A Good(?) Use Case for AI

Victory Gardens of home-cooked coding

This is what I got using substack’s AI tool to respons to the prompt “woman barefoot in kitchen programming on the command line” I don’t use AI art here unless I am commenting on the practice.

What does software development look like as part of a home ec practice?

That’s the vibe I get from Maggie Appleton’s essay on “Home-Cooked Software” (h/t to Jake Meador of Mere Orthodoxy for flagging this for me). Appleton is interested in how AI programming tools could make building software a casual, cheap endeavor, and thus accessible to a wider range of developers (who don’t think of themselves as developers) interested in solving small-scale local problems.

Early in her presentation, she points to what Clay Shirky called situated software:

He painted a vision of applications that could be used by dozens of users rather than thousands or millions. This is an absurd target population both then and now.

The dream was that communities would get very form-fit tools for their particular needs, rather than trying to adapt generic software to solve them.

A number of thoughtful people work in this space. Alan Jacobs is a writer who is very thoughtful about how he connects the pipes and tools that let him write without being swept away by the internet maelstrom. (And then, appropriately enough, receive his missives via my RSS reader Feedly).

Appleton approvingly cites Robin Sloan’s homebrew family photosharing app, whose only users are his own family:

The app is a “magic window” that captures photos and videos and shuttles them around. Messages wait in a queue and, once viewed — always full-screen, with no distractions, no prods to comment or share — they disappear. That is literally it. The app has basically no interface. There’s a camera button and a badge in the corner, calm green, that indicates how many messages are waiting.

What excites Appleton about the advent of AI coding assistants is the chance that building a light, limited user app like this could be much more accessible to people who don’t think of themselves as programmers.

Without an expectation you can get under the hood and build what you need yourself, you’re dependent on whomever thinks they can profit from your need. That’s not intrinsically bad—I don’t resent my grocery for selling me food! But it is a reason I picked an RSS reader I could pay for with money rather than paying by having my attention sold to advertisers.

I’m much further out on the “comfort with programming” spectrum than the median person. I’ve coded in Java, Python, Scheme, R, STATA, SPSS, and possibly some others I’ve forgotten. But, for the most part, they were toy problems for class. I’ve only picked up code to solve a problem a couple times—mostly for data journalism or Project Euler.

When I’ve wanted to solve a problem related to how I use my computer and phone, I’ve more often turned to If This Then That, a service that sets up domino chains where an action here triggers a reaction there. But everything goes through pipes you don’t control yourself, and Appleton hopes that AI coding assistants can let you build tools you own, that keep your data local and private.

Anything that lets us think about our software and our computers as tools we can alter offers us an invitation to ask tools about how we’d like them to be different. Once I can opt out of a total, locked down system, I can ask where I’d like to add friction and how I want my tools to shape me as a user.

Appleton’s post reminds me I should take another look at building the tool I’ve wanted for a long time, but have not gotten over the new programming tools hurdle to actually build (esp with an expanding family).

I’d like to write a Chrome extension that delays opening certain webpages (like Twitter) and shows an interstitial page for 30 seconds or so, with a randomly chosen prayer intention from a list I can edit.

It sounds like that’s definitely something I should be able to snap together from well understood code snippets, but brief, casual poking around hasn’t gotten me there. I’m curious if Copilot will.

What micro app would you build for yourself or your family? How would it change how you approach tech?

Where (if anywhere) have you tinkered with the tech you use (with code or otherwise) to make it serve your goals, not the goals of the company who sold it to you (or sells you to advertisers?

What do you think of AI as a way to leapfrog deep coding knowledge?