Sub-millisecond VM sandboxes using CoW memory forking
I wanted to see how fast an isolated code sandbox could start if I never had to boot a fresh VM. So instead of launching a new microVM per execution, I boot Firecracker once with Python and numpy already loaded, then snapshot the full VM state. Every execution after that creates a new KVM VM backed by a `MAP_PRIVATE` mapping of the snapshot memory, so Linux gives me copy-on-write pages automatically. That means each sandbox starts from an already-running Python process inside a real VM, runs the code, and exits. These are real KVM VMs, not containers: separate guest kernel, separate guest…
Sub-millisecond VM sandboxes using CoW memory forking launched on March 17, 2026 with 311 votes, #77 of 23,380 launches that month and more than 99% of that year's launches.
In plain words
This project demonstrates a method for creating isolated code sandboxes that start in sub-millisecond time using Firecracker microVMs. Instead of booting a fresh VM for each execution, it snapshots a pre-loaded VM with Python and numpy, then uses copy-on-write memory forking to spawn new isolated KVM instances that inherit the snapshot state. Each sandbox runs as a complete virtual machine with its own kernel and memory space, with pages copied only when modified. The implementation is written in Rust and released under Apache 2.0.
written from the facts on this page · September 2026
From the sources
Sub-millisecond VM sandboxes for AI agents via copy-on-write forking - zerobootdev/zeroboot
curl -X POST https://api.zeroboot.dev/v1/exec \ -H ' Content-Type: application/json ' \ -H ' Authorization: Bearer zb_demo_hn2026 ' \ -d ' {"code":"import numpy as np; print(np.random.rand(3))"} ' Benchmarks Metric Zeroboot E2B microsandbox Daytona Spawn latency p50 0.79ms ~150ms ~200ms ~27ms Spawn latency p99 1.74ms ~300ms ~400ms ~90ms Memory per sandbox ~265KB ~128MB ~50MB ~50MB Fork + exec (Python) ~8ms - - - 1000 concurrent forks 815ms - - - Each sandbox is a real KVM virtual machine with hardware-enforced memory isolation. Firecracker snapshot ──► mmap(MAP_PRIVATE) ──► KVM VM + restored CPU state (copy-on-write) (~0.8ms) Template (one-time): Firecracker boots a VM, pre-loads your…from github.com
In the maker’s words, at launch
I wanted to see how fast an isolated code sandbox could start if I never had to boot a fresh VM. So instead of launching a new microVM per execution, I boot Firecracker once with Python and numpy already loaded, then snapshot the full VM state. Every execution after that creates a new KVM VM backed by a `MAP_PRIVATE`…
Does a similar job
all alternatives →
Tiny VM sandbox in C with apps in Rust, C and ZigDec 2025 · github.com · ▲196Minimalist, dependency-free virtual machine sandbox for microcontrollers and other resource-constrained devices. Single C file, no dynamic memory allocations, asynchronous design, pure C99 - ringtailsoftware/uvm32
Smol machines – subsecond coldstart, portable virtual machinesApr 2026 · github.com · ▲499Branchable computing by using a portable, lightweight, self-contained virtual machine - smol-machines/smolvm
Ephemeral VMs in 1 Microsecond2024 · github.com · ▲84Contribute to libriscv/drogon-sandbox development by creating an account on GitHub.
AVR-VM – VM with JIT-compiler for ATMega32 written in Rust2017 · github.com · ▲78VM with JIT-compiler for ATMega32 written in Rust. Contribute to MackieLoeffel/avr-vm development by creating an account on GitHub.
npm i -g hotcellAug 2026 · github.com · ▲97Hi HN! Sina here. i wanted to share this open source project (apache 2.0) that i've been working on for the past month or so. it's called hotcell and it lets you create/pause/manage sandboxes on any device (your laptop, linux vm, bare metal). i've tested it against various benchmarks that i found on computesdk (https://www.computesdk.com/benchmarks/sandboxes/) as well as the one shared by dax (cofounder of open code) on twitter recently (link to his original thread and my reply https://x.com/sinasanm/status/2078158598996443346) and…
Clone, a small Rust VMM, forks VMs in under 20ms via CoWApr 2026 · github.com · ▲11We needed a secure, multi-tenant way to offer shell accounts to users, but most VMMs were using too much memory and containers are unsafe. With clone, VMs are now more memory efficient than containers in most cases. Since many other projects on HN looked like they were doing this too, open sourcing this was the right thing to do. Feel free to use in whole or in part as you see fit!
More dev tools this month
the category →
open-source IDE for thoughtful software design. Contribute to devdotfast/whiteboard development by creating an account on GitHub.
Dev tools · 27d ago · github.com
Contribute to reladraw/reladraw development by creating an account on GitHub.
Dev tools · 25d ago · github.com
OpenBot▲362Grok Bot alternative: free, local, open-source, multiplayer
Dev tools · 22d ago · openbot.run
Talanoa▲341Hi HN, I’m a fullstack developer, and honestly, I’ve lost count of how many side projects I’ve worked on over the years. But one idea kept coming back to me: a different kind of email client. It first came to me in 2014: “What if we could handle emails more like conversations?” Every year, I’d revisit this concept in my “Build Notes.” Ten years later, no one had built exactly what I had in mind — so I finally decided to do it myself. That’s how Talanoa was born. I wanted something focused on people, not timelines. I was tired of marking messages as unread just to remember to reply later, or…
Dev tools · 20d ago · talanoa.email
CoreSpeed▲332One MCP for everything your agents need: apps, memory, tools
Dev tools · 28d ago · corespeed.io
Source: Show HN post ↗ · launch text quoted in part
Launched alongside, March 2026
the whole month →
Switch from ChatGPT to Claude with import memory feature
AI · Mar 2026 · claude.com
Littlebird▲709Littlebird remembers everything across your meetings, messages, docs, and apps, then takes action so you can recall any detail and never lose context.
AI · Mar 2026 · littlebird.ai
