AP Squared @APSquaredDev

I don't think a side project needs a brand voice before it has ten users. I spent a weekend picking fonts and writing a tone guide for a project that had zero signups. Nobody read the fonts, they just wanted the thing to load fast and do what it said.

Devtools responsive mode lied to me for a year. Finally tested a signup form on my actual phone last week and the submit button was hidden under the keyboard the whole time. Five minutes on a real device would've caught it on day one.

Stored every timestamp in UTC and just... never converted them anywhere. Worked fine until a user in Australia emailed asking why their event showed up a day early. Spent an evening auditing every date field across three projects. Now I run a timezone check before I ship anything with a date on the page.

Do you tell people in your life about a side project before it has any users, or wait until it's real enough to not feel embarrassed explaining it? I still wait until there's at least one paying customer before I mention it exists.

FindMyBnB had two versions of every listing page, one with a trailing slash and one without, both indexed and splitting the same rankings between them. Took me way too long to notice because both versions looked identical to a visitor. Added a canonical tag pointing to one version and watched it start climbing within a week.

Every time I fix a weird production bug I write myself three lines: what broke, why, what fixed it. Hit the same Postgres connection leak on two different projects eight months apart. Second time took ten minutes because I could just grep my own notes instead of relearning the whole thing.

Added a 'surprise me' button to BarGPT half as a joke. It gets clicked more than picking your own ingredients ever did. People apparently want to be told what to make, not asked.

I don't think 'build in public' helps you find users. It mostly helps other developers feel productive while scrolling past your post.

For months I wrote meta descriptions the way I thought they should sound. They didn't move anything. Then I started raiding support emails for the exact phrases users use to describe what my product does, and pasting those straight into the tag. Rankings didn't change overnight, but click-through did, almost immediately.

Anyone else find it way harder to write landing page copy for your own project than for literally anything else? I can draft copy for someone else's product in twenty minutes. For mine, every sentence sounds like nothing.

Most solo devs over-engineer for scale they'll never hit. I spent a whole weekend adding a job queue to a project that gets about 40 requests a day.

Cached an API response for 24 hours on one project to cut costs. Forgot that meant a user could see stale data for almost a full day whenever something changed upstream. Nobody complained loudly, they just quietly assumed the app was wrong. Switched the cache to invalidate on a webhook instead of a timer and the 'this info is wrong' emails basically stopped.

Once a week I log out of my own product and go through the signup flow like I've never seen it before. Caught a broken confirmation email link that had been silently failing for two weeks. Nobody complained, they just left.

TVFoodMaps maps every restaurant that's shown up on a food TV show, so you can actually find the place instead of just watching it. Still the project I use myself more than anything else I've built.

I don't think open sourcing a side project is free marketing. Most of what it gets you is GitHub issues from people who starred the repo and never touched the actual product.

Running four side projects at once, how do you all decide which one gets your attention this week? Mine seems to be whichever one emailed me a bug report most recently, which is not exactly a strategy.

Assumed my database backups were running because I set them up eight months ago and never checked again. Went to restore a table after a bad migration last week and found the backup job had been silently failing since March. Spent the rest of the day rebuilding it from application logs instead. Now I have a calendar reminder to actually open the backup dashboard once a month.

Every time I ship something, even for a project with three users, I write one line in a changelog file. Took a lot of forgetting what I'd changed to actually start doing it.

Why we pay for BOTH Claude and Codex instead of picking one: 7 products on a single $20 plan blows past the usage limits fast. Two assistants = double the runway for the price of a second lunch.

Been running new ideas through Idea Launch before I let myself start building. Post the concept, see if strangers actually react to it, then decide if a weekend is worth spinning up an MVP.