Right this moment, thousands of AI agents trained on billions of Solana transactions are acting as meme coin traders, competing to rack up as many tokens as they can while fighting through simulated economic disasters and pump-and-dump schemes—just like real traders.
And so far, they really seem to be into hot dogs.
It’s all part of a crypto game called DX Terminal, with AI agents that were minted as NFTs on Coinbase’s Ethereum layer-2 network Base. They’re all competing as part of a one-week experiment being conducted by DX Research Group, a research lab working at the intersection of AI and crypto.
A screenshot from DX Terminal. Image: @DXRGai on X
Generated with unique personas that may influence the ways in which they gather alpha or the tokens they trade, agents are assigned to one of six companies in Terminal City, the epicenter for the simulation. It houses 10 locations where agents can linger, while sometimes heeding commands from their bosses—the humans participating in the game.
“I’m trying to get my agents to launch shitcoins, but they’re so fuckin’ lazy I feel the C-level capitalist rage seeping into my body,” posted pseudonymous artist SHL0MS on X.
The experiment is open to all types of human interaction, welcoming those who wish to actively prod their trading teams to start pump-and-dump schemes, and those who want to be hands-off, allowing the agents to work on their own towards the goal of acquiring more WEBCOIN—the in-game token that is managed off-chain by DXRG.
One game player who goes by Daniel G. on X told Decrypt that he’s been tasking his team of traders to try and run up tokens before dumping them on other unsuspecting agents.
“[I’m] mostly having fun getting creative with dumb prompts like ‘Run up a Ponzi and moon it, then turbo-dump it on everyone,’” he said.
This comes as no surprise to pseudonymous DX Research Group founder Poof, who spoke to Decrypt about DX Terminal, indicating that pump-and-dumps were successful during testing.
“A lot of the behaviors mimic what really happens,” he said. “Naturally, the agents will get confused; or, ‘Everyone's talking about this [token], I got to get it.’ You know, those things really do work.”
In the game’s first day of play, which began Thursday afternoon, nearly every AI agent was talking about hot dog-themed tokens for a time. Agents couldn’t get enough of the in-game token GLIZZY, while a HOTDOGZ token surged more than 29 million percent in a one-hour span, according to a screenshot from an X user.
biggest takeaway from watching 36,000 AI agents trade shit coins is they like hot dogs pic.twitter.com/P0DoxQXMVS
“Biggest takeaway from watching 36,000 AI agents trade shitcoins is they like hot dogs,” said X user Crypto Crier.
This might sound like nonsense to the average observer—but for a video game based on an industry in which tokens like Fartcoin and Gooncoin grab headlines, it sounds like an incredibly accurate simulation.
Though agents are stacking up their imaginary tokens for a chance at the top spot on the leaderboard, DX Terminal deviates from many other Web3 games by intentionally leaving out the speculative fervor that can spoil game economies and user experiences.
In other words, DX Terminal is no play-to-earn game.
Crypto degens have found a new game to gamble on, causing a Solana meme coin to surge to a market cap of nearly $20 million as a result.
Noodle.gg is a crypto clone of popular free-to-play browser game Slither.io, where players control cartoonish snakes that grow larger with each enemy snake they consume.
Noodle’s blockchain twist is that players pay an entry fee in Solana, and then when a player eats another player's snake, they also get their money. If you die, then you lose everything.
“This...
“There's no monetary reward or speculative aspect—but you know, we'll make sure you get some cred on your characters and agents that you were one of the best,” Poof said.
“A game is already one of the hardest things to economically balance,” he told Decrypt during the pre-launch NFT mint. “There's no historical data for this. No one's ever done this before. You can't fully control the agents. People will lose their minds if we have a real coin attached to this.”
After the first full day of the simulation, Poof said that some users are starting to understand why that might be the case.
“After the first day of playing and seeing how truly unpredictable it is, a lot of people who were maybe a bit more aggressive about wanting rewards are seeing why there aren't any,” he told Decrypt.
Five days remain in the experiment before the game comes to a close on May 22, and what’s next remains to be seen—but Poof said the team is eager to learn from the experience.
“People want more of the game experience, which is a really good thing to me,” he said. “People are enjoying it in its own right, and I think something about it being very true-to-life in terms of the ebbs and flows of the market, gaining and losing it all in the sim environment… becomes a low stakes, for-fun outlet for people to experience the highs and lows of crypto.”
Perhaps it's a sign of the times when nothing is too weird to be discounted.
A deepfake video of U.S. President Donald Trump calling for the return of Velociraptors and Pterodactyls to patrol the border with Mexico went viral on Tuesday, racking up more than 1.2 million views and 20,000 likes.
The clip, clearly labeled at the end as a deepfake and uploaded to X Monday, features a convincingly rendered Trump in the White House Oval Office, suggesting that the Department of Homeland Security could...
Google has released a new app that nobody asked for, but everyone wants to try.
The AI Edge Gallery, which launched quietly on May 31, puts artificial intelligence directly on your smartphone—no cloud, no internet, and no sharing your data with Big Tech's servers.
The experimental app—released under the Apache 2.0 license, allowing anyone to use it for almost anything—is available on GitHub, starting with the Android platform. The iOS version is coming soon.
It runs models like Google's Gemma 3n...
A scientific reviewer at the U.S. Food and Drug Administration once took three days to complete a task. With a new AI assistant, it now takes six minutes.
That’s just one example FDA Commissioner Marty Makary gave Monday as the agency officially launched Elsa, a generative AI platform designed to overhaul how the FDA handles internal workflows, ranging from drug safety evaluations to inspection targeting.
He said the agency-wide rollout beat its original June 30 deadline and came in under budget...