move fast, break nothing.

a rapid app building league with automated stress testing

We took a hackathon and messed up the units. You have 24 ~~hours~~ minutes to build, and then everything you built gets put to the test.

[ get notified ]how it works →

hack•let(hæk-lət)n.

An app built in a very short amount of time, typically with AI assistance: My roommate built a hacklet while waiting for his flight.
A tightly compressed hackathon with an audience possibility: Come attend a hacklet; it lasts no longer than a cs club meeting.

# how it works

HackLet League runs three formats. All three compress engineering into 24 minutes with AI, all three run the same QA catalog at time expiry, all three score across slop and communication. They differ in what you do during the build phase.

see exactly how scoring works →

$ ./hacklet --formats
  vibe            build a working web app from scratch
  unslop          diagnose and fix a broken ai-generated app (the
                  kind you'd call "slop")
  underspecified  build to a vague client brief, and defend how you
                  read it

## HackLet Vibe: build from scratch

Build a working web app with AI assistance of your own choice.

$ ./hacklet --format vibe --timeline
  5 min     opening: round prep
  24 min    build: ai-assisted building (aka vibecoding)
  18 min    defend: qa testing + pitch prep
  5-30 min  pitch: judging
  8 min     awards: closing

HackLet Vibe is akin to a traditional hackathon except with time compressed to minutes instead of hours. As AI is ubiquitous, HackLet Vibe tests if you can build apps of function, instead of apps of just form.

## HackLet Unslop: fix the slop (hence the name ;) )

Receive a broken AI-written codebase at the beginning. Figure out what’s wrong across multiple dimensions and improve upon it. Ship a defended version.

$ ./hacklet --format unslop --timeline
  5 min     opening: broken codebase revealed
  24 min    remediate: ai-assisted diagnosis and repair
  18 min    defend: same qa testing + pitch prep
  5-30 min  pitch: same judging
  8 min     awards: closing

HackLet Unslop reflects the reality of engineering work: working with existing codebases that may or may not be functioning ideally. HackLet Unslop tests if you can deal with existing code you see at work... under pressure.

## HackLet Underspecified: read the brief

Get a vague, half-formed client brief. Work out what they actually need, build it, and defend the call you made. The brief does not hand you the answer, and reading it wrong is part of the test.

$ ./hacklet --format underspecified --timeline
  5 min     opening: the vague brief drops
  24 min    interpret + build: decide what it means, then ship it
  18 min    defend: same qa testing + pitch prep
  5-30 min  pitch: defend how you read the brief
  8 min     awards: closing

HackLet Underspecified tests the thing a real client meeting tests: turning an unclear ask into a defensible plan. Most engineering that fails does not fail because the code was bad. It fails because someone built the wrong thing well.

# want to run one?

HackLet is built to be hosted. start a chapter for your school, club, or company, and run your own rounds. we review new chapters before they go live.