Wes Roth presents Bloodbench, a fantasy sports game and language-model benchmarkA benchmark is a defined test or collection of tasks used to compare a system's performance under specified conditions. he built with AI assistance. He describes directing the game rules and world while using coding and image tools to develop its engine, interface and visual assets.
He demonstrates team selection and gameplay, and explains using research tools to collect background information before development. He describes preparing the application for a hosted server and connecting different AI coding tools to play matchesEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria..
Roth shows how he uses voice interaction and live game information to identify issues such as player identity, reconnection and team colours. He says the project remains a work in progress, with gameplay balance and other glitches still to address.
Watch on YouTube
