Stop Learning AI. Start Using It.
The best way to learn AI is to skip the learning and solve a real business problem with it. Here's a concrete action plan.
MaGong
Insights on AI, coding, and tech from a software engineer perspective
The Vertical SaaS Moat That Might Be Empty
A popular argument says AI startups should hide in niches too small for big companies. But what if those niches are barren? Construction SaaS data tells a sober
How History Teaches AI Scaling: Lessons from Zhu Rui
When AI projects stall at scale-up, we blame budgets and leadership. But history shows the real problem is courage. Here's how a Chinese general scaled artillery from zero to decisiveness in three years—and what that teaches us about AI adoption in 2026.
Configuration Management: The Silent Technical Debt
Most teams treat configuration management as an afterthought, but it's one of the most overlooked sources of production chaos. Here's why the industry still hasn't figured it out.
Music and Affective Computing: Programming Human Emotions
Agent Management Forum Episode 14 — exploring whether music is a form of emotional programming, and what affective computing can learn from it.
Ensuring AI Code Quality: Lessons from Statistical Process Control
What a 1924 factory quality method teaches us about controlling AI code — and why effective constraint equals constraint power times independence.
The Missing Role: Why Software Needs a Code Supervisor
Civil engineering has independent supervisors with veto power. Software doesn't. AI finally makes this role affordable.
Testing: AI Coding's Last Unsolved Problem
Five questions about testing in the AI coding era that nobody could answer at our engineering forum — and why at least three of them won't be solved by better m
Beyond the Alpha Wolf: Building AI-Native Dev Teams That Scale
The alpha-wolf model of AI coding works brilliantly — if you can find a unicorn. Product Tri-Ownership offers a replicable alternative.
Forward Deployed Engineer: Silicon Valley's Rebranded Delivery Engineer
Silicon Valley is hyping Forward Deployed Engineers as innovation. I've seen this movie before — it's called zhongtai.
Alibaba's Qwen Code: A Hasty Fork
Alibaba's qwen-code CLI is a fork of Google's Gemini CLI. The licence is Apache 2.0, so legally there's nothing wrong with this. But legality aside, the problem
Iterative Methodology for Industrial-Quality AI Software
I keep hearing people complain that AI-written code is unusable. I think the reason is simple: they're not applying iteration as a methodology. Iteration means ...
Treat AI Like Humans, Not Software
When AI fails to meet our expectations, we often blame it for being "buggy" or "unreliable". But perhaps the problem isn't with AI - it's with our mental model.
The Core Problems of AI Coding
After five months of building production systems with AI coding tools, I've identified three core problems that actually matter. This post focuses on ensuring q
The Burnout Effect: Treating AI as Human (Episode 2)
AI gets tired. Not physically, but functionally. Understanding this burnout effect is key to working effectively with LLMs.
Pseudo-Problems and Core Problems in AI Coding
As an AI coding enthusiast, I've noticed many smart people spending time on pseudo-problems. Here's what actually matters.