A notebook, kept in public
small, practice-based, and open to correction
No products and no courses. Just notes kept while doing the work, published in case they are useful to someone doing the same thing.
What we do
the short version
AI Learn Lab is a working notebook. We build AI-assisted software, and we write down what we learn while doing it — usually in the form of a single decision, its trade-offs, and whether it turned out to be right.
Most of what we write about sits in unglamorous territory: how to tell whether a change to a prompt helped, what to log so that failures are visible, where to draw the boundary around a first version so that it can actually be finished and measured.
How we try to write
four habits we hold ourselves to
We aim for notes that would still be useful in a year. In practice that comes down to a few habits.
specificity
Describe the specific case rather than generalising from one project to an industry.
trade-offs
Include what an approach costs. If it buys accuracy with time, or flexibility with complexity, say so.
uncertainty
Say when we are unsure, and say when something we wrote earlier turned out to be wrong.
no predictions
Avoid forecasting where the field is going. We are not well placed to make that call.
What this site is not
worth saying plainly
It is not a product site and there is nothing to sign up for. We are not selling tooling, running a course, or offering a platform. If a note mentions a specific tool, it is because we used it, not because of any arrangement with whoever makes it.
We also try to avoid the two default registers of AI writing — that everything has changed, or that none of it works. Both are easier to write than an accurate account of a particular problem.
Corrections
we would rather know
If something here is wrong, we would like to know. Notes get updated when we learn better, and substantive corrections are noted in the text rather than quietly edited in.
Email hello@ailearnlab.org, or see the contact page for what is useful to include.