blog
Blog — page 38 of 39
Posts 889–912 of 928, newest first.
Serverless on Firecracker: How FaaS Really Works
Every "serverless" function is somebody's untrusted code running next to a stranger's untrusted code on shared hardware. That's a containment problem, and the industry's answer was microVMs. Here's how FaaS is really built — and why snapshot-restore is the part that matters.
Ajay KumarDatabase-per-Tenant Isolation with MicroVMs
One tenant's runaway query shouldn't become everyone's incident. A tour of the multi-tenancy spectrum — shared schema, schema-per-tenant, database-per-tenant, VM-per-tenant — when full per-tenant DB isolation is actually worth it, and how a microVM makes it a clean boundary.
Ajay KumarBrowser Isolation in MicroVMs: Headless & Remote
A browser is a giant attack surface that executes arbitrary remote JavaScript — and now the model picks which links to click. Here's why one browser per disposable microVM is the clean answer for remote isolation and agent web automation.
Ajay KumarReproducible Builds in Disposable MicroVMs
A long-lived CI runner is a shared apartment where every previous tenant left something in the fridge. A fresh microVM per build is a hermetic, reproducible alternative.
Ajay KumarSandboxing LLM Tool Calls and MCP Servers
A tool call is the model reaching out of the chat and touching your systems. Here's how to box in function calling, MCP servers, and computer use so a hijacked call destroys a throwaway VM, not your infra.
Ajay KumarMicroVM vs VM vs Container: A 2026 Comparison
A traditional VM gives you strong isolation and a heavy bill. A container gives you density and a shared kernel you're trusting with your life. A microVM is the third option — VM-grade isolation at near-container speed. Here's the honest three-way.
Ajay KumarFirecracker Networking Explained: TAP, netns, NAT
The guest thinks it has a normal network card. It has a polite illusion of one. Here's the full path — virtio-net to TAP to a per-sandbox network namespace to NAT'd egress — and why isolating the network per sandbox is the difference between a sandbox and a liability.
Ajay KumargVisor vs Firecracker: Which Isolation for Agents?
Both are real steps up from a plain container — but a software syscall interceptor and a hardware-virtualized microVM are not the same boundary. Here's the fair, honest comparison for AI-agent workloads.
Ajay KumarHow to Build a Sandboxed AI Coding Agent (2026)
Your agent will eventually run code it wrote at 3am with the confidence of a junior engineer and the judgment of a dice roll. Here's how to build the loop so that's fine.
Ajay KumarThaw: How We Made a Cold Start Take 164ms
Vercel Fluid pre-warms and shares instances so you rarely hit a cold start. We took the other path: scale fully to zero, then thaw a frozen microVM in ~164ms. Here's how.
Ajay KumarBest Code Execution Sandboxes for AI Agents (2026)
What a code execution sandbox is, and how PandaStack, E2B, Modal, Daytona, Vercel Sandbox, Northflank, Blaxel, Cloudflare and Fly.io Sprites compare on isolation, self-hosting and persistence, from each vendor's own docs.
Ajay KumarBest Open-Source Sandboxes for Running Untrusted Code
The honest set of open-source, self-hostable options for isolating untrusted code execution — characterized by license and isolation model, not a leaderboard.
Ajay KumarSelf-Hosted Code Execution Sandbox for Production AI
Why self-host a code execution sandbox — data residency, VPC isolation, cost at scale — what it actually takes to run, and the honest cases where hosted wins.
Ajay KumarThe Real Cost of Hosted Sandboxes at Scale
Hosted sandboxes are cheap until they aren't. The cost structure of per-second sandboxes — idle, per-create overhead, egress — and where the math flips toward self-hosting.
Ajay KumarCopy-on-Write Rootfs: Why MicroVM Create Is O(metadata)
Cloning a multi-gigabyte rootfs sounds expensive. With a copy-on-write filesystem it isn't: XFS reflink makes a clone that shares data extents until a write copies a single block, so the clone is O(metadata), not O(data). Here's the mechanism, and why ext4 can't do it.
Ajay Kumaruserfaultfd: Lazy Memory for Instant VM Restore
userfaultfd hands a page fault to your own code instead of the kernel resolving it. That one inversion is what lets a microVM resume before its whole memory image has arrived — PandaStack uses it to stream vm.mem from object storage on restore.
Ajay KumarHow PandaStack Creates a MicroVM in Under 200ms
A cold Firecracker boot is genuinely fast — minimal device model, no firmware, a stripped kernel. But PandaStack's 179ms create isn't a boot at all: it restores a frozen machine's memory and device state and resumes. Here's the difference, with real numbers.
Ajay KumarWebAssembly vs MicroVMs for Sandboxed Code
WASM isolates one program's memory behind a software boundary; a microVM isolates a whole OS behind a hardware one. They're more complementary than competing — here's how to choose, honestly.
Ajay KumarCode Interpreter API Pricing, Compared
A model-by-model map of code interpreter API pricing — no invented competitor numbers. What you actually pay for, and the self-host break-even at scale.
Ajay KumarStop an AI Agent Touching the Host Filesystem & Network
An agent should not read your host files, reach your internal services, or quietly exfiltrate. Here's the per-sandbox boundary that enforces that — own kernel, own filesystem, own network namespace — and the parts you still have to configure yourself.
Ajay KumarHow to Sandbox Untrusted & AI-Generated Code
You have to run code you don't trust — user submissions, CI, and now LLM output — without it owning your host or your other tenants. This is the decision framework: what threat you're defending against, what isolation actually holds, and where each option fits.
Ajay KumarWhy Docker Isn't a Sandbox
A container is excellent isolation for software that cooperates with you. It is not a security boundary against software that's actively trying to break out — because it shares the one thing it's supposed to be protected from: the host kernel.
Ajay KumarThe Code Isolation Hierarchy
Bare process, container, gVisor, Kata, microVM, confidential VM — a clean walk up the isolation ladder, with each rung's real threat model and overhead, and why higher isn't automatically better.
Ajay KumarMulti-Tenant Code Execution: Isolation Requirements
If you run other people's code on shared hosts, the requirements aren't a wishlist — they're the line between a platform and a breach. Hardware isolation between tenants, network isolation, resource limits, ephemerality, and blast-radius containment, with honest notes on residual risk.
Ajay Kumar