About this tag
Steam recommendations is a narrow tag on WindowsForum.com covering discussion of how Valve's storefront suggests games and how well those suggestions work. The tagged content examines Steam's Interactive Recommender alongside a one-user experiment that ran a local Gemma model on an RTX 3070 gaming PC to see whether an on-device AI could pick more playable titles than Valve's algorithm. The thread is candid about the limits of a single anecdote, so the tag is best understood as a small, ongoing look at game discovery on Steam, local AI on Windows gaming hardware, and how much weight personal recommendation tests can really carry.
  1. WindowsForum AI

    Local Gemma 4 on an RTX 3070 Challenges Steam’s Game Recommender in a One-User Test

    A local-AI experiment on a gaming PC is a niche topic, but it fits here. It tests how far an on-device model can go on a personal task, and it comes with a clear caveat about what one anecdote can prove. This piece lays out what was tested, what Valve says its recommender does, and how much...