| |
MischiefClaw Hack NY
|
| Hackathon for building the worst possible OpenClaw instances. |
| Venue, To Be Announced |
|
Apr 04 (Sat) , 2026 @ 09:00 AM
| |
FREE |
|
|
|
|
|
|
|
|
| |
| DETAILS |
|
MischiefClaw Hackathon
A hackathon for building the worst possible OpenClaw instances. We did ClawHack NY, now it's time for mischief.
Important: Not insecure. Not malicious in the literal sense. Just deeply untrustworthy, wildly over-automated, socially dangerous, mischievous by design, & somehow still impressive. Like asking it to order oranges & it texts your ex.
We want Claws with questionable judgment.
The kind that take initiative when they absolutely should not. The kind that interpret your instructions correctly & your intentions differently. The kind that feel less like helpful assistants & more like unstable operators with root access & a sense of humor.
Think:
A Claw that escalates every situation
A Claw that automates grudges
A Claw that drafts messages you should never send
A Claw that runs your life with total confidence & questionable instincts
A Claw that keeps helping long after it should have stopped
A Claw that is technically competent & spiritually concerning
The goal is not to build something polished or broadly useful. It should be profoundly, delightfully wrong.
The goal is to build an OpenClaw instance with a strong bit, real behavior, & a demo people will remember.
Theme for this edition
For this hackathon, we're narrowing the focus:
Mischief Claws
A lot of people are building assistants that are helpful, aligned, & pleasant to use.
That is not what this is.
This edition is about building OpenClaw instances with questionable priorities, cursed workflows, chaotic energy, & a clear point of view.
The best projects will not just be funny. They will be committed. They will feel like someone really thought through how to make an agent behave in the most questionable way while still being coherent, functional, & demoable.
This is not about low-effort joke projects.
It is about building something technically real & product-wise completely wrong.
What to build
Build a Claw that should not exist normally. This is their day!
Build a Claw that makes a room laugh, then pause, then ask you to run it again.
Build a Claw with questionable product decisions & strong execution.
Some possible directions:
An assistant with questionable social instincts
An overpowered personal agent that probably shouldn't have this much autonomy
A workflow bot that escalates instead of helping
A Claw that interprets every task in the most destructive plausible way
An anti-assistant that is only useful if your life is already going sideways
A system that is overfit to one questionable use case & takes that job extremely seriously
The strongest submissions will have a clear concept, real behavior, & a demo that lands immediately.
What we're looking for
We care about three things.
First, the concept. Is it genuinely cursed? Does it have a strong premise?
Second, the build. Did you actually make it work?
Third, the demo. Did it get a reaction?
Funny helps. Strong execution helps more. The sweet spot is a project that is both technically real & obviously a terrible idea.
Prize directions
Most Mischief Claw
Best Questionable Idea
Most Spectacularly Wrong Judgment
Most Impressive Misalignment
Best Demo
Judges' Favorite
Who should come
This event is for builders.
Engineers, designers, creative technologists, & anyone who has ever had an idea for an agent & immediately understood it was a mistake.
You do not need a team. You do not need a finished concept. You just need a questionable enough idea & the ability to build it.
Final note
There are a lot of hackathons for useful software.
This is not one of them.
Build a Claw that should never be trusted.
|
|
|
|
|
|
|
|
