AI software development tools are fundamentally changing the roles played by software developers in several ways. Software developers are going to be writing less-- or, often, no-- code moving forward. The tools, whether in an IDE or a terminal window, handle writing standard code better and faster than any developer. In this mode, the developer's role is to translate application requirements for the AI agents and make sure that the AI does it's job. Effectively, the AI is a

For a second experiment using an AI agent to build an application, I wanted something more ambitious than a basic app with text in a database. The goal was to create a system that would aggregate files from remote users, automatically pull them into a database, run configurable analysis tools on them, and make everything searchable and viewable with proper access controls. The inspiration came from real life. I sifted through hundreds of photos to create slide shows for three

For my first experiment using an AI agent to code a complete application, I wanted to build something practical—a web-based database app where users could log in, enter information, store it, search through it, view and edit records, and import/export data in various formats. I've used multiple, commercially available databases over the years. They all have strengths and weaknesses. In software engineering, there's a general principle: the best application to write is one tha

Clearly specifying requirements and detailed testing. That's really all you need to do. Want more? Write your specification with an eye towards what needs to be in the software. How you want it to run. What's the look-and-feel. The more specific you are, the more likely you are to get a final package that meets your needs. AI agents are like the Genies in fairy tales. In theory, at least, you always get what you asked for. But not necessarily, what you wanted-- unless you're

The AI agents that write code are all built on something called Large Language Models, or LLMs. If you dig into the technical literature, you'll encounter discussions of vectors, matrices, attention mechanisms, and massive datasets used for training. But let's skip the linear algebra and talk about what this actually means in practice. Building an LLM requires two things: a model architecture and training data. Lots of training data. For software development capabilities, tha

My experience suggests that AI software development works best when you let the AI agent own the entire development process, start to finish. Trying to micromanage or constantly intervene usually just slows things down and introduces inconsistencies. That said, your job as the developer doesn't disappear—it shifts. You're responsible for three things: 1. Give the AI clear ground rules. For Claude, this means writing a detailed CLAUDE.md file. 2. Provide a tight specification.

Let's start with "traditional" software development—though that phrase is about as useful as "traditional soup making." Soup can be meat-based or vegetarian, thin broth or hearty stew, clear consommé or creamy bisque. Software development has just as much variation. But soup, at its core, is a liquid with stuff in it. Usually some fat for richness and salt to bring out the flavors. Everything else is details. Likewise, software development involves some planning—extensive doc