claude performance-issues developer-tools ai-agents

Anthropic Cowork feature creates 10GB VM bundle on macOS without warning

Cowork feature creates 10GB VM bundle that severely degrades performance · Issue #22543 · anthropics/claude-code

github.com

March 2, 2026

1 min read

🔥🔥🔥🔥🔥

64/100

Summary

The cowork feature in Claude Desktop creates a 10GB VM bundle that leads to significant performance degradation, including slow startup, UI lag, and slow responses. The VM bundle file, located at ~/Library/Application Support/Claude/vm_bundles/claudevm.bundle/rootfs.img, grows to 10GB and is not cleaned up, causing performance issues to worsen over time during a single session.

Key Takeaways

The cowork feature in Claude Desktop creates a 10GB VM bundle that significantly slows down performance, causing slow startup, UI lag, and delayed responses.
Performance degrades over time even after cleaning up the VM bundle, with CPU usage increasing from 24% at idle to 55% after several minutes of use.
Deleting the VM bundle and cache provides a temporary performance improvement of approximately 75%, but the degradation recurs shortly after.
The expected behavior includes stable CPU usage without degradation and automatic cleanup of VM bundles after cowork sessions.

Read original article

View story →

Coverage & discussions

Hacker News · github.com

Anthropic Cowork feature creates 10GB VM bundle on macOS without warning ↗

Community discussions

Hacker News discussion ↗

Historical count: 5 · 5 saved in sample · Historical sample

Analysis ready

Community Sentiment

Mixed

Positives

The Claude Cowork feature leverages a VM to provide a dedicated environment for AI agents, enhancing their ability to write custom scripts and solve user problems effectively.
Using a VM for AI development allows separation from the user's local environment, which can lead to improved safety and performance in executing tasks.

Concerns

There are concerns that the quality of AI-generated products may decline as reliance on AI tools increases, potentially diminishing the role of professional developers.
The lack of effective storage management in applications leads to excessive disk usage, which can hinder user experience and efficiency.