Adeia

The 2 Streaming patents filed in H1 2026 tackle two distinct friction points in the gaming experience. The first describes a system that lets any viewer scan a screen showing a livestream and immediately join the game session at the exact state displayed, with no account login or prior synchronization required. Rather than relying on QR codes or Platforms authentication, the technology captures frames passively, embeds game-state metadata directly, and reconstructs the session statelessly on the joining device. The second patent addresses a different moment of friction, when players get stuck and have to stop playing to hunt for help. This filing describes a system that monitors real-time performance metrics during gameplay and automatically surfaces contextually matched tutorial videos without the player ever leaving the game, including stitching together clips across sub-levels and continuously pruning the video database as player performance data evolves.

Adeia's 2 AI & Machine Learning patents both apply machine learning to problems of game session management and character behavior, though in quite different directions. One filing embeds time constraints directly into a Monte Carlo Tree Search algorithm, allowing an AI opponent system to fit a session into a user-defined time window while keeping difficulty consistent throughout, rather than switching models mid-session or making jarring real-time difficulty adjustments. The other describes a personality system for non-player characters that uses a relationship graph to propagate behavioral traits across an entire game world, then generates dialogue through a large language model, with knowledge base constraints tied to player progression preventing characters from responding outside their defined scope.

Cloud gaming accounts for 3 patents in this period, with 2 of them converging on the same core engineering challenge: encoding video at ultra-low latency without sacrificing quality. Both filings describe architectures that run multiple encoder instances in parallel across distributed quantization parameter ranges, using a variational autoencoder to predict scene complexity before encoding begins, which eliminates the trial-and-error of traditional rate control. The third cloud gaming patent takes a different angle, describing a system that identifies non-interactive gameplay moments such as cutscenes and replaces them at the encoder level with targeted video ads selected using real-time player data and network conditions, requiring no integration on the client side.

The single Monetization patent filed in H1 2026 addresses the timing problem at the center of in-game advertising. Rather than inserting ads at fixed intervals, the system continuously quantifies player mood and attention using engagement and satisfaction metrics, then schedules supplemental content for moments when a player is most likely to be receptive. The filing also describes matching ad tone to player state, surfacing upbeat content after accomplishments and comfort-oriented content following failures, while holding off entirely during moments of high active engagement.