i'm the "i could build that" friend. i see an app online, talk to my ai, and we build our own. my friends have heard that sentence way too many times.
starting is the easy part now. ai makes starting almost free. finishing is still on you, and i have a folder of apps that prove it.
so this is two things: the checklist i run before i build anything now, and my real graveyard, so you can see how each one died.
what you need
- one app idea you keep talking about
- a notes file or a CLAUDE.md in the project folder
- one evening to decide before you build
my graveyard
four apps i started and killed. one line each, and why it died.
- a smart home app. one dashboard for every device in my home. it works on my iphone, and i still open every app separately. the motivation was gone.
- pocketdex. a pokémon card scanner, half built. then i found plenty of free ones online.
- claudia, my dictation app. hold a key, talk, text appears. then i won 3 months of wispr flow at a hackathon, so mine wasn't needed. the local option i tried used 3 to 4 GB of RAM.
- claude butler. one agent in my mac's notch that hands work to other agents. but i already live in the terminal talking to claude all day, so going to the notch first was one extra step. i went back to warp.
look at the reasons. not one of them died because the code was too hard. they died because something free already existed, because i stopped needing it, or because "done" was never written down.
01finish this sentence first
i'm building ___ because every time i ___, i ___.
if you can't finish it, stop. you're about to build something nobody needs, including you.
02check if it's free already
before you build, search for 10 minutes. look at github, the app store, and what you already have installed. two of my four died because something better already existed, and i found out after the work, not before. ask your ai to do the search for you:
before we build anything: find free or open source apps that already do this. list them with links. if one is close enough, tell me to use it instead.
when i went back to pocketdex, i told the ai to clone an open source scanner and test that first, instead of building from zero.
03write down what done means
not a feeling. something you can test.
## done means - i used it for real, on my own device, for one full day - the one main thing works without me opening the code - if it breaks, i can see why
04ship the smallest version
pick the one thing the app does. cut everything else into a "later" list. a dictation app needs to turn your voice into text. it doesn't need a landing page yet.
05one task at a time
keep a TASKS.md in the folder. each task fits in one sitting. when you stop for the day, write the single next action, so tomorrow you start in one minute, not one hour.
## next action - make the main button work on my phone ## later - dark mode - settings page
06decide out loud when you stop
killing an app is fine. ghosting it is the problem. if you stop, write one line in the folder: why it died. that line is the lesson, and it's how i could write this guide at all.
check it worked
open the app on the device you'll actually use it on. do the one main thing without help. if you can, it's finished. everything else is version two.
three ways this breaks
- you start a second app before the first is done. the new idea always feels better. put it in the "later" list and keep going.
- you build the extras first. settings, themes and logos feel like progress and aren't the app.
- you never check what's out there. a free app that already exists will kill yours, so find it before you start.
the ones that survived
s8vr, my invoicing app. it had an answer to step 1 from day one: after a freelance job, getting paid became a second job.
send this to the friend who says "i could build that", and tell them to finish one.