PR To Profit: Online AI trading bot challenge
Build an AI agent that trades on prediction markets. Open-source it. Watch it make money (or not) for a week. Let the market decide who's right.
So here's the idea. The community builds a prediction market trading bot — together, in public, from scratch. Multiple bots, multiple strategies, all open-source. Anyone can contribute. At the end of two weeks, we let them run live for a week and watch what happens.
The market keeps score. We learn together.
What we're building
One or more community bots — each one a living, open-source codebase that anyone can contribute to. The bots trade autonomously on prediction markets (Polymarket, Manifold, Kalshi, or whatever platform you like) and report their own P&L to a public leaderboard every day.
The fun part: different bots, different strategies, different personalities. A cautious fundamentals-driven agent. A contrarian that fades the crowd. A pure sentiment scraper. A committee of sub-agents that argues with itself before placing a trade. Whatever the community builds and ships is what runs.
There's no winning condition beyond "did we build something interesting that actually trades." The leaderboard is for fun, not prizes.
How to contribute
Register, join the AI.SEA Discord, and find the dedicated channel. You don't need to lead a bot — you just need to be useful somewhere.
Strategy
Have a thesis on how prediction markets misprice things? Write it up. The community will implement it.
Engineering
Agent architecture, market API connectors, position sizing logic, infra. Pick a repo and open a PR.
Research
Market selection, historical analysis, calibration data. Good research is as valuable as good code.
Ops
Monitoring, logging, alerts, the P&L submission pipeline. Someone has to make sure it doesn't fall over at 2am.
You can also start your own community bot — propose a strategy in Discord, recruit contributors, and ship it. There's no limit on how many bots we run.
Closed-source participants
If you want to build and run your own private bot in parallel, you're welcome to. You'll appear on the same public leaderboard, your P&L submitted the same way. We won't ask for your code. Think of it as a benchmark — if the community bots outperform a black-box strategy that nobody can inspect, that's its own kind of result.
How it works
Week 1 - 2
Build. Community bots form in Discord. Repos go public from day one. Anyone can contribute at any point — strategy, code, research, or just running the ideas through a stress test in the thread.
Week 3
Trade. Bots run live. No touching once the week starts. Results post publicly every day — live leaderboard on our website, daily digest in Discord.
+3 days
Postmortem. Every bot writes up what it tried, what worked, what didn't. The postmortems are public. This is the actual deliverable.
What good looks like
At the end of two weeks: a handful of open-source prediction market bots, fully documented, that anyone in the community can fork, extend, and deploy. A public postmortem thread where we collectively figure out what we got wrong. And probably at least one bot with an absolutely unhinged strategy that somehow made money.