testy.coolI build LLM automations, research agents, and web dev agents. This site is where I write notes about all of it.
What I work on
Most of my time goes into making LLMs do useful things - not chatbots, but actual pipelines that take messy inputs, run multi-step reasoning, and produce structured outputs. I work with Claude, GPT, and Gemini across different use cases, and I write about what works, what breaks, and what I learned along the way.
I'm especially interested in agent architectures - how to give LLMs tools, how to manage context across long workflows, and how to build systems that recover from errors instead of falling over. A lot of the posts here come out of debugging sessions or experiments that turned up something surprising.
What's on this site
The blog is the main thing. Most posts are one of:
- Tutorials that build something up from zero, usually with interactive tools embedded
- Lab notes from experiments - short, concrete, sometimes inconclusive
- Troubleshooting writeups for specific problems and the fixes that worked
There's also a tools page with standalone utilities like the LLM Price Calculator and some browser extensions.
Why this exists
I write primarily for myself. Documenting what I learn forces me to understand it properly, and having searchable notes saves me from re-solving the same problems. If other people find these notes useful, that's a bonus.
The site runs on Next.js with Fumadocs and deploys to Cloudflare Pages.