Simon Willison tests smolvm: hardware-isolated sandbox with 0.6s cold starts
yawnxyz · x · 2026-08-24
Simon Willison tasked Fable 5 in Claude Code for web with stress-testing smolmachines/smolvm as a fast, secure sandbox.
- Verdict: smolvm 1.8.3 is well suited for sandboxing untrusted Python/JS using hardware-isolated VMs rather than shared-kernel containers.
- Verified: offline local images, no-network execution, CPU/RAM limits, guest-enforced timeouts, storage quotas, read-only input mounts, writable output mounts, and --unprivileged all worked as intended.
- Performance: cold starts 0.6–1.5s, warm executions 50ms.
- Twist: the Claude Code container lacked /dev/kvm (no nested virt), so the agent pivoted to a GitHub Actions runner with KVM to run the full test battery — a notable display of agent autonomy in itself.
Related event: Simon Willison Tests smolvm's Hardware-Isolated Sandbox(2 posts)→
More from coding & agent
- Prompt Engineering: Enforce Single-Turn Questions to Prevent Info Overload — dejanseo · 2026-08-24
- More memory made my AI agent worse — a developer's case for write-side memory governance — Many_Audience7660 · 2026-08-24
- Gemini CLI Fix: Prevents Output Inflation on Negative maxChars — Kanika0306 · 2026-08-24
- Day 4 of Cloud Agents Migration: Tackling Constant PR Rebasing — jarrodwatts · 2026-08-24
- Graph Engineering organizes multi-agent systems via dynamic structures — Yuyuan Feng · 2026-08-24
- RecVerse agent simulates realistic e-commerce shopping sessions — Jiakai Tang · 2026-08-24