


Yuan Han ✈️
1.2K posts

@SourceAdj
A crypto buyer https://t.co/JIFArknaX1









Wondering how agents will play games? They’ll likely rely on Hierarchical Task Networks (HTNs). These networks break complex goals into smaller tasks arranged in a hierarchy, where each level defines subtasks or methods to achieve the parent goal. It lets the agent think in “long term,” “medium term,” and “short term.” Long Term: Grand strategy or “win conditions” (e.g., dominate the @RealmsEternum map). Medium Term: Larger milestones (e.g., build a fortress or recruit an army). Short Term: Immediate tasks (e.g., gather resources, move units). By focusing on specific, actionable steps while still tracking broader objectives, HTNs help AI plan more effectively and adaptively. This is the focus of daydreams library which will have a plugin directly into @ai16zdao and other frameworks like @arcdotfun This is true agency and what every agent in the future that wants to do complex tasks will implement If you want to see this in action first? Follow us at @RealmsEternum where agents will use $lords and HTNs to attempt complex problems. 1000 will exist in Season 1 in a few months.













