Lab Notes

Introduction

Welcome to the lab notes.

These are short, plain guides to using AI well. Not magic tricks. A method.

Most AI advice online is folklore. Someone says "try this prompt, it works great," and you have no way to know if that is true for you. We do it differently. We treat using AI like lab work. We test things. We measure them. We write down the steps and the numbers so you can repeat them and see for yourself.

You do not need a degree to read these. We write at about an 8th grade level on purpose. Plain words, real examples, and the math shown.

How to use this page. If you are new, read the notes in order. Each one builds on the last. If you need one thing, jump straight to it. Every note stands on its own and ends with a "Try it yourself" so you can run it.

Lab principles

These are the rules we hold ourselves to. They also make a good way for you to work.

1. We test, we do not guess. We never just tell you something works. We show you how to check it.
2. Every note can be repeated. If we claim a result, we give you the steps and the numbers to get the same one.
3. Measure before you trust. Keep a few test cases you can run again. That is how you know a change really helped.
4. Change one thing at a time. It is the only way to learn what actually made the difference.
5. Start small and cheap. Use the smallest model and the shortest prompt that does the job. Step up only when you must.
6. The right context beats more context. A few useful details help more than a wall of text.
7. Write it down. A result you did not record is a result you cannot repeat. Keep a lab notebook.
8. Be honest about limits. We say what we know, what we do not, and what might change.
9. Protect privacy. We keep people's names and private data out of public notes.
10. Facts go stale. Prices and tools change. Always check the source for today's number.

Where this is going

More notes will land here over time. Each one is the same method pointed at a new piece: writing better prompts, picking the right model, running your own private model on your own machine, and building systems that do many steps at once.

The goal is simple, and a little strange for a lab to admit. We want you to stop needing us. When you can test your own AI work and trust the result, you are doing real lab work. That is the whole point.

This is an open lab. Try the steps. Push on them. If something breaks, or you find a better way, we want to hear it.

Thanks for reading

Your attention is the rarest thing you own. Thank you for spending some of it here.

We will keep these notes honest, plain, and worth your time. That is the deal.