When I learn about a new technology that I believe is sufficiently interesting and impactful, I create a project to build something with that tool.

I try to spend a little bit of time understanding a tools affordances, where it can work, what it’s limits are, and then think about how it can be used practically in its current form, not in some future better state.

As an example - when Stable Diffusion was released, I create an app called Fresco, a simple game that to create artwork using diffusion models. But instead of trying to mimic hyper-realism (which, at the time, models weren’t good at), I made the app focused on creating abstract art, so that the hallucinations and inaccuracy became a feature more than a bug.

Sometimes I will choose to blend learning two things together into one project. For example, when I want to try a new framework or way of coding, I’ll force myself to use that framework in the creation of this new utility. In the Fresco example, it was an exercise in learning a new framework called Hotwire for interactive web app.

Creating an artifact of your learning that is more than a set of notes feels more satisfying too.