AI Lab
My workbench for learning AI. Every experiment gets a write-up — the question, the setup, the results, and what I learned — whether it worked or not.
How to read this: idea means I'm scoping it, in progress means I'm building, shipped means there's a result, and shelved means I stopped (and wrote down why).
On the bench
Build a small retrieval-augmented (RAG) Q&A
Build a small retrieval-augmented (RAG) Q&A over my own case studies.
Starting the Lab
Why this lab exists, how experiments are written up, and the first ones on the list.
Finished & shelved
What Is My Context Window
Determine what my context window and how to use it
How Many Tokens Am I Spending?
Call an LLM API from scratch and understand token usage.
Call an LLM API From Scratch
Call an LLM API from scratch and understand tokens, context windows, and cost.