Thinking about AI-assisted software engineering
AI is changing the economics of software implementation.
Our focus is what happens around the code: architecture, system comprehension, verification, engineering rules, multi-team coordination, and the operating model required to use coding agents safely at scale.
Loop engineering · 2026-10-06 · 1 min read
From copilots to loop engineering
The next step after coding copilots is not simply more autonomous coding. It is an explicit engineering loop with defined roles for agents, tools, rules, and people.
Read article: From copilots to loop engineeringArchitecture · 2026-10-06 · 1 min read
AI can write the code. Who owns the architecture?
AI reduces the cost of implementation but does not remove architecture. Organizations need stronger mechanisms to preserve boundaries, ownership, and compatibility.
Read article: AI can write the code. Who owns the architecture?Practice · 2026-10-06 · 1 min read
The Room Technique: separate thinking modes in AI-assisted development
AI-assisted development works better when different forms of engineering work are intentionally separated into explicit working modes, or rooms.
Read article: The Room Technique: separate thinking modes in AI-assisted developmentVerification · 2026-10-06 · 1 min read
Building reliable systems when the AI is not reliable
Production engineering should assume model output can be wrong. Reliability comes from the system around the model: narrow roles, deterministic checks, evidence, and feedback.
Read article: Building reliable systems when the AI is not reliableEngineering risk · 2026-10-06 · 1 min read
Comprehension debt: when software is created faster than teams understand it
Technical debt is not the only risk of faster code generation. Teams may accumulate software that works but is poorly understood by the people who must change, review, debug, and operate it.
Read article: Comprehension debt: when software is created faster than teams understand it
Frequently asked questions
Articles on loop engineering, architecture, software assurance, coding agents, verification, and the operating model needed to use coding agents safely at scale.
The articles are published by Value AI Labs.
The initial articles are drafts that set out each article's argument and outline. They will be expanded over time.