Many Worlds 2T Innovations LLC v. Spotify Technology SA

I. Executive Summary and Procedural Information

• Parties & Counsel:
  • Plaintiff: Many Worlds 2T Innovations LLC (Texas)
  • Defendant: Spotify Technology S.A. (Luxembourg); Spotify AB (Sweden); Spotify USA Inc. (Delaware)
  • Plaintiff's Counsel: CADWELL THOMAS LLP
• Case Identification: 2:26-cv-00303, E.D. Tex., 04/16/2026
• Venue Allegations: Plaintiff alleges venue is proper because Defendant Spotify USA Inc. maintains a regular and established place of business in the district, has committed acts of infringement there, and registered to do business in Texas. Venue is also alleged as proper for the foreign defendants under 28 U.S.C. § 1391(c)(3).
• Core Dispute: Plaintiff alleges that Defendant's "Two Tower (2T)" recommendation engine infringes five patents related to computer-implemented recommendation systems that use concepts such as fuzzy networks, dual-analysis pipelines, and vector-based search.
• Technical Context: The technology at issue involves sophisticated computer-implemented recommendation systems, which are of critical market significance for digital content providers like music streaming services.
• Key Procedural History: The complaint asserts that the patents-in-suit claim priority to a family of provisional applications filed in 2011. Plaintiff also states that it has not made or sold unmarked articles practicing the patents, which it argues entitles it to collect pre-filing damages for the full period allowed by law.

Case Timeline

Date	Event
2011-03-29	Earliest Priority Date for all Asserted Patents
2014-02-18	U.S. Patent No. 8,655,829 Issues
2014-03-18	U.S. Patent No. 8,676,742 Issues
2014-09-23	U.S. Patent No. 8,843,433 Issues
2016-09-27	U.S. Patent No. 9,454,729 Issues
2020-06-30	U.S. Patent No. 10,699,202 Issues
2026-04-16	Complaint Filed

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 8,655,829

• Patent Identification: U.S. Patent No. 8,655,829, "Activity Stream-Based Recommendations System and Method," issued February 18, 2014.

The Invention Explained

• Problem Addressed: The patent background addresses the limitations of prior recommendation systems, which often rely on simple collaborative filtering (analyzing user behavior patterns) while ignoring the actual characteristics of the content being recommended Compl. ¶1 This led to poor recommendation quality, a "cold-start" problem for new content, and a lack of transparency for the user Compl. ¶32
• The Patented Solution: The invention proposes an "adaptive system" that generates recommendations based on a user's activity stream (a temporally sequenced series of objects) US 8655829, abstract When a user requests a recommendation, a specific object from the stream is selected as the context. A recommender function then generates a recommendation based on inferences of the user's interests (derived from usage behaviors) and a "contextualization" that uses "fuzzy network-based affinities" between the selected object and other objects in the system US 8,655,829, claim 1 This combines behavioral analysis with network-based relationships to provide more relevant results US 8,655,829, col. 2:20-30
• Technical Importance: The technology aimed to improve recommendation relevance by integrating user activity streams with a network-based understanding of content relationships, moving beyond simplistic user-to-user similarity.

Key Claims at a Glance

• The complaint asserts at least independent claim 1 Compl. ¶59
• Claim 1 of the '829 Patent breaks down into the following essential elements:
  • Receiving an activity stream of temporally sequenced computer-implemented objects.
  • Selecting one of those objects as a context for a recommendation based on a direct user request.
  • Receiving a recommendation generated by a recommender function.
  • The recommender function generates the recommendation based on (a) an inference of user interests from usage behaviors and (b) a contextualization comprising "fuzzy network-based affinities" between the selected context object and a second plurality of objects.

U.S. Patent No. 8,676,742

• Patent Identification: U.S. Patent No. 8,676,742, "Contextual Scope-Based Discovery System," issued March 18, 2014.

The Invention Explained

• Problem Addressed: Prior recommendation engines were often "black boxes" that were content-agnostic and unable to recommend new content effectively (the "cold-start" problem) Compl. ¶32 They also lacked the ability to explain recommendations to the user Compl. ¶31
• The Patented Solution: The invention describes a "fuzzy network-based structure" where objects are linked by weighted affinity relationships Compl. ¶33 Compl. ¶35 The system uses a "contextual scope function" to establish a "contextual neighborhood" of related items within this network, and a "discovery function" then generates a recommendation based on this neighborhood and inferences from user behaviors Compl. ¶37 Figure 3 from the patent, included in the complaint, depicts an example fuzzy network with content and topic objects linked by weighted affinities Compl. ¶33, p. 10 The system can also transform data from "originating structures" like hierarchical folders into this flexible fuzzy network Compl. ¶39 US 8676742, Fig. 5C
• Technical Importance: This architecture improves computer functionality by converting rigid, hierarchical data into a flexible, weighted graph structure, enabling more accurate recommendations through graph-based similarity computations Compl. ¶36

Key Claims at a Glance

• The complaint asserts at least independent claim 1 and dependent claim 5 Compl. ¶73
• Independent Claim 1 of the '742 Patent breaks down into the following essential elements:
  • Invoking a "contextual scope function" on a processor to establish a "contextual neighborhood" within a "fuzzy network-based structure."
  • Receiving a recommendation generated by a "discovery function" that uses the contextual neighborhood and an inference from user usage behaviors.
  • The recommendation is generated in accordance with the contextual neighborhood, which is based on a selected object.
• The complaint notes that dependent claim 5 adds the requirement that the fuzzy network be transformed from one or more originating structures Compl. ¶39

Multi-Patent Capsule: U.S. Patent No. 8,843,433

• Patent Identification: U.S. Patent No. 8,843,433, "Integrated Search and Adaptive Discovery System and Method," issued September 23, 2014.
• Technology Synopsis: The technology describes a dual-analysis system that generates recommendations by algorithmically combining two sets of values: a first set based on a user's inferred interest in topics (derived from behavior), and a second set based on the relevance of content to those topics (derived from content analysis) Compl. ¶42 The patent also claims generating natural-language explanations for these recommendations and a behavioral search process using vector representations Compl. ¶¶44, 46
• Asserted Claims: Claims 8, 9, 14, and 17 Compl. ¶87
• Accused Features: The complaint alleges infringement by Spotify's dual-tower neural network (combining user and content analysis), its systems for generating explanations, and its behavioral indexing and vector-based search system Compl. ¶¶87-88

Multi-Patent Capsule: U.S. Patent No. 9,454,729

• Patent Identification: U.S. Patent No. 9,454,729, "Serendipity Generating Method, System, and Device," issued September 27, 2016.
• Technology Synopsis: This patent addresses the "filter bubble" problem where users only receive recommendations for things they have already experienced (Compl. ¶50). The patented solution involves selecting two users with highly similar "affinity vectors" (arrays of topic-interest values), but then identifying "contrasting corresponding topic affinity level values" between them. The system then generates a recommendation for the first user that is weighted toward these contrasting dimensions, thereby enabling serendipitous discovery (Compl. ¶¶51, 53).
• Asserted Claims: Claim 1 (Compl. ¶101).
• Accused Features: The complaint accuses Spotify's system that uses affinity vector analysis to identify similar users with contrasting topic preferences and generates serendipitous recommendations based on those contrasts (Compl. ¶102).

Multi-Patent Capsule: U.S. Patent No. 10,699,202

• Patent Identification: U.S. Patent No. 10,699,202, "Inferential-based Communications Method and System," issued June 30, 2020.
• Technology Synopsis: This patent claims a method for automatically generating a communication (e.g., an explanation for a recommendation) for a user. It accesses a first set of values from analyzing text associated with a content object and a second set of values from analyzing user behaviors. It then selects words based on both value sets and syntactical rules to generate the communication, specifying a technical pipeline for how explanations are constructed (Compl. ¶¶47-48).
• Asserted Claims: Claims 1, 15, and 17 (Compl. ¶115).
• Accused Features: The complaint alleges infringement by Spotify's content analysis, user behavior analysis, and AI DJ explanation systems that combine content-based and behavior-based inferences (Compl. ¶116).

III. The Accused Instrumentality

• Product Identification: The accused instrumentalities are collectively identified as Spotify's "Two Tower (2T) recommendation engine" and the associated "Spotify applications, web players, [and] server infrastructure" Compl. ¶55
• Functionality and Market Context:
  • The complaint describes the 2T model as consisting of two feed-forward deep neural networks Compl. ¶55 A "user tower" processes user demographic and historical interaction data to generate "user embedding vectors," while a "content tower" processes content metadata (e.g., language, genre, descriptions) to generate "content embedding vectors" Compl. ¶55 The model's output vectors are optimized to be close in vector space for content the user has engaged with, which enables recommendations through "nearest-neighbor vector similarity search" Compl. ¶55
  • The complaint alleges that recommendations are "central to its business model and competitive differentiation" and that Spotify's success in the "discovery business" is built on implementing the patented technologies (Compl. ¶2; Compl. ¶3). Figure 5C from the '742 Patent, included in the complaint, illustrates the transformation of data from originating hierarchical systems into the fuzzy network's structure (Compl. ¶37, p. 12).

IV. Analysis of Infringement Allegations

U.S. Patent No. 8,655,829 Infringement Allegations

Claim Element (from Independent Claim 1)	Alleged Infringing Functionality	Complaint Citation	Patent Citation
using a first computer-implemented system to receive an activity stream, wherein the activity stream comprises a first plurality of computer-implemented objects, wherein the first plurality of computer-implemented objects are temporally sequenced;	Spotify's service receives activity streams of user listening behavior, which comprise temporally sequenced content objects Compl. ¶59	¶59	col. 51:1-8
selecting a first computer-implemented object of the first plurality of computer-implemented objects as a context for a recommendation... wherein the selecting of the first computer-implemented object is performed in accordance with a direct request for the recommendation by a user who receives the activity stream; and	The system selects content objects as recommendation contexts based on user requests Compl. ¶59	¶59	col. 51:9-16
receiving the recommendation... generated by a recommender function... based, at least in part, on an inference of the user's interests from a plurality of usage behaviors and a contextualization associated with the context, wherein the contextualization comprises fuzzy network-based affinities between the selected first computer-implemented object and a second plurality of computer-implemented objects...	The system generates recommendations via a recommender function that infers user interests from behavioral patterns and applies "contextualization through weighted network relationships between content objects" Compl. ¶59	¶59	col. 51:17-36

U.S. Patent No. 8,676,742 Infringement Allegations

Claim Element (from Independent Claim 1)	Alleged Infringing Functionality	Complaint Citation	Patent Citation
invoking a contextual scope function executed on a processor-based computing device, wherein the contextual scope function establishes a contextual neighborhood within a computer-implemented fuzzy network-based structure;	Spotify invokes "contextual scope functions that establish contextual neighborhoods within fuzzy network-based graph structures using node2vec algorithms" Compl. ¶73	¶73	col. 51:38-44
receiving a recommendation, wherein the recommendation is generated by a computer-implemented discovery function that generates the recommendation in accordance with the contextual neighborhood and an inference from a plurality of usage behaviors; and	The system receives "recommendations generated by discovery functions operating on these contextual neighborhoods combined with inferences from user usage behaviors" Compl. ¶73	¶73	col. 51:45-51
receiving the recommendation, wherein the recommendation is generated in accordance with the contextual neighborhood, wherein the contextual neighborhood is based on a selected object.	Recommendations are generated based on contextual neighborhoods derived from user interactions with content, which serve as the selected objects Compl. ¶73 Figure 3 from the '742 Patent, included in the complaint, depicts an example fuzzy network with content and topic objects linked by weighted affinities Compl. ¶33, p. 10	¶73	col. 51:52-56

• Identified Points of Contention:
  • Scope Questions: A central question for the court will be whether Spotify's "Two Tower (2T)" model, which uses "dual deep neural networks" to generate "embedding vectors" Compl. ¶1 Compl. ¶55, falls within the scope of the patents' "fuzzy network-based structure." The patents appear to describe a graph-based data model with explicit nodes and weighted edges Compl. ¶33, and the dispute may focus on whether a deep neural network architecture is equivalent to, or an implementation of, this claimed structure.
  • Technical Questions: The complaint alleges that Spotify's use of "node2vec graph learning algorithms" Compl. ¶2 satisfies the "contextual scope function" limitation of the '742 patent Compl. ¶73 A key technical question will be whether the operation of Spotify's algorithm is functionally equivalent to the claimed process of traversing weighted edges from a selected node to establish a neighborhood, as described in the patent Compl. ¶40

V. Key Claim Terms for Construction

• The Term: "fuzzy network-based structure" (from '742 Patent, Claim 1)

• Context and Importance: This term is the core architectural element of several asserted patents. The outcome of the infringement analysis may depend on whether Spotify's deep neural network architecture is considered a "fuzzy network."

• Intrinsic Evidence for Interpretation:

  • Evidence for a Broader Interpretation: The specification of the related '202 patent defines a "fuzzy network" as "a computer-implemented plurality of nodes, with relationships among the nodes that have affinities that are by degree" US 10699202, col. 2:64-67 Plaintiff may argue this broad, functional definition encompasses any system with weighted connections between data points, including a neural network that generates and relates embedding vectors.
  • Evidence for a Narrower Interpretation: The figures and specific embodiments in the patents depict a graph with discrete "content objects" and "topic objects" connected by explicit numerical weights US 8,676,742, Fig. 3 Defendant may argue that the claims, when read in light of these embodiments, are limited to this specific type of weighted-graph data structure and do not cover the more abstract, high-dimensional vector spaces and implicit relationships inherent in a deep neural network model.

• The Term: "contextual scope function" (from '742 Patent, Claim 1)

• Context and Importance: This term defines the mechanism for creating a set of related items for recommendation. The infringement analysis may turn on whether Spotify's method for identifying related items is equivalent to the claimed "function."

• Intrinsic Evidence for Interpretation:

  • Evidence for a Broader Interpretation: The claim language is functional, defining the term by its result: it "establishes a contextual neighborhood" (US 8,676,742, claim 1). Plaintiff may argue that any algorithmic process that identifies a set of related items based on a starting point, such as Spotify's alleged use of "node2vec graph learning algorithms" Compl. ¶2, meets this functional definition.
  • Evidence for a Narrower Interpretation: The complaint itself characterizes the claimed function as "traversing weighted edges from a selected object node to identify related objects within a defined affinity threshold" Compl. ¶37 Defendant may argue that this specific traversal mechanism is a required step and that its own "nearest-neighbor vector similarity search" Compl. ¶55 operates differently and is not a "traversal" of a graph in the manner disclosed in the patent.

VI. Other Allegations

• Indirect Infringement: The complaint alleges induced infringement for all asserted patents. The allegations are based on Spotify providing its application and web player to users with instructions, marketing materials promoting the infringing features, and user interfaces and help documentation designed to facilitate and encourage user interaction with the accused recommendation systems (Compl. ¶63; Compl. ¶77; Compl. ¶90; Compl. ¶104; Compl. ¶118).
• Willful Infringement: The complaint alleges that Spotify has had knowledge of the patents "at least as of the filing of this Complaint" and that its continued infringement will be willful (Compl. ¶67; Compl. ¶81; Compl. ¶95; Compl. ¶109; Compl. ¶123). The inducement counts also allege Spotify acted with "specific intent to induce infringement, or at minimum with willful blindness" (Compl. ¶63; Compl. ¶77; Compl. ¶90; Compl. ¶104; Compl. ¶118).

VII. Analyst's Conclusion: Key Questions for the Case

• A core issue will be one of technical and definitional scope: can the term "fuzzy network-based structure", which is rooted in the patents' descriptions of weighted graph structures with discrete objects and topics, be construed to cover the architecture of Spotify's "Two Tower" model, which uses deep neural networks to generate and compare high-dimensional embedding vectors?
• A key evidentiary question will be one of functional equivalence: does Spotify's recommendation engine, which allegedly uses algorithms like "node2vec" and "nearest-neighbor vector similarity search," perform the specific, multi-step computational processes required by the claims-such as a "contextual scope function" that "traverses weighted edges" from a starting node-or is there a fundamental mismatch in technical operation?
• A further question will be one of patent eligibility: while the complaint argues the claims are concrete technical improvements under 35 U.S.C. § 101 (Compl. ¶¶28-30; Compl. ¶54), a court will need to analyze whether the claims are directed to an abstract idea (like generating recommendations) and, if so, whether they contain a sufficient inventive concept that transforms the abstract idea into a patent-eligible application.