OfCourse by AgileIdeas.dk
Course outline

Agents That Write Code You Can Live With

Getting started 7 pages · 1 h 1 min
Free taster: the one-hour quality harness 5 pages · 1 h 10 min
Part 1: Habits Coming soon 4 modules · 5 h 45 min
Part 2: Measure and guard Coming soon 2 modules · 2 h 40 min
Part 3: Going deeper Coming soon 3 modules · 4 h 45 min

Overview, about 10 minutes

The course robot climbing a rope that hangs from a big stopwatch

The one-hour quality harness

Building software used to take a lot of effort. But now... We simply get the AI to do a lot of the heavy lifting. And voilà, we're done much faster than we thought possible.

The real problem is, like with my memory, when I see things months later I can't remember why or how I did things. This is where the quality harness comes into play.

In the next hour you will give your agent two things it is missing: the conventions of your codebase, and a way to find its own mistakes. It's a small change, and most teams feel the difference the same day.

Why agent code rots

I am really not good at repetitive work. I tend to miss even obvious mistakes. I simply zone-out and just do it mindlessly.

An agent however doesn't get careless the way a tired developer does. It gets things wrong in more predictable ways:

  • It starts from zero every session. Your team's conventions, the decisions you made last spring and the helper that already exists are all invisible to it unless something tells it.
  • It stops at "looks done". If nothing checks the result, a plausible diff counts as finished. A missing test or a type error only shows up later, when you run into it.
  • It adds instead of changing. An agent that doesn't know a function exists writes a new one. Duplication builds up quietly.
  • It is fast enough to hide the cost. You ship features faster and change the codebase slower, and it takes months before the second part becomes visible.

None of these are fixed by better prompts. They are fixed by the environment the agent works in. Agents amplify whatever feedback loops a codebase already has: a codebase with fast, strict checks gets better code from the same agent than one without.

What you get in this hour

Part Time
The seven layers of a quality harness 15 min
Exercise: a starter AGENTS.md and one check command 30 min
Score your codebase with the self-assessment 10 min

The agent does the coding. You set up what it works in, give it a task and watch what happens. You need a codebase and a coding agent that can run commands in it: your own repository, or the practice codebase if you don't have one you can use. Getting started helps you choose and get set up. The examples use PHP and Laravel. The ideas work for any stack.

Start with the seven layers.

You've marked this page as done.

Kept in this browser. Sign in to keep it on every device.