Fetching from the wire…
Top 5 · 2026-04-11 · source-backed
Anthropic released Claude Managed Agents in public beta on April 8 and the pricing stopped me mid-scroll. $0.08 per runtime hour. That's roughly $58/month to run an agent 24/7, plus token costs. For context, I've spent more than that on a single debugging session with Claude Code.
The architecture is what caught my attention though. Anthropic published a separate engineering blog post detailing the Brain/Hands/Session separation. Brain handles reasoning. Hands are stateless containers for code execution, provisioned on demand. Session is an append-only event log that enables wake() recovery after crashes. By starting inference immediately from the session log rather than waiting for container provisioning, they hit a 60% TTFT improvement at p50 and over 90% at p95.
Real companies are already shipping on this. Sentry built a fully autonomous bug-to-PR agent. A bug comes in, the agent triages it, writes the fix, opens the PR. Notion lets teams delegate coding and spreadsheet tasks to Claude without leaving their workspace. Asana's CTO says they shipped advanced features "dramatically faster" than before.
This eliminates months of infrastructure work. I've built agent systems from scratch. The sandboxing alone takes weeks. Permissions, state management, error recovery, container orchestration. All of that is now absorbed into the platform. If you're a team that was planning to build agent infrastructure, the build-vs-buy calculus just flipped hard toward buy.
The timing with Osmani's Agent Skills isn't coincidental. You can now define what an agent should do (Agent Skills), deploy it without managing infrastructure (Managed Agents), and pay by the hour instead of burning engineering time on plumbing. The full agentic engineering stack is assembling itself in real time.
For builders: if you've been running agent workloads on raw API calls with your own state management, evaluate whether Managed Agents can replace your infrastructure layer. The $0.08/hr pricing makes the comparison straightforward. Calculate your current infrastructure cost per agent-hour and compare.
Each link below shares sources, entities, or timing with this story.
Anthropic released Claude Managed Agents in public beta on April 8. Hosted infrastructure. Automatic scaling. Sandboxed code execution. Scoped permissions. Persistent sessions with checkpointing. The things you'd need to build yourself if you wanted to run agents in production...
Barry Zhang and Mahesh Murag, the engineers who built Claude Skills at Anthropic, published a talk and engineering post that's gotten 14K+ likes and is reshaping how I think about agent development. The core argument: most agent approaches fail because they lack domain experti...
Anthropic's biggest product event since the Claude 4.6 launch happens tomorrow at 9:30am EST. Expect new Cowork features, live demos of "Claude's newest capabilities," and technical sessions on enterprise agent deployment. The Cowork desktop preview (macOS and Windows) now inc...
DeepClaude hit 470 points on Hacker News. It swaps Claude Code's API backend to DeepSeek V4 Pro while preserving the full agent loop: file editing, bash execution, git tooling, the whole workflow. DeepSeek V4 Pro scores 96.4% on LiveCodeBench at a fraction of Anthropic's prici...
For a month, Claude Code users were convinced the model had been "nerfed." Forums lit up. Conspiracy theories multiplied. People switched tools. Then on April 23, Anthropic did something unusual: they published a detailed post-mortem that named three specific bugs with exact d...
Today's enterprise event in NYC premiered live demos of Claude's newest capabilities and the Cowork enterprise evolution. Self-serve Enterprise (launched Feb 12) bundles Claude + Claude Code + Cowork in a single seat with SSO/SCIM, audit logs, compliance API, and Claude Code S...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.