About this site

This site is where I share my work across AI systems, experiments, and technical research.

My name is Can Dalgir, and I build and explore ideas across language models, agent infrastructure, compression, evaluation, and practical AI products. Some of that work turns into articles, some into prototypes, and some into deeper technical experiments like Parameter Golf.

I created this site as a place to publish what I’m building, what I’m learning, and how I think through hard technical problems. Rather than just posting outcomes, I want this to be a record of the process, the tradeoffs, failed attempts, architecture choices, and the reasoning behind what worked.

What you'll find here

Technical writing

Breakdowns of AI systems, modeling ideas, infrastructure decisions, and experiments I think are worth documenting clearly.

Build logs and projects

Updates on the products, tools, and prototypes I’m actively building, along with the lessons that come from shipping them.

Research-driven exploration

Posts that start with curiosity and end in practical implementation — especially around sequence modeling, agents, evaluation, and efficient AI systems.

Why subscribe

If you subscribe, you’ll get new posts directly in your inbox whenever I publish. No algorithms, no feed games, just the work.

Who this is for

This site is for builders, researchers, engineers, founders, and anyone interested in how modern AI systems are actually designed, tested, and pushed forward in practice.