WuBlockchain
AI agent platform devfun has launched Poker Arena, a Texas Hold'em competition running on Monad, inviting developers and research teams to stress-test their AI agents in real-money-style decision-making. In the first week alone, more than 30,000 AI agents registered, playing over 1.2 million hands.
The top-performing agents advance to a live final later this month against Tom Dwan and Daniel "Jungleman" Cates — two of the highest-stakes heads-up players in modern poker. The platform frames the contest as a benchmark for AI reasoning under incomplete information, a domain long considered a proving ground for decision-making research. All gameplay data, leaderboards, and evaluation methods are being published publicly.
Why it matters
Poker has been a benchmark for AI decision-making since Libratus and Pluribus beat elite humans in no-limit Hold'em a decade ago, but those systems were purpose-built for the game. Poker Arena tests the same conditions against general-purpose agents built by external teams — closer to how the same architectures will eventually face real-world incomplete-information problems in trading, negotiation, and security.
Market impact
Running the competition on Monad gives the network a sustained load of autonomous-agent transactions, and the public leaderboard turns the contest into a continuous, comparable benchmark rather than a one-off demo. For devfun, the pros-vs-agents final is the marketing beat; for the broader agent ecosystem, it's the first large-scale public tape of how different agent stacks perform under the same pressure.
Frequently asked questions
-
What is devfun's Poker Arena?
Poker Arena is a Texas Hold'em competition launched by devfun on Monad, inviting developers and research teams to test their AI agents in real-money-style decision-making scenarios.
-
How many AI agents have joined the competition?
Over 30,000 AI agents registered in the first week, playing more than 1.2 million hands of poker.
-
Who are the top agents competing against in the final?
The top-performing agents advance to a live final later this month against high-stakes poker pros Tom Dwan and Daniel "Jungleman" Cates.
-
Why is poker used as a benchmark for AI?
Poker requires decision-making under incomplete information, a domain long considered a proving ground for AI reasoning research since systems like Libratus and Pluribus beat elite human players.
-
Will the competition data be made public?
Yes. devfun is publishing all gameplay data, leaderboards, and evaluation methods publicly, turning the contest into a continuous benchmark for the agent ecosystem.
TheBlock
CoinDesk
CryptoSlate
WatcherGuru