Welcome to Minimalizt Dev

Latest Posts

I gave five AI models the same product engineering task (aka the slop evals)

I gave five AI models the same product engineering task (aka the slop evals)

A home experiment comparing GPT-5.6 Sol, Fable 5, Kimi K3, GLM-5.2 and DeepSeek V4 Pro on code quality, time and cost.

Code is cheap. Restraint is priceless.

Code is cheap. Restraint is priceless.

With AI, code is cheap. You can build any feature in no time. But knowing what NOT to build may be more important for product success.

Principles for Building AI-Native Web Applications

Principles for Building AI-Native Web Applications

As the web development landscape is shifting, most apps get AI features bolted on top of their existing architectures (with varying degrees of success). If I had to build a truly AI-native web app today, based on my learnings over the past 3 years, these are the principles I would follow.