What is the Lab, and why we publish it
The Lab is where we share the tests and checks we run while operating an AI API gateway, so you can use what we learn instead of repeating the work.
We run an AI API gateway, and we are always testing something. How fast does a model answer? How does a coding tool behave with a custom endpoint? What does a change do to the bill? Why did that request fail?
Most teams keep these results in a private note or a chat thread. We decided to publish ours.
Why we share them
Everyone who builds with AI ends up running the same tests and making the same mistakes. That wastes time and money, and much of it is easy to avoid when someone has already checked.
If one of our results saves you an afternoon, or stops you from picking the wrong tool, it was worth writing down.
What you will find here
- Tests. The same task run on different models, tools or settings, then compared.
- Checks. What happens when something goes wrong: errors, slow replies, limits.
- Setup guides. What works, with the exact steps, and the date we last checked.
- Plain advice. What we would pick, and why.
How to read a post
Every post starts with the short answer, so you can stop there if that is all you need. If you want more, the rest shows how we tested it, the numbers, and where we might be wrong.
Every post shows when it was last updated. Tools and models change quickly, so an older post says so.
What we never publish
- Anyone's prompts or responses, or anything that identifies an account.
- Secrets, or details of our setup that would make the service easier to abuse.
Our interest
We sell API access, so some results matter to our business. When a post touches something we sell or compete with, we say so near the top. We also publish results that do not favor us.
The Lab is in beta
We are still shaping it, and it will grow. If there is something you want tested, or something here is confusing or wrong, tell us. The best suggestions are specific: what you would test, and how you would judge the result.