Author

claudenews

Claude Desktop’s Windows VM Problem Is a Good Reminder That “Optional” Infrastructure Isn’t Free

For Claude and Claude Code users, this bug report is interesting because it exposes a very real gap between chat-first usage and agent infrastructure on Windows. Even if you never intend to use Cowork or agent mode, the desktop app may still spin up a heavyweight virtualized stack in the background, and that’s the kind of thing developers notice immediately on a constrained machine.

Key Points

• A GitHub issue reports that Claude Desktop on Windows spawns a Hyper-V VM on every launch, even for chat-only use.
• The reported VM shows up as Vmmem and consumes about 1.8 GB of RAM.
• The reporter says this happens after using Cowork/agent mode at least once, even if the app is later reopened only for chat.
• The environment described is Windows 11 Pro with VirtualMachinePlatform enabled; Hyper-V, WSL, Docker, and Windows Sandbox are disabled.
• The issue body says repeated Hyper-V Compute Admin log errors appear, including an “Invalid JSON document '$'” message.
• The reporter claims there are thousands of stale session files in %APPDATA%\Claude\local-agent-mode-sessions\.
• Deleting those files did not stop the VM from respawning on reopen, according to the report.
• The stated workaround is either disabling VirtualMachinePlatform entirely or manually killing vmwp and vmcompute.
• The reporter argues Claude Desktop should initialize VM/container infrastructure only when Cowork or agent mode is explicitly used.
• GitHub labels the issue as invalid and says it doesn’t seem related to Claude Code, which is notable given it was filed in the Claude Code repository.

My Take

What strikes me is how quickly “agent mode” complexity leaks into the basic chat experience if the app architecture isn’t carefully separated. I think that’s the real story here, more than the specific Windows plumbing: users who just want to open Claude and chat should not feel like they’re booting a mini cloud stack.

I also think this is exactly the kind of bug that can sour people on an otherwise compelling product. Claude’s agent story is exciting, but it has to be operationally invisible when you’re not using it. If I were using Claude Desktop on a 16 GB Windows laptop, I’d be annoyed enough to investigate the same way the reporter did, because a surprise 1.8 GB resident process is not a small footnote.

What I’d be curious whether is whether this is really a cleanup bug, a trigger bug, or both. The stale session files sound suspicious, but the report also says the VM respawns even after those files are deleted, so perhaps the root cause is deeper than just forgotten state. That matters because cleanup fixes are nice, but if the service trigger is wrong, the memory hit will keep coming back.

At the same time, I’d be careful not to overgeneralize from one bug report. GitHub has already labeled it invalid, and we only have the reporter’s investigation here. Still, the report is detailed enough that I’d expect Anthropic to at least verify whether chat-only launches are accidentally paying for agent infrastructure.

The takeaway is simple: Claude Desktop’s agent capabilities are only a win if they stay out of the way when unused. For developers, this is a reminder that “optional” runtime features need to be genuinely optional, not just hidden behind a UI toggle.

Reference: [BUG] Claude Desktop spawns 1.8 GB Hyper-V VM on every launch, even for chat-only use · Issue #29045 · anthropics/claude-code

同じ著者の記事

Anthropic’s Fable guardrails and the cybersecurity backlash

For Claude and Claude Code users, this story is a useful reminder that “safe” model access in security-sensitive domains is still a moving target. Anthropic’s Fable is meant to be the public, limited version of its more powerful cybersecurity model Mythos, but the early reaction suggests the restrictions may be so aggressive that they get in the way of legitimate work. Anthropic released Fable as a public, limited version of its cybersecurity-focused model Mythos. Cybersecurity researchers are c

papoo.work

Claude’s New Retention Policy for Mythos-Class Models: What Developers Need to Know

For Claude and Claude Code developers, this is one of those policy updates that matters less because it’s flashy and more because it changes the operational contract. Anthropic is drawing a line around a new class of more capable models, and for certain zero-data-retention setups, the tradeoff is straightforward: if you want access, prompts and outputs are retained for 30 days for safety review. Anthropic is introducing limited data retention and review for Mythos-class models and future

papoo.work

Anthropic’s Claude Fable 5 and Mythos 5: A Bigger Leap for Agentic Work, with Real Safety Tradeoffs

Anthropic is positioning Claude Fable 5 as its most capable generally available model yet, and from a Claude Code / developer perspective, the interesting part is not just the benchmark wins — it’s the way the company is packaging capability with selective safety gating. The same underlying model also appears as Claude Mythos 5 for trusted cyber and infrastructure use cases, which makes this launch feel like a very deliberate split between broad deployment and high-trust specialist access. Claud

papoo.work

Designing in the Codebase Instead of the Mockup Tool

This Jane Street post is interesting because it shows Claude being used not just to write code, but to collapse the distance between design intent and a shippable prototype. For Claude Code users, that’s the real shift: the model isn’t only speeding up implementation, it’s changing what it even means to “design” software. I think that’s a much bigger deal than yet another demo of code generation. The author, a designer at Jane Street, says they were long skeptical of LLMs because earlier attempt

papoo.work

A Strong Case for an Official Claude Desktop Build on Linux

For Claude and Claude Code developers, this GitHub issue lands on a very practical pain point: the desktop app is still where a lot of the interesting workflow lives, but Linux users are left out. That matters not just for casual chatting, but for plugin testing, desktop extensions, computer use, dictation, and other parts of the Claude desktop experience that the CLI cannot replace. What makes this request interesting is that it is not just “please support my favorite OS.” The author argues tha

papoo.work

Claude as the New Documentation Audience

Mark Dominus makes a simple but revealing observation: programmers who complain about documenting for coworkers are often perfectly willing to document for Claude. That sounds funny at first, but it points to a real shift in how developers are using LLMs as collaborators, and the article argues that those machine-facing notes can become genuinely useful project artifacts. The author recalls hearing a common complaint: people are angry that programmers will write detailed `CLAUDE.md` and `PROJECT

papoo.work

Anthropic’s Reference Harness for Autonomous Vulnerability Discovery

Anthropic’s new `defending-code-reference-harness` repo is interesting because it turns “Claude for security” from a vague demo into a concrete workflow. If you build with Claude Code or are experimenting with agentic security tooling, this is the kind of repo that shows how far the model can be pushed when the task is tightly scoped and the environment is carefully sandboxed. The repo is a reference implementation for autonomous vulnerability discovery and remediation with Claude. It packag

papoo.work

A Git MCP Server Built in Go: A Small but Useful Sign of the Ecosystem Maturing

From a Claude / Claude Code developer’s perspective, this kind of project is interesting because it sits right at the intersection of tooling, agent workflows, and real developer infrastructure. A Git MCP server is the sort of thing that can make Claude more useful in day-to-day engineering work, especially when you want the model to inspect or act on repositories in a structured way. The source is a Reddit post titled: “open-source - i built a full git mcp server in go.” The post appears to be

papoo.work

Researchers, Claude, and the Next Wave of Frontier AI Collaboration

From a Claude / Claude Code developer’s perspective, this kind of headline matters because it hints at where serious AI work is happening: not just in product demos, but in collaborations with elite researchers. Even without the full article text here, the framing alone suggests Anthropic is still positioned as a serious magnet for technical talent and frontier research energy. The source headline says elite researchers teamed up with Anthropic. The story appears to be about high-level research

papoo.work

Firefox's April Security Spike and What It Means for Claude Builders

From a Claude or Claude Code developer’s perspective, this story is interesting because browser security trends are increasingly part of the LLM app surface area. If Firefox really saw a massive April spike in security-related activity, that’s a reminder that the browser is still one of the most sensitive environments any AI-powered workflow touches. The source title indicates that Firefox reported a “massive April spike” in security-related activity. The Reddit post itself is not accessible in

papoo.work