A3S Lab

Blog

Engineering articles, release notes, and deep dives from the A3S team.

2026-03-05ddddomain-driven-designnestjscqrsevent-sourcingarchitecturetypescript

Domain-Driven Design: Engineering Practice from the Ground Up

This article traces DDD back to its core motivations, combining the Nestify project source code to dissect every key concept — not just "how", but "why".

2026-03-05libkrunlibkrunfwvirtualizationwhpxwindowskvmhvfvirtiotsi

libkrun Deep Dive: Architecture, Modules, and the Windows WHPX Backend

A comprehensive analysis of libkrun's layered architecture, the design principles behind its 13 core modules, and every engineering challenge solved to boot a Linux kernel on Windows

2026-03-01http-402generative-uiai-agentsopenclawpayment-protocolsoftware-economics

The 30-Year Sleeping Standard: HTTP 402, Generative UI, and the Coming Age of Agent-Native Software

HTTP 402 was defined in 1996 and marked "reserved for future use." Thirty years later, the AI agent paradigm is finally making it relevant. This article starts from the ground up to explain how agentic software, machine-negotiated payments, and generative UI are quietly dismantling everything we assumed about how software works.

2026-02-27a3s-codecoding-agentembeddabletypescriptsdk

20 Lines of Code to Build Claude Code's Core Capabilities

A3S Code is an embeddable Rust coding agent framework. Build a terminal assistant in 20 lines, build production-grade Agentic apps with the full SDK.

2026-02-23gatewayllmstreamingautoscalingai-infrastructurerust

Your nginx Is Killing Your AI Service — Why You Need to Redesign the Traffic Layer

Web-era gateways were built on four assumptions. AI services break all four. This article starts from the physical reality of traffic and derives the five capabilities an AI-native gateway must have — and why existing tools cannot provide them.

2026-02-22powerllmteerustinferenceprivacyconfidential-computing

A Privacy LLM Inference Engine That Runs on $10 Hardware

A deep dive into A3S Power's architecture — a privacy inference engine that runs 10B models on $10, 256MB hardware, featuring TEE hardware isolation, model attestation binding, layer-streaming inference, and zero-trust security

2026-02-19boxmicrovmrustteedockerai-agent

A 40MB MicroVM Runtime Written in Rust — A Perfect Docker Replacement for AI Agent Sandboxes

A deep dive into A3S Box architecture — 200ms cold start, 52 Docker-compatible commands, hardware-level isolation, and confidential computing