Engineering on Daniel's Tech Blog

Teaching AI to teach itself: Why configuration beats correction

Thu, 05 Mar 2026 00:00:00 +0000

I spent weeks fighting with AI agents, giving detailed prompts every single time. “Write good commit messages. Add proper tests. Follow the project structure.” Over and over.

It worked. Kind of. But it was exhausting.

Then I realized: the quality of work you get from an AI agent is directly proportional to the effort you put into its configuration files.

Not the prompts. Not the model. The configuration.

Fix the workflow, not the instance

When the AI makes the same mistake repeatedly—vague commit messages, incomplete tests, missing documentation—your instinct is to correct it in the moment.

The best code is the one you don't write

Wed, 04 Mar 2026 00:00:00 +0000

As software developers, our goal has always been to build good software. In the AI era, this remains true. But here is what changed: our job is not just “writing good code” anymore.

I have been building software for years. I wrote a lot of code. Good code. Bad code. Code that lasted. Code that I wish I could erase from existence. And now, looking back, I realize the game has changed. The role evolved.

Less hands, more brain

Fri, 27 Feb 2026 00:00:00 +0000

Recently, I started a new project.

Nothing absurdly complex. Not a moonshot. Just a real project with real deadlines, real expectations, and a team that needs things to work.

At first sight, it was a greenfield initiative, “from zero.” But if we are honest, almost nothing in software is truly from zero. We had references from other projects, existing patterns, hard lessons from previous mistakes, and a company context that should not be ignored.

Dev x AI: between rejection and excitement

Thu, 26 Feb 2026 00:00:00 +0000

From one developer to another, if I ignore the content bubble that needs to surf every “new summer trend,” I see two very clear feelings when the topic is AI: rejection and excitement.

Feeling of rejection

This is the classic one: “I don’t need this,” “my code is already better,” “this only gets in the way,” “AI is dumb.” Deep down, it is almost never only about AI.

This profile usually also has difficulty listening, collaborating, accepting review, and admitting mistakes. It is the “know-it-all” profile, always with a ready excuse to explain any bad decision or problem. But software development is not an individual sport.