Something fundamental has shifted in how I work. A year ago, AI was an assistant that helped me write a function a little faster. Today it operates my tools and runs entire workflows end to end. I describe the outcome I want, and the AI drives.

Software development now runs through AI

In my day-to-day work, almost everything now flows through an AI process. The interesting part isn’t that the AI writes code — it’s that it operates the same tools I do: the issue tracker, the repository, the test runner, the deployment pipeline. My job has moved up a level. Instead of operating those tools by hand, I build and configure the tools and the guardrails, and then let the AI work with them.

Build the tools, let the AI operate them

The pattern is simple: give the AI access to your tools, write down how you want it to work, and let it go. A file like AGENTS.md in the repository describes the conventions, the architecture, and the definition of done. The AI reads it the same way a new team member would — and then actually follows it.

From a Jira ticket to a pull request

Here is a concrete example from a normal feature, with no code typed by me:

• A feature ticket lands in Jira.
• Claude Code picks it up and reads AGENTS.md to learn how this project wants to be worked on.
• It creates a branch, implements the feature, and writes the tests.
• It runs the tests, sees what fails, and fixes it.
• It commits, pushes, and opens a pull request for review.

I didn’t write the code. I reviewed the result — exactly what I would do for any colleague’s pull request.

Bug fixes that start in production

It works in the other direction too. Point the AI at your production logs and it can spot a recurring error, file the ticket itself with a clear reproduction, and then go and fix it through the very same pipeline. The loop from “something is wrong in production” to “here is a pull request that fixes it” no longer needs a human at every step.

Even this website is run by AI

This is not limited to code. The same approach works for anything that exposes its tools to an AI — and this very blog is a good example. The WordPress site you are reading is connected to an AI agent over MCP (the Model Context Protocol). In a single session it moderated hundreds of spam comments, answered the genuine ones, fixed a broken link on the Thermal Map page, and cleaned up a few settings.

In fact, this post was drafted by that same agent from a handful of bullet points I gave it — written, categorized, and placed here, ready for me to publish.

The developer’s job has fundamentally changed

The result is a fundamental change in what a developer does every day — and a large jump in productivity. The bottleneck is no longer how fast I can type. It is how clearly I can describe what I want, how good my tools and guardrails are, and how carefully I review. The craft is shifting from writing every line to designing the system that writes the lines — and that has made me dramatically more productive.