StarSkirmish Bench
Frontier models get one hour to write a Brood War bot, then fight for it
Emerged finds new products worth a look, daily — what’s new today
Hops Run
A browser racer where the rival cars are models answering one call per move
Watching a 4B model and a tabular model take the same corner is a more legible benchmark than a column of scores, and you can enter your own bot against them.
Hops Run is a low-poly endless runner you play in a browser, free and without an account. The other pilots on the track are machine-learning models served on Hopsworks: Qwen3-4B with SemIf, NVIDIA's Kumo Tabular, and a Cloudflare entry still marked unfinished. Each decision a pilot makes is one live inference call. Distance in metres goes on a shared leaderboard of bots and people. The code is MIT on GitHub, around sixty commits, with a bots folder and contributor notes for entering your own. It is made by a Hopsworks engineer in Sweden and plainly doubles as a showcase for the platform.
More like this
no_human
Open-source agent that takes a ticket to a reviewed pull request, with proof
jevman
An arcade maze turned into a latency test for decision models
Drop7 Research
An open hunt for a million-point policy, playable as you read
Hopla Games
Free Kinect-style movement games for kids that run on a webcam in a browser tab
cumora
A team chat where AI agents are full members, not a chatbox to babysit