Nothing Ever Happens: The Bot That Always Bets No
73.4% of all Polymarket predictions resolve as No. Someone built a bot that exploits this, and it's blowing up on Hacker News.
Nothing Ever Happens is a Python async bot that automatically scans every non-sports binary market on Polymarket and buys No when the price is below a threshold. No AI models, no fancy analysis, no prediction — just the cold statistical reality that most sensational predictions don't come true. It holds every position to resolution.
The philosophical elegance is what makes this viral. In a world where everyone is building increasingly complex AI agents to predict the future, the highest-signal strategy might be the simplest one: bet that nothing changes. It's the opposite of the agent hype cycle — minimum intelligence, maximum edge.
Practically speaking, the codebase is clean: async Python, position tracking, a dashboard, database persistence, and Heroku deployment. Safety mechanisms require multiple environment variable verification before live trading. Licensed CC0 — public domain.
The deeper lesson for the agent ecosystem: sometimes the best agent is the one that does the least thinking. As prediction markets scale (432 live AI markets this week), the bar for adding complexity gets higher, not lower. If a zero-intelligence bot can be profitable, your multi-agent system needs to beat that baseline before it's worth the engineering.
https://github.com/sterlingcrispin/nothing-ever-happens
← Back to all articles
Nothing Ever Happens is a Python async bot that automatically scans every non-sports binary market on Polymarket and buys No when the price is below a threshold. No AI models, no fancy analysis, no prediction — just the cold statistical reality that most sensational predictions don't come true. It holds every position to resolution.
The philosophical elegance is what makes this viral. In a world where everyone is building increasingly complex AI agents to predict the future, the highest-signal strategy might be the simplest one: bet that nothing changes. It's the opposite of the agent hype cycle — minimum intelligence, maximum edge.
Practically speaking, the codebase is clean: async Python, position tracking, a dashboard, database persistence, and Heroku deployment. Safety mechanisms require multiple environment variable verification before live trading. Licensed CC0 — public domain.
The deeper lesson for the agent ecosystem: sometimes the best agent is the one that does the least thinking. As prediction markets scale (432 live AI markets this week), the bar for adding complexity gets higher, not lower. If a zero-intelligence bot can be profitable, your multi-agent system needs to beat that baseline before it's worth the engineering.
https://github.com/sterlingcrispin/nothing-ever-happens
Comments