What I learned about products from making tacos

When I was a teenager, my mother was out a lot getting a master’s degree. So I would sometimes cook our family tacos - the midwestern standard Ortega taco mix, hard shells, ground beef, tomatoes and cheese and such. My family always told me they liked when I made them, for some reason. It was one of the first times I remember being complimented for making something.

But it’s an easy kit to make. So, why did they like mine better? It turned out to be really simple: my mom would slice up tomatoes so they were fairly big and would fall out of the taco. I would cube them into smaller, more uniform pieces, that just worked better. User design! It’s simple, it’s stupid, it’s obvious, anyone can do it…and yet, often no one does.

I see a huge amount of this kind of neglect in the current crop of AI based products. I wrote last week (and I’ve written before about this - sorry, on a rant I guess) about AI being a weird coworker that doesn’t do anything. Some of this is just because the builders of these products are ignoring so many basic bits of fit and finish.

Sometimes it’s much more egregious than just fit and finish. One of the three early image generators required Discord to use - much harder than the other two, so after looking once, I never went back. Not enough value to justify the hassle. I just saw another one this week that’s much better than the rest, but requires yet another third party service (this time GitHub) to use it. Which is fine, but janky enough that instead of it being 10 seconds to ask ChatGPT or Bing, it’s usually a minute of “dammit, what’s my password again, let me reset it” bs. The tomato is falling off the taco.

We as engineers are enamored of novelty and being able to build cool stuff. Users are not. They are enamored of us helping them get shit done effectively. To quote my favorite writer* “Users are lazy”. Many teams are building these fantastic new capabilities, and then shipping them in the software equivalent of a splintery box nailed together in haste. It’s just silly!

If you haven’t done this with your product lately, one of the best things you can do is give it to someone you don’t know (buy them a latte or something), tell them to use it, shut your mouth and sit on your hands while they do, and watch what happens. Don’t pick your programmer friend who’s just like you. Find someone in a coffee shop or a friend of a friend who’s not in the tech industry, but who you want to have as your user. See if your solution actually works and is smooth for them.

As a maker, for a long time, I was obsessed with getting the object “done”. So, I was more focused on speed and less on quality. I made a TON of stuff - most of it crappy. I’ve learned to sand and polish and focus and be still and careful and look and do all the things I can to make what I’m making as good as I can make it.

Find the parts of your product that aren’t right for your users and ruthlessly work on them until they’re perfect.