Blog
Ben Monk on software and AI engineering
What an AI engineer actually does
The job is less about models than people expect, and far more about the system around them.
Read
Why AI features fail in production
Most of them do not break for AI reasons. They break for ordinary software reasons nobody was watching for.
Read
Retrieval augmented generation, explained without the diagram
What it is, why it is usually the right first answer, and where it quietly falls apart.
Read
How to evaluate model output without guessing
You cannot assert equality on a model's answer. Here is what to measure instead.
Read
Prompting, fine tuning or retrieval: picking the cheap one first
Three ways to make a model do what you need, and most teams climb the ladder in the wrong order.
Read
Building agents that stay in their lane
Loops, drift and unbounded cost. Every agent failure mode is a design decision, not a model limitation.
Read
The real cost of a long context window
Big windows made it easy to stop thinking about what goes in the prompt. That has a price.
Read
The software habits that survive AI
Assisted coding changed how fast code appears. It did not change what keeps a system alive.
Read