How I build with AI

September 2026

I use AI tools as part of learning and building software. What interests me is the whole process: understanding a problem, making something small, seeing where it breaks and improving it.

Start with something concrete

A small goal is easier to reason about. Read a file. Find relevant information. Display a result. Each step should have an input, an expected outcome and a way to check it.

My personal projects explore different parts of this process: a Turkish model connected to a Linux tool, an assistant that searches a knowledge base, and experiments with web interfaces.

Use the tools, understand the result

Coding assistants help me explore alternatives and write an initial version. I still need to read the changes, understand the assumptions and check the result. A generated answer or a successful build is one piece of evidence, not the whole picture.

For an interface, that means opening it and trying it on different screen sizes. For a model, it means testing examples and looking at the failures as well as the successes.

Keep learning visible

I like small, reviewable steps. Notes help me remember why a decision was made; version history helps me revisit it. Sharing source code makes the result easier to inspect and improve.

That is what this site is for: a place for my projects, experiments and things I’m learning. The daily Research section is collected automatically by an AI agent; these personal notes are a separate part of the site.