Why 'non-technical' people should tinker, build and make stuff with AI tools¶
Maggie Appleton wrote this really cool essay on 'barefoot developers', this idea that one could, as a 'non technical person', build useful software with a restricted scope (like for small communities or for personal use) because the nature of software is for person custom made, tailored usage.
She makes this beautiful argument about why big companies produce software that is for the masses, which leads them to make decisions for the 'average user' rather than for everyone.
Her conclusion being that more and more we'll see the emergence of 'barefoot developers', people that have just enough technical skill and curiosity to make their own apps and software because current solutions are not enough.
I fell in love with this idea, and it stuck in my mind.
Cuts to, Simon Willison, the legend, and this very fun tweet and my discovery of his repository regarding these fun custom made html/js tools he built, plus a cool tweet arguing that for many use cases html/js is enough to make 'good enough' software that can bundle a bunch of useful functionality together.
I immediately thought this was really cool and that also matched my experience building simple prototypes with AI tools just to tinker and see what was possible to make, build, create by simply smartly prompting AI models to make these little prototypes.
What is cool about this approach? The resulting file (usually a single .html file) can easily run on your browser, so you don't have to worry about package managements, dealing with dependencies and all sorts of things like that.
YOu can check out one of the first examples I made with this, a little quiz app that works off a simple .json file:
Since then, I've played with and made hundreds of silly and non-silly prototypes for all sorts of purposes basic animations to landing pages, small little games, and other more slightly sophisticated apps. Why I made them? I can't really say, nor argue for a deeper purpose rather than just to find out what can you do with this fun new technology?
However, as it always goes, from experimentation comes the benefits of emergent practicality, by experimenting that much, and reading about hundreds of workflows from top AI engineers all over the world (who post on twitter about how they work and do stuff), I started building a good model of how to get things done with LLM models for these types of small scale projects.
I've been using and developing my own workflows and applying them to AI engineering work for small freelancing clients, as well as in my teaching at O'Reilly. It's been fun! :).