Index
Case 04 / 2024 / Weekend build

Braindump — type what's in your head, get back one thing to do.

A small web app for the moment when your head is too full to start. You write down everything bouncing around — tasks, worries, half-ideas — and it hands back a single, concrete action you can start right now. Built and deployed in a day.

Next.js Supabase Vercel OpenAI Tailwind
RoleDesign · Build
Year2024
Scope~1 day
StatusLive · personal use
braindump.app — landing

Unload your mind.
Start one thing.

Turn mental noise into one clear action.
Get started — it's free
No credit card required. Takes 30 seconds.
The case in one paragraph

Most productivity tools ask you to organize when what you actually need is permission to stop organizing. Braindump is the smallest app I could imagine that does the opposite: one text box, one button, one answer. I built it in a day to have it for myself, and to practice shipping the whole loop — frontend, auth, model, deploy — on a stack I'd never used together.

01 The problem 02 How it works 03 Three screens 04 How it was built 05 Honest outcomes
01 · The problem

You open a planner when you're stuck. Then the planner makes you more stuck.

The moment
Your head is full and you can't start anything.
Ten things feel equally urgent. You know you're avoiding the important one. You open a notes app hoping it'll help and close it five minutes later with a longer list.
Why lists don't help
A list is still ten things.
Writing it down organizes your anxiety, but it doesn't reduce it. You still have to pick, and picking is the thing your brain can't do right now.
What you actually need
Someone to just tell you what to do first.
Not a suggestion, not a ranked list, not a “here are your options”. A single concrete sentence you can do in five minutes, so the stuck feeling breaks.
The design bet
One output is the product.
Offering two choices reintroduces the original problem. So there are no choices. You get the one thing, a line explaining why, and permission to ignore the rest.
02 · How it works

Three screens. Under two minutes end to end.

You land, you dump, you get one thing. That's the full loop — no dashboard, no history, no settings. Here's what each step does before we zoom in on the screens.

Step 01 · You arrive
Landing page.
A short promise and one button. No feature list. If the idea doesn't land in one line, the product isn't the problem.
Step 02 · You write
One text box.
Type whatever's in your head. No formatting, no tags, no length limit. The button says Unload.
Step 03 · You get one thing
A single action.
Plus a short reflection naming the pattern, and a small “you don't need to think about these” list for everything else.
03 · The three screens

Recreated from the live app.

These are rebuilt in HTML using the same colors and type as the shipped product — not screenshots. I'll explain what each screen is doing and the one or two design choices that matter on it.

Screen 01 · Landing

What it is: the front door.

A visitor's first ten seconds with the app. One sentence, one button. A dark cocoa background makes it feel like a quiet room rather than a cheerful productivity tool — nobody shows up here when they're calm.

The main choice. No screenshots or feature bullets. If the sentence “unload your mind, start one thing” doesn't explain it, adding more copy won't either.

braindump.app — /

Unload your mind.
Start one thing.

Turn mental noise into one clear action.
Get started — it's free
No credit card required. Takes 30 seconds.
A ritual for the overwhelmed mind.
Screen 02 · The dump

What it is: one text box, no rules.

The prompt is a question: “What's in your head right now?”. You type whatever comes out — tasks, feelings, distractions — all in one paragraph. The button says Unload, not Submit.

The main choice. No formatting, no tags, no minimum length, no character counter. Every widget I didn't add is one less thing to think about while trying to think less.

braindump.app — /dump
Log out
What's in your head right now?

Honestly I've been avoiding the hard stuff and just doing easy busywork. The video is the thing that actually matters but I keep finding excuses. Maybe I should just watch some tutorials first? No that's just more procrastination. Also need to call mom back and figure out what's happening with the car insurance renewal.

Just feeling scattered today. Too many tabs open in my brain.

Unload
Screen 03 · The result

What it is: one reflection, one thing, a short list of “not now”.

Top: a single italic sentence naming the pattern in the dump — not advice, just observation. Middle: the bordered card holding the one thing — verb-first, startable in five minutes. Bottom: a quiet shadow list of the other stuff so it doesn't feel lost.

How it decides. The dump gets sent to a language model with a strict set of rules: pick exactly one action, make it startable immediately, phrase it as a commitment, never a suggestion. The instructions to the model are more carefully written than anything else in the product.

braindump.app — /result
Log out
You're recognizing the pattern of avoiding the video editing by finding easier tasks and justifications.
Your One Thing
Open your video editing software and load the project file you need to finish.
You don't need to think about these right now.
  • Respond to Sarah's email about the project timeline
  • Schedule the dentist appointment
  • Call mom back
  • Handle car insurance renewal
Do this one thing and leave the rest at the door.
Dump again
04 · How it was built

Four off-the-shelf services, wired together.

There's no clever architecture here. The goal was practice — running the whole loop from login to deploy using pieces I hadn't used together before. Each piece below does one job.

The pages you see
Next.js + Tailwind
Three screens. Landing, dump, result. Styled with Tailwind so I didn't have to name CSS classes.
Login + saving
Supabase
Email magic-link login. One table that stores each dump and what the app wrote back, so I can re-read old results later.
The “one thing”
OpenAI gpt-4o-mini
The dump is sent to a language model with a short, strict instruction. The instructions are the product opinion — the UI just holds the output still.

The real takeaway. The unglamorous plumbing — login redirect URLs, environment variables across local and production, not running up a model bill — took more hours than the product itself. That was the point. The next time I ship something on this stack it'll be faster, because I know where the sharp edges are.

05 · Honest outcomes

What it is, and what it isn't.

~1d
Idea to deployed
Single build session
3
Screens total
Landing · dump · result
1
Output per session
On purpose
n = 1
Regular user
Me, when I need it

This is a weekend experiment that happens to run in production, not a product with a growth curve. It works for me when my head is full. A few friends have the link bookmarked. That's the full scoreboard.

What I took away: the boring middle of shipping — auth, hosting, billing, deploy — is the part that slows you down the first time you touch a new stack. Getting through it once on a throwaway project means the next real thing I build on Next.js and Supabase won't cost a weekend of setup.

What was hard. Not the code. The editorial voice. How confident should the app sound when it picks your one thing? If the pick is wrong, does it apologize or stand behind it? Braindump lives between a coach and a task runner — too much of either and it breaks. Writing the instructions to the model was the product design.