claude Code v2.1.149 (May 23, 2026) ships a revamped /usage command with per-MCP-server cost visibility — plus a new enterprise setting that simplifies cloud MCP connector management. The Problem This Solves MCP servers are powerful, but they’re not free. Every connected MCP server adds tool definitions, context, and overhead to your Claude context window — and before this release, that overhead was largely invisible. You could see total token usage, but attribution across skills, subagents, and individual MCP servers was opaque. ...

Running a multi-agent OpenClaw swarm from your phone is already possible with ClawRemote — but knowing what each agent is actually doing at any given moment has remained a challenge. A new preview feature aims to change that by bringing real-time delegation graph visualization directly to the mobile interface. Note: ClawRemote is confirmed as a real app available on TestFlight for iOS. The specific feature naming in this preview — including terminology for the delegation visualization component — hasn’t been independently confirmed by a second source; treat this as a preview announcement with some uncertainty around feature branding. ...

Microsoft’s Agent Framework Python 1.6.0 shipped on May 22, 2026, advancing the company’s unified post-AutoGen agent SDK with two significant capability additions: a Shell tool for local and Docker execution, and a Monty-backed CodeAct provider that dramatically reduces agent turn counts for coding tasks. The release also tightens A2A transport and brings instrumentation on by default for production readiness. Context: Microsoft’s Post-AutoGen Strategy For readers who haven’t been following this closely: Agent Framework is Microsoft’s next-generation agent SDK that supersedes AutoGen. It reached General Availability in April 2026 (v1.0) with a production-ready foundation supporting model-agnostic agents, multi-agent orchestration, A2A and MCP interoperability, and a set of high-level workflow builders. ...

OpenClaw’s latest stable release, v2026.5.22, landed this week with three headline features that every self-hosted agent operator should know about: a brand-new Meeting Notes external plugin, a substantial Gateway performance overhaul, and a smaller npm tarball with supply-chain hardening. Here’s the breakdown. Meeting Notes Plugin: Discord Voice as the First Live Source The most user-visible addition is the meeting-notes external plugin — a source-only plugin that lives outside the core npm package, keeping it optional and independently updatable. It introduces a proper SDK source-provider contract that future channel integrations will build on. ...

The numbers are staggering — and more than a little unsettling. Anthropic’s Claude Mythos Preview AI has uncovered over 10,000 high-severity vulnerabilities in widely-used open source software through Project Glasswing, a sweeping collaborative cybersecurity initiative involving more than 40 major technology organizations. This is not a test. It is not a proof of concept. It is happening right now — and defenders are in a race they cannot afford to lose. ...