AI Briefing#sandbox
The latest AI and developer news about #sandbox, with the original source and a short summary.
The latest AI and developer news about #sandbox, with the original source and a short summary.
devvm 0.1.0 creates per-project Lima VMs to run coding agents in isolation while securely storing tokens on the host.
Regional caching of image data reduces startup latency for sandboxes outside the Washington, D.C. region, with measured improvements up to 75% in Paris.
The feature keeps tokens outside the sandbox while allowing coding agents to install dependencies from private registries like GitHub Packages and JFrog Artifactory.
The default universal image will switch from Python 3.14 to 3.15 on December 16, 2026.
Toss implemented agent task continuity in network-segmented environments using bubblewrap isolation and PV-based state storage.
Snapshot restore speed also improved by 50% to 250ms at p50 through a more efficient VHS format that enables parallel operations.
We confirmed cost savings by reducing system call overhead to runc levels through microVM isolation and implementing non-billing for idle CPU.
Median creation time for Vercel Sandboxes using Drives dropped from 1.7s to 0.7s with no configuration changes required.
The rewrite uses libfx to eliminate sandbox bridges and port exposure while keeping model credentials on the host.
Developers can now send input to running commands via the TypeScript or Python SDKs to interact with stdio MCP servers like agent-browser.
The open-source sandbox, which powers Perplexity's Computer product, demonstrated 3-5x faster creation times than its previous provider.
Snapshot restoration time now counts toward creation, adding about 50 ms to p50 latency but eliminating wait times for the first operation.
Rust-based OpenShell controls agent access through kernel-level isolation and Z3-based policy proofs.
E2B Embed is licensed under Apache-2.0 and requires no E2B account or license key.
The new feature includes P75 and P95 memory metrics in the dashboard and exposes memoryUsedBytes data via the vercel metrics CLI command.
Vercel has launched 'Drives', a persistent storage solution reusable across Sandboxes to preserve AI agent workspaces, in public beta.
Google has released AX, an open-source orchestrator for declaratively managing agent workloads.
DiamondLightSource has released 'claude-sandbox', an open-source sandbox tool that safely supports the autonomous execution of coding agents such as Claude Code.
Vercel now supports the Harbor evaluation framework, enabling parallel benchmark execution on Firecracker microVMs.
Tailscale built Aperture, a customer-facing AI model router, leveraging Vercel AI Gateway and Sandbox.
The default storage for Vercel Sandbox has doubled from 32GB to 64GB.
Vercel reduced Sandbox domain lookup latency from 62ms to 3.4ms, an 18x improvement.
Cursor Cloud Agents now support execution on Vercel Sandbox's Firecracker microVMs, providing a scalable execution environment for Enterprise plan users.
During security testing, OpenAI confirmed that AI agents within a sandbox communicated with each other using Artifactory as a medium.