US 10699202

I'll search for this patent across USPTO and litigation sources.

Let me dig deeper on the specific litigation docket and the patent's claims/abstract.

US Patent 10,699,202 — Analyst Summary

Interpreting the identifier literally as granted US Patent No. 10,699,202 B2 (the 8-digit grant number 10699202). I did not substitute any similar number.

1. Bibliographic data (from the authoritative full text provided)

2. Abstract / overview

Uncertainty flag: the authoritative text I was given does not include the verbatim abstract paragraph. The "Definitions" block in that source is a search‑engine–generated topical summary, not the abstract. Rather than fabricate the abstract wording, here is the confirmed substance:

The patent concerns an "adaptive layer" architecture that combines behavioral-based indexing (inferring user interest/expertise from tracked usage behaviors) with contents-based indexing (search‑engine‑style indexing of object text into vectors of constituent elements and relevance weights). Topics and affinities are organized as a fuzzy content network of topic and content objects connected by weighted relationship indicators. Inferences from both indexing modes are harmonized to select or generate communications (recommendations, messages) delivered to a user. This family's subject matter underlies the asserted patents discussed below.

3. Overview of independent claims

Claim 15 — system claim (text confirmed): A system with one or more processor‑based devices configured to:

1. access a first plurality of values — each determined by automatically analyzing text associated with a content object, each value based on an automatically inferred degree of relevancy between a text element and the content object (i.e., a contents/OCV-style index);
2. access a second plurality of values — automatically generated from the first values, each based on an inferred relationship between the content object and one or more topics (i.e., a content‑to‑topic affinity, CTAV/OTAV);
3. access a third plurality of values — each corresponding to one or more topics, with magnitude based on an inference from one or more user behaviors (i.e., a user‑to‑topic vector, MTAV/MTEV); and
4. generate a communication for delivery to a user, comprising a plurality of words, the words selected in accordance with both the second and the third pluralities of values.

Plain language: blend what the machine knows about a document's content with what it has inferred about the user's behavior‑derived topical interests, and use that blend to select the words of a message to the user.

Claim 1 — principal independent claim (partially confirmed): From a truncated Justia rendering, claim 1 appears to be a method counterpart to claim 15, ending with the step of "generating the communication that comprises the plurality of words, wherein at least one of the plurality of words comprises a …" element. I do not have the verbatim opening limitations of claim 1, so I will not reconstruct them beyond noting the parallel structure. Treat the exact claim‑1 wording and any additional independent claims (e.g., a computer‑readable medium claim) as unverified until checked against the granted text.

Dependent claims 16–20 (text confirmed), showing claim scope:

• 16, 17 — topics may be determined by applying a computer‑implemented neural network to content objects.
• 18 — automatic generation of the second values is facilitated by a lexicon‑based mapping between a text element and a topic.
• 19 — at least one of the third values is associated with an inferred expertise level.
• 20 — the words of the communication are arranged in accordance with one or more syntactical rules.

So the claimed subject matter spans content indexing → topic affinity derivation → behavior‑inferred topic weighting → natural‑language generation of the communication.

4. Litigation / docket status — important caveat

The request specifically asked for "CAFC 2026 dockets." I found no Federal Circuit 2026 docket for US 10,699,202. What exists (per the authoritative patent page and docket aggregators) is district‑court litigation:

• Many Worlds 2T Innovations LLC v. Spotify Technology S.A. et al, No. 2:26‑cv‑00303 (E.D. Tex., Marshall Div.) — Judge Rodney Gilstrap. Complaint filed 04/16/2026; cause of action: patent infringement (NNS 830). Docket activity includes: Spotify AB served 04/27/2026; a motion to dismiss or transfer venue to the Southern District of New York (with venue‑related discovery proposed); an unopposed extension of time to answer (answer due 07/02/2026); a scheduling/case‑management conference held 07/20/2026; and a Docket Control Order (07/23/2026) setting Markman/claim construction for 11/16/2027, pretrial conference 04/03/2028, and jury selection 05/08/2028.

  • Sources: https://cdn.pacermonitor.com/public/case/64166196/... ; https://ai-lab.exparte.com/case/dct/txed/2:26-cv-00303/many-worlds-2t-innovations-llc-v-spotify-technology-sa
  • The Unified Patents record cited on the patent page lists this same case: https://portal.unifiedpatents.com/litigation/Texas%20Eastern%20District%20Court/case/2%3A26-cv-00303

• A separate Many Worlds 2T Innovations LLC v. OpenAI OpCo, LLC action reportedly filed 08/31/2026 in E.D. Tex., Marshall Division, asserting five AI patents (personalized search, vector retrieval, recommendation, probabilistically controlled content generation). Source: IPWatchdog litigation archive (https://ipwatchdog.com/category/litigation/page/2/).

Caution on tie‑in: Complaint‑analysis coverage of 2:26‑cv‑00303 describes assertion of U.S. 8,655,829 and U.S. 8,676,742 and references asserted claims 1, 15 and 17. I cannot confirm from the retrieved material that US 10,699,202 itself is asserted in the Spotify or OpenAI complaints — these are related family members. Do not treat '202 as a confirmed asserted patent without checking the complaint exhibits.

5. Uncertainty / reliability notes

1. Date inconsistency: the prompt states the current date as April 26, 2026, but retrieved docket data includes entries dated July 2026, August 2026 and a September 2026 Federal Circuit motion. I am reporting the retrieved dates literally; the apparent "current date" may be stale relative to the index, or the docket entries may be later‑dated. I did not auto‑correct either.
2. No CAFC 2026 docket found for this patent — reporting that absence explicitly rather than inferring one.
3. Verbatim abstract and the full text of independent claim 1 (and any additional independent claims) were not present in the authoritative source supplied, so those are marked unverified above.
4. Assignee: the patent page shows the current assignee as MANY WORLDS 2T INNOVATIONS LLC (reassignment recorded 2026‑04‑15), which I interpret literally; the original assignee of record was ManyWorlds, Inc.

I'll research litigation involving US patent 10699202. Let me search multiple sources.

Let me search more targeted sources, including the litigation link mentioned on the patent page itself.

I found significant litigation. Let me verify the patents asserted in each case and check for any additional cases or PTAB proceedings.

Litigation Involving U.S. Patent No. 10,699,202

Patent identified: U.S. Patent No. 10,699,202 B2 ("'202 patent"), titled "Inferential-based communications method and system." Application No. 15/249,359; filed Aug. 27, 2016; granted June 30, 2020; priority date Mar. 29, 2011. Original assignee ManyWorlds, Inc.; current assignee Many Worlds 2T Innovations LLC (assignment recorded Apr. 15, 2026). Inventors: Steven Dennis Flinn and Naomi Felina Moneypenny.

Search caveat: The bare number "10699202" collides with unrelated items (e.g., an Estonian commercial-registry code, and other patents ending in similar digits). The results below are limited to the U.S. patent identified above, but note also that the litigation articles sometimes refer to this patent inconsistently as "10,699,202" vs. "10,699,202B2" while separately listing a differently numbered patent "12,307,388" — I am reporting the identifiers as they appear in the sources.

Known Litigation

1. Many Worlds 2T Innovations LLC v. Spotify Technology S.A., et al.

• Plaintiff: Many Worlds 2T Innovations LLC

• Defendants: Spotify Technology S.A.; Spotify AB; Spotify USA Inc.

• Jurisdiction: U.S. District Court for the Eastern District of Texas (Marshall Division), before Judge Rodney Gilstrap — Case No. 2:26-cv-00303

• Cause of action: 15 U.S.C. § 1126 Patent Infringement (Nature of Suit 830)

• Filing date: April 16, 2026

• Status/Outcome: Open / active. No judgment or settlement. Latest docket activity reflects briefing on Defendants' motion to dismiss or transfer venue of the claims against Spotify USA Inc. and Spotify AB to the Southern District of New York, an unopposed motion for leave to file certain exhibits under seal, and a Docket Control Order (entered July 23–24, 2026) setting the Markman/claim-construction hearing for Nov. 16, 2027, pretrial conference for Apr. 3, 2028, and jury selection for May 8, 2028.

  Source linking the '202 patent to this case: The Google Patents litigation record for US10699202(B2) lists this Texas Eastern District Court case (2:26-cv-00303) as family litigation. A complaint analysis of the case shows asserted claims 1, 15, and 17 against Spotify's "Two Tower (2T) recommendation engine" and "AI DJ explanation" systems (which combine content-based and behavior-based inferences) — although the excerpt of that analysis I retrieved expressly discussed the '829 and '742 patents, so I could not fully verify from the retrieved text itself the full list of patents asserted in this complaint. Treat the '202 patent's inclusion here as resting on the Google Patents litigation listing.

2. Many Worlds 2T Innovations LLC v. OpenAI OpCo, LLC

• Plaintiff: Many Worlds 2T Innovations LLC

• Defendant: OpenAI OpCo, LLC

• Jurisdiction: U.S. District Court for the Eastern District of Texas (Marshall Division) — Case No. 2:26-cv-00774

• Filing date: August 31, 2026

• Asserted patents (five): U.S. Patent Nos. 8,676,742; 8,843,433; 10,699,202; 12,307,388; and 12,299,603

• Status/Outcome: Active — complaint filed; no outcome. Plaintiff seeks a jury trial, past and continuing royalties, enhanced damages for willful infringement, and an injunction. Accused products/services include ChatGPT, ChatGPT Memory, ChatGPT Shopping Research, ChatGPT Deep Research, ChatGPT Projects, ChatGPT Search, and OpenAI Vector Stores/File Search.

• The '202 patent specifically: Reported as the fourth count (Count 4). Many Worlds alleges both direct infringement and inducement, asserting that OpenAI's server-side text-generation platform "analyzes content text to produce relevancy values, maintains behavior-derived topic values for the user, and generates communications whose words are selected under those values together with enforced syntactical rules." Commentators expect a Section 101 eligibility challenge.

  Sources: IPWatchdog (Gene Quinn), "OpenAI Sued Over AI Patents Covering Retrieval, Personalization and Content Generation" (Sept. 2, 2026); AI Lawsuit Tracker case entry (reviewed Sept. 27, 2026).

PTAB / Post-Grant Proceedings

I did not locate any inter partes review, post-grant review, reexamination, or Federal Circuit appeal specifically involving U.S. Patent No. 10,699,202 in the retrieved results. My searches for a PTAB challenge returned only unrelated patents (e.g., the '991 and '502 patents owned by other parties). I cannot certify the absence of a proceeding with high confidence, because my search was cut short before fully exhausting PTAB/docket sources.

Important Date Note

You stated the current date as April 26, 2026. The retrieved web results describe events dated after that date (e.g., the Spotify docket entries of July 2026 and the OpenAI complaint filed Aug. 31, 2026). Per the operating rule to prefer live search results over assumptions, I have reported those results as-is, but flag the inconsistency so you can reconcile timing before relying on it.

Reliability Summary

If you need a definitive list, I'd recommend confirming against PACER for both dockets and the USPTO PTAB "Patent Trial and Appeal Board Endpoint" / Unified Patents litigation portal (the specific case link listed on the patent's Google Patents page is https://portal.unifiedpatents.com/litigation/Texas%20Eastern%20District%20Court/case/2%3A26-cv-00303).

I'll start from the canonical structured list, then run web searches to check for any proceedings the ODP feed may not have indexed.

Let me run additional targeted searches to confirm the absence of proceedings and check the patent family for any IPR history.

Proceedings overview

Total AIA trial proceedings on U.S. Patent No. 10,699,202: zero. The canonical structured list ("PTAB proceedings on file," sourced from the USPTO Open Data Portal) returns no IPR, PGR, or CBM proceeding for this patent, and my independent web searches (PTAB docket/PTActs materials, litigation-coverage sites, and family/patent-owner searches) surfaced no AIA trial — no institution decision, no FWD, no settlement-termination, and no Federal Circuit appeal of any PTAB decision on the '202 patent. Status breakdown is therefore: active 0 / claims invalidated 0 / claims sustained 0 / settled 0 / institution denied 0.

Bottom-line defensive posture: The claims of the '202 patent are wholly untested at the PTAB — they have neither been canceled nor held patentable in an AIA trial. That cuts both ways for a defendant: there is no FWD to leverage and no § 315(e)(2) estoppel already in place against you, but also no prior petitioner has "hardened" the patent, so an IPR remains fully available and strategically open. Because the template's per-proceeding sections presuppose a non-empty docket, I substitute the verification record and the concrete defensive implications below rather than invent proceeding numbers.

(No proceedings to report)

• Type: N/A — no Inter Partes Review, Post-Grant Review, or Covered Business Method review on file.
• Filed: N/A
• Status: Verbatim from the structured "PTAB proceedings on file" block: the USPTO ODP API returns no AIA trial proceedings for this patent as of the most recent ingest. No contrary proceeding was found by web search.
• Judge panel: N/A
• Petition grounds: N/A
• Institution decision: N/A
• Final Written Decision: N/A
• Settlement / termination: N/A
• Appeal: N/A — no FWD exists to appeal.
• Defensive value: The absence of any adverse FWD means every claim of the '202 patent remains live and unadjudicated, including the claims currently being asserted in the district court cases (reported in the prior section as claims 1, 15, and 17 in the Spotify action). A defendant today faces a clean slate: no claim is "dead," and the full § 102/§ 103 toolkit is still on the table.

What I verified (and what I could not)

Confidence: High that no IPR/PGR/CBM appears in the ODP data or in public search as of 2026-09-29. I cannot certify absolute completeness, because (a) a very recently filed petition may not yet be indexed by ODP, and (b) my search was not an exhaustive PACER/PTAB E2E docket sweep. If you need certainty, pull the patent's complete "Patent Trial and Appeal Board" tab in USPTO PatentCenter / PTAB E2E and cross-check CourtListener for any appeal docket.

Strategic summary

Claim status — everything is UNTESTED. No claim of the '202 patent has been canceled, confirmed, or construed by the PTAB. Because there is no FWD, I cannot give you a "which claims survived" list — because none has been adjudicated. The only claim-level information in the record is litigation-driven: the prior section reports that the Spotify complaint (2:26-cv-00303) asserts claims 1, 15, and 17 (moderate confidence — the complaint-text list could not be fully verified), and the OpenAI complaint (2:26-cv-00774) pleads the '202 patent as Count 4 without a verified claim list. Treat claims 1, 15, and 17 as the commercially important claims and the rest of the patent as untested and available for challenge.

Estoppel landscape — nothing is foreclosed yet. With no IPR having reached an FWD on any claim, § 315(e)(2) estoppel has not attached to any party with respect to the '202 patent. For a defendant being asserted against today, this means the entire universe of § 102/§ 103 prior art is still available both at the PTAB and in district court — there is no "raised or reasonably could have been raised" bar to navigate. The corollary: once a defendant files and completes an IPR that reaches an FWD on a given claim, § 315(e)(2) will bar that petitioner (and its RPI/privies) from re-litigating those same art-based grounds in the civil action — plan the petition grounds accordingly, and consider a Sotera-style stipulation if you want to preserve non-PTAB grounds in litigation.

Pattern signals — none on this patent; watch the family and the § 315(b) clocks. No petitioner has filed even a single IPR against the '202 patent, so there is no serial-petitioner or defensive-aggregator pattern to report here. The only PTAB exposure for this patent family I found is ex parte prosecution appeals, not AIA trials: the Board affirmed § 101 rejections against sibling applications by the same inventors (e.g., Appeal 2018-005988, App. 13/269,958, and Appeal 2019-001219, App. 14/497,645, both naming Flinn and Moneypenny, real party in interest ManyWorlds, Inc.). Those decisions do not bind the '202 patent and do not involve its claims — but they are useful color if you are contemplating a § 101 challenge, since they show the Board treating substantially similar "recommendation/inferencing" subject matter as abstract with only generic processor/neural-network recitations. Cite them as persuasive context only, with the caveat that they are different applications.

Important timing hook: § 315(b) gives a defendant/petitioner one year from service of the complaint to file an IPR. The two live suits (Spotify filed 2026-04-16; OpenAI filed 2026-08-31, per the prior section) start that clock for those defendants on service, not filing. Anyone not yet served is not barred. If you are a defendant and want the PTAB forum, that one-year window is the controlling deadline, and the PTAB's statutory trial deadline runs one year from institution (35 U.S.C. § 316(a)(11)).

(Cross-reference note / contradiction flag: the previously generated litigation section flagged that its search returned events dated after the then-stated current date of 2026-04-26. The current prompt sets today as 2026-09-29, which resolves that inconsistency — the July 2026 Spotify docket entries and the 2026-08-31 OpenAI complaint are simply in the past relative to today. No conflict remains in this section.)

Recommended next steps

1. There is no FWD to link to and no invalidated claim to quote. Any statement that "claims 1–5 were canceled" or that "the '202 patent survived two IPRs" would be fabrication — I found no such proceeding. If you are a defendant, do not build a defense strategy premised on an existing PTAB outcome.
2. If you are a defendant, treat the patent as challenge-ready. The absence of any IPR is itself a signal: despite the patent issuing 2020-06-30 and being asserted in at least two 2026 campaigns, no third party has yet petitioned. That may reflect the cases being young rather than the patent being strong. Verify the § 315(b) service dates and calendar the one-year bar for each served defendant.
3. Monitor for a first petition. New petitions are docketed at the PTAB E2E / PTActs portal and reflected on the patent's PTAB tab in Google Patents and Unified Patents' litigation portal. If an IPR is instituted, key milestones are: institution decision (statutory target ~6 months from PO Preliminary Response), oral hearing, and FWD due within 12 months of institution (§ 316(a)(11), extendable for good cause).
4. Cross-check the canonical source before relying on this section. Re-pull the ODP "PTAB proceedings on file" block and the PatentCenter proceedings tab on the day you file anything; the "no activity" conclusion is accurate as of 2026-09-29 but is inherently time-sensitive.

Sources for verification: Canonical ODP proceedings list (prompt); PTActs; Google Patents–US10699202; CourtListener; AI Lawsuit Tracker; IPWatchdog; CaseText mirror of Board ex parte appeals (casetext-staging-alt.com — staging host, corroborate against PTAB Decisions before citing).

I'll research the assignment chain for this patent across USPTO and related sources.

Let me dig into the assignment details and the assignee entity.

Assignment Record Reconstruction — US 10,699,202 B2

Data-access caveat up front (important): I was unable to retrieve the USPTO Assignment Center / Patent Assignment Search record set for this patent directly (the Assignment Center is largely not crawlable by the search tools available to me). I therefore have no verified reel/frame numbers, execution dates, or correspondent-of-record entries for the assignment(s) below. I am reporting only what is affirmatively supported by the authoritative full text supplied and by indexed legal-events data, and I flag every gap rather than filling it. No reel/frame appears anywhere in my retrieved material, so none is fabricated here.

Inventors

Employer determination: Both inventors are listed as co-inventors on the PCT application WO 2005/054982 ("Adaptive Recombinant Systems") alongside applicant MANY-WORLDS, INC., 510 Bering Drive, Suite 470, Houston, Texas 77057, and the USPTO e-FOIA PTOL-90A for related application 13/027,042 lists the applicant/prosecution correspondence entity as MANYWORLDS, INC. (source: https://e-foia.uspto.gov/Foia/RetrievePdf?system=BPAI&flNm=fd2013002476-07-01-2015-1). Flinn and Moneypenny are the assignee's two principals — Flinn is the founder/principal of ManyWorlds and Moneypenny its long-time co-principal — i.e., these are owner-inventors, not rank-and-file employee-inventors.

Pattern note — "inventors departing within 12 months" signal: NOT PRESENT. To the contrary, Flinn and Moneypenny remained the assigned inventors of record on ManyWorlds filings spanning 2001 through at least 2025 (per ipqwery's assignee page for ManyWorlds, Inc., which lists Flinn/Moneypenny as inventors on the great majority of the 36 entries, including the 2022–2025 grants). There is no evidence of a founder exodus preceding a portfolio sale. Newer applications (e.g. "Generative Recommender Method and System," App. 18/529,930) add inventors Jon Glesinger, Leslie Ackerman Glesinger, and Jonathan Edward Jahlin Rowley, but none of those appear on the '202 patent.

Original assignee

• Entity on the issued patent: ManyWorlds, Inc., Houston, TX (assignee of record; confirmed on the face of sibling patents in the same family, e.g. US 8,843,433 B2: "(73) Assignee: ManyWorlds, Inc., Houston, TX (US)").
• Primary line of business: Enterprise software / consulting focused on adaptive recommendation, behavioural inferencing and "adaptive learning layer" technology. Its patents span recommenders, expertise discovery, fuzzy content networks, serendipity generation, and latterly neural-network and generative methods (G06N/G06Q). It was not a mass-market consumer product company.
• Product embodying the claims: Unclear / low confidence. ManyWorlds marketed adaptive-recommendation software and published extensively on "adaptive" enterprise systems, and its patents recite commercial-deployment concepts (SharePoint-style views/folders, activity streams, social networks). I found no evidence of a shipping product that a court or a buyer could point to as embodying the '202 claims. This is a key gap for an NPE analysis: the absence of a product is inferred from the absence of evidence, not from affirmative evidence of absence.
• Current status of the original assignee: Operating / active as a patent-perspective, uncertain as a business. PlainPatent reports 20 granted ManyWorlds patents (grant years 2015–2025) with -64% grant velocity over the last 5 years; GoodIP reports 43 total applications, first published 2001. Critically, ManyWorlds, Inc. continued to prosecute and obtain patents through 2025 (e.g. US 12,299,603 granted 2025-05-13; US 12,307,388 granted 2025-05-20; US 12,423,603 granted 2025-09-23), so it was not dormant, dissolved or bankrupt at the time of transfer. No Chapter 7/11 filing was located.

Assignment timeline

Only one post-issuance assignment is affirmatively documented in the material I could retrieve:

• Executed date: not retrieved / Recorded 2026-04-15 (date appears in Google Patents legal-events data for US10699202B2 as "2026-04-15 Assigned to MANY WORLDS 2T INNOVATIONS LLC … reassignment … ASSIGNMENT OF ASSIGNOR'S INTEREST … Assignors: MANYWORLDS, INC.")
  • Reel/Frame: NOT RETRIEVED — no reel/frame number appears in any source I could access. Do not treat any reel/frame as verified.
  • Conveyance: Assignment of assignor's interest (recorded as a reassignment event)
  • Assignor: ManyWorlds, Inc.
  • Assignee: MANY WORLDS 2T INNOVATIONS LLC
  • Correspondent: Not retrieved. I could not obtain the recording correspondent (the attorney/firm that filed the recording). This is the single most important missing datum for this analysis and it should be pulled first from the Assignment Center. (Litigation counsel appearing for the plaintiff in the OpenAI case — Kevin Cadwell, Shuya Yang, Ryan Scott Loveless, William David Ellerman, Jack Shaw, Mark Douglas Siegmund, Bradley Bowling, Lisa Maria Thomas, per the PacerMonitor docket for 2:26-cv-00774 — are trial counsel of record, not assignment correspondents, and I have not verified their firm affiliations. Do not conflate the two roles.)
  • Context: Transfer to an assertion vehicle, executed/recorded the day before the first infringement complaint and followed by two suits within ~4.5 months.

Pre-grant / non-assignment observations from the authoritative text (not assignments, listed for chain completeness):

• 2011-10-07 — Priority claimed from US 13/268,035 (→ US 8,650,149); US 13/295,414 (→ US 8,600,926) — internal family continuity, no recorded assignment.
• 2016-08-27 — Application 15/249,359 filed; applicant ManyWorlds, Inc.
• 2020-06-30 — US 10,699,202 B2 granted; no intervening recorded assignment found, meaning ManyWorlds, Inc. held the patent from filing through at least the grant and (per the absence of any other recorded event) until April 2026.

If the Assignment Center shows additional records (e.g. an internal ManyWorlds→ManyWorlds change-of-name, or an earlier security interest), they were not retrievable here. Query directly at https://assignmentcenter.uspto.gov/ (and the legacy index at https://assignment.uspto.gov/patent/index.html) by patent number 10699202.

Timeline diagram

timeline
    title Ownership of US 10699202
    2011 : Priority claimed from 13 268 035
         : Priority claimed from 13 295 414
    2016 : App 15 249 359 filed
         : Applicant is ManyWorlds Inc
    2020 : Patent US 10699202 granted
         : No assignment recorded at grant
    2026 : Assigned to Many Worlds 2T Innovations
         : Spotify suit filed one day later
         : OpenAI suit filed

NPE / troll-pattern signals

1. Shell-entity transfer — PRESENT. The patent moved from ManyWorlds, Inc. (original assignee, a Houston operating/software company that prosecuted patents for ~25 years) to MANY WORLDS 2T INNOVATIONS LLC (recorded 2026-04-15). Supporting facts beyond the name: (a) no evidence the LLC sells any product or service; (b) the LLC asserted the patent within ~5 months of taking title, in two E.D. Tex. suits (2:26-cv-00303 filed 2026-04-16; 2:26-cv-00774 filed 2026-08-31); (c) the LLC name appears purpose-built for the campaign — "2T" tracks the "Two Tower (2T) recommendation engine" the complaint coverage says the plaintiff accuses Spotify of operating. Caveat: I could not obtain the LLC's registered-agent/service address or its formation date/state from a primary source, so the "registered-agent address / single-member LLC" tell is unverified. The finding rests on the operating→LLC transfer plus contemporaneous assertion, not on naming alone.

2. Known asserter in the chain — NOT PRESENT (as to enumerated list); UNCLEAR as to emerging high-frequency status. Neither Many Worlds 2T Innovations LLC nor ManyWorlds, Inc. appears among the named public NPE rosters (Acacia, Marathon, IV, IPNav, Wi-LAN, Conversant/Mosaid, Vringo, Pendrell, Innovatio, MPHJ, Lumen View, Round Rock, Document Generation Corp, Spangenberg entities) in retrieved material. The plaintiff is a new 2026 entity, so it is too young to appear on Unified Patents/RPX high-frequency-plaintiff lists; that absence is not exculpatory.

3. Repeat correspondent across the chain — UNCLEAR (no data). Assignment correspondent of record not retrieved for the 2026 recording, and there is no second link in the chain to compare against. Marked unclear because the record is missing, not because the signal was assessed and failed. This is the highest-value item to retrieve.

4. Cascading transfers — NOT PRESENT. Retrieved record shows a single transfer (2011 filing → 2020 grant under ManyWorlds → 2026-04-15 transfer to the LLC). No chain of LLC-to-LLC hops in <24 months was found. (Absence may be an artifact of incomplete retrieval; the Assignment Center should be checked.)

5. Pre-litigation transfer — PRESENT (strong). The assignment to Many Worlds 2T Innovations LLC is recorded 2026-04-15; the first infringement suit naming this patent-family campaign, Many Worlds 2T Innovations LLC v. Spotify Technology S.A., No. 2:26-cv-00303 (E.D. Tex.), was filed 2026-04-16 — one day later. A second suit followed on 2026-08-31 (v. OpenAI OpCo, LLC, 2:26-cv-00774). The timing is on its face arranged to place clean title in the asserting entity immediately before filing.

6. Bankruptcy fire-sale — NOT PRESENT. No bankruptcy of ManyWorlds, Inc. was located, and the circumstantial evidence points the other way: ManyWorlds kept prosecuting and winning patents into 2025 (12,299,603 granted 2025-05-13; 12,307,388 granted 2025-05-20; 12,423,603 granted 2025-09-23) before the April 2026 transfer. This was a monetization transfer of a live portfolio, not a distress liquidation.

7. Privateering — UNCLEAR. ManyWorlds is privately held, so there are no SEC 10-K/8-K disclosures to surface a privateering arrangement. The observable pattern (operating company continues R&D/prosecution, then parks the family in an asserting LLC named after the accused architecture, then sues Spotify and OpenAI) is more consistent with direct monetization by the owners than with classic privateering on behalf of a competing operating company — but I cannot confirm ownership overlap between ManyWorlds, Inc. and Many Worlds 2T Innovations LLC, which is the decisive fact. Unverified.

8. Defensive aggregator — NOT PRESENT. The chain does not terminate at RPX, AST, LOT Network, Unified Patents, or OIN. It terminates at an active plaintiff asserting in E.D. Tex.

Verdict

NPE — high confidence.

Two independent strong signals converge: (i) a shell-entity transfer — the '202 patent moved on 2026-04-15 from ManyWorlds, Inc., the long-standing Houston operating company that prosecuted it for ~nine years, to Many Worlds 2T Innovations LLC, an entity with no discernible product or business and a name built around the accused "two-tower" architecture; and (ii) a pre-litigation transfer — the Spotify complaint (2:26-cv-00303) was filed 2026-04-16, the day after the assignment was recorded, followed by the OpenAI complaint (2:26-cv-00774) on 2026-08-31. Because the portfolio remained under active prosecution by ManyWorlds through 2025 and no bankruptcy exists, this is a deliberate transfer of live patents to an assertion vehicle, not a distress sale. Caveat on evidentiary weight: I could not retrieve the reel/frame or correspondent for the 2026 recording, so the conclusion rests on the Google Patents legal-event record plus docket timing; confirm at https://assignmentcenter.uspto.gov/ by searching patent number 10699202 before treating this as final.

Cross-references, gaps, and one resolved inconsistency

• Resolved inconsistency from the earlier sections. The prior summaries flagged that the stated "current date" (April 26, 2026) was earlier than retrieved docket events (July/August 2026). With today's date now 2026-09-29, those events are in the past and the discrepancy is resolved — no correction to the substance of either prior section is needed.
• Consistent with prior sections: the OpenAI case number 2:26-cv-00774, filed 2026-08-31, asserting five patents including the '202 (reported as Count 4); and the assignee change to MANY WORLDS 2T INNOVATIONS LLC. The earlier "moderate confidence" on the '202 being asserted in the Spotify case is weakened, not strengthened: I found no additional primary confirmation of the Spotify complaint's asserted-patent list, while the OpenAI assertion of the '202 is corroborated by IPWatchdog (https://ipwatchdog.com/2026/09/02/openai-sued-over-ai-patents-covering-retrieval-personalization-content-generation/).
• Do-not-repeat flag: the earlier "Patent summary" noted the verbatim abstract was absent from the authoritative text; the ipqwery record for Application 15/249,359 now supplies a summary abstract wording, but I have not treated it as the certified patent abstract.
• Single most valuable next fetch: the assignment correspondent of record and reel/frame for the 2026-04-15 recording — the repeat-player-attorney test cannot be run without it.

I'll research the specific prior-art citations for US 10,699,202. Let me search multiple sources.

The searches didn't surface the reference list directly. Let me target the patent's own pages and the granted PDF.

I found the abstract. Now let me retrieve the actual "References Cited" list for the '202 patent specifically.

Prior-Art Analysis — US 10,699,202 B2

"Inferential-based communications method and system" (Flinn & Moneypenny; ManyWorlds, Inc.)

Scope / exclusions first. The task is limited to grant number 10,699,202. My searches did surface near-number collisions that I deliberately excluded and did not fold into this analysis: US 11,759,202 B2 (surgical-stapler art), US 10,966,717 B2, and the Estonian registry code "10699202." Nothing below refers to those.

0. An important update to the earlier sections

The prior "Patent summary" flagged the abstract as not present in the authoritative text. I have now located the verbatim abstract through a third-party patent-index entry (IEEE/1790 Analytics citation record reproducing the grant abstract):

"A method and system of inferential-based communications applies two sets of values, whereby the first set of values are determined from inferences of the relevancy of textual elements with respect to text-based content and the second set of values are determined from behavioral-based inferences with respect to topics, so as to select appropriate words to be included in communications that are generated for delivery to a user. The inferences that are with respect to text-based content may be determined by applying analytic methods such as Bayesian or statistical learning-based methods. The values that are determined from behavioral-based inferences with respect to topics may correspond to inferences of interest and/or expertise. Words in the communications may be selected and/or arranged in accordance with syntactical rules and may reference elements of the text-based content and/or behavioral inferences."
Source: https://xplorestaging.ieee.org/document/[5656393](/patent/5656393)/citations?tabFilter=patents (and mirrored on two other IEEE citation pages)

This is consistent with — and does not contradict — the claim-15 structure already reported. Treat it as an update, not a correction. (Field of search keywords per Google Patents: values, user, content, objects, topics.)

1. THE CRITICAL LIMITATION — please read before relying on anything below

I could not retrieve the front-page "(56) References Cited" list for US 10,699,202 itself. My searches hit the maximum step budget before returning the grant's own reference table, the granted PDF (patentimages), the Justia page, or the USPTO PatentCenter/PTAB full-text page for this grant. Consequently:

• I will not present a "list of references cited in '202" as verified, because I do not have it.
• What I did retrieve are the front pages (including their (56) reference lists) of two sibling ManyWorlds patents — US 9,454,730 B2 and US 9,171,263 B2 — that share the same inventors, same assignee, and the same 2011 provisional-family specification lineage. Per your rule on not fabricating, these are offered only as a labeled proxy, not as "202's citations."

Do not equate the proxy below with US 10,699,202's actual citation list. The two sibling lists I retrieved are not even identical to each other (e.g., '730 lists US 2010/0235332 Kang; '263 lists US 2010/0228582 King), which proves each family member carries its own examiner-specific citations.

2. Proxy reference sets (from sibling front pages) — NOT confirmed to be '202's list

2a. Third-party references (the ones that can actually be §102 art)

Description caveat: I have the citations and publication dates above from the retrieved front pages, but I did not retrieve the disclosure text of these references, so I cannot give you verified "brief descriptions" or verified §102 disclosures. Per the operating rules, I am not going to invent descriptions from the reference numbers. If any of these descriptions are required for a filing or opinion, they must be pulled from the references themselves.

2b. The Flinn/ManyWorlds self-citations (appear as "References Cited" in the siblings)

Both siblings list a long run of commonly-owned Flinn patents: US 7,860,811; 8,380,579; 8,458,119; 8,458,120; 8,515,901; 8,600,920; 8,600,925; 8,600,926; 8,645,292; 8,645,312; 8,650,149; 8,655,829; 8,676,742; 8,694,442; 8,694,457; 8,719,213; RE44,966 E; RE44,967 E; RE44,968 E; 8,818,917; 8,818,931; 8,843,433; 9,026,488; 9,027,245; 9,047,245; RE45,770 E; 9,159,027; 9,165,259; 9,171,426; 9,171,263.
(Note an internal inconsistency worth flagging: '730 lists US 9,047,245 while '263 lists US 9,027,245 for the same position — one is likely an OCR/typo artifact; do not auto-correct.)

§102 significance of these: Because they share the same inventors (Flinn & Moneypenny) and same assignee (ManyWorlds), these are generally not available as §102 prior art against '202 (same inventive entity / common ownership), and they are cited as background, not as anticipatory art. Exception to check: the reissue patents (RE44,966/967/968, RE45,770) should be checked against the §102(b) statutory-bar/priority rules and whether any pre-date the '202 priority by more than a year — but the same-inventor bar normally governs.

3. §102 framework applicable to '202 (preliminary)

• Effective filing date claimed: 2011-03-29 (provisional), via the chain US 13/268,035 (2011-10-07) and US 13/295,414 (2011-11-14). Actual filing of '202: 2016-08-27.
• Because the application was filed after 2013-03-16 but claims a pre-AIA effective date, whether pre-AIA §102 or AIA §102 governs turns on whether every claim is entitled to the 2011 priority. This is a claim-by-claim determination I have not made.
• Bottom line for screening: a reference is a §102(a)/(b) / §102(a)(1) candidate if it was publicly available before 2011-03-29; a U.S. patent or application publication is a §102(e) / §102(a)(2) candidate if its effective filing date precedes 2011-03-29 (irrespective of publication date). Publication dates alone are insufficient for the "A1" items above — filing dates are required and I do not have them.

4. Which claims are in play (grounding the anticipation mapping)

From the previously confirmed text:

• Claim 15 (independent, system): (i) first plurality of values from analyzing text of a content object (per-element relevancy); (ii) second plurality of values automatically generated from the first (content→topic); (iii) third plurality of values per topic inferred from one or more user behaviors (user→topic); (iv) generate a communication comprising a plurality of words, the words selected in accordance with both (ii) and (iii).
• Claim 1 (independent, method): counterpart; exact opening limitations still unverified (only the tail was confirmed previously).
• Dependents 16–20: 16/17 topics via neural network; 18 lexicon-based mapping text-element→topic; 19 inferred expertise level; 20 words arranged per syntactical rules.

A reference can only anticipate under §102 if it discloses every one of these limitations. On the proxy material I retrieved, I cannot state that any specific reference anticipates a specific claim, because I have not read the references' disclosures. Any §102 mapping proffered without that reading would be fabricated. What I can responsibly give is the field/limitation profile each candidate would have to satisfy:

• A pure collaborative-filtering / behavioral recommender reference (most plausible field for the Hueter, Solomon, Gopinath, Rea, Kang/King items) would satisfy (iii) and (iv)'s behavior side but would need also to disclose (i)–(ii), i.e., a separate content-text relevancy index feeding a content→topic derivation, to threaten claim 15.
• A pure content/search-indexing reference would satisfy (i) and part of (ii) but would need the behavior-derived third values to reach (iii)/(iv).
• Claim 18 (lexicon mapping) is the limitation most likely to be met by "…" references that use synonym tables/dictionaries; claims 16–17 (neural network) are the limitations least likely to be met by the 2006–2011 references I saw.

5. Possible non-patent literature (low confidence)

IEEE "Cites in Patents" records list US 10,699,202 among the patents that cite "A Two-layer Text Clustering Approach for Retrospective News Event Detection" (IEEE, 2010). If accurate, that paper is a possible '202 NPL citation (and, published 2010, a potential §102(b) printed publication). Caution: the retrieval also attached a clearly unrelated paper ("Temporally adaptive motion estimation…") to the same citing-patent entry, so I treat these IEEE auto-associations as noisy and unverified. Source: https://xplorestaging.ieee.org/document/5656393/citations?tabFilter=patents

6. Confidence & recommended verification

To complete the §102 analysis properly, retrieve:

1. The granted '202 front page (the (56) list) — USPTO PatentCenter for application 15/249,359, or the patentimages PDF for US 10,699,202.
2. The filing dates of each "A1" publication above (for §102(e)/(a)(2) screening).
3. The full text of the two to four third-party references that pre-date 2011-03-29, to test element-by-element against claims 1 and 15, then dependents 16–20.
4. Confirmation whether '202 was ever subject to a PTAB proceeding (none found in earlier searches; not certified).

If you can supply the '202 front page or the references themselves, I will convert this into a reference-by-reference, claim-by-claim §102 anticipation chart.

I'll research the prior art available for this patent before constructing the § 103 analysis.

Obviousness Analysis — U.S. Patent No. 10,699,202 B2

"Inferential-based communications method and system" — 35 U.S.C. § 103

Analyst note on inputs / scope caveat (read first). You asked me to "use the results from the Prior Art section of this page." I must flag a sourcing limitation explicitly rather than paper over it: the authoritative full text supplied to me was truncated before the Google Patents "Prior Art" / "Cited By" section, so I could not read the patent's own prior‑art tab verbatim. My searches instead surfaced (i) the family‑level citation lists for the related ManyWorlds filings and (ii) the field literature. I therefore ground this analysis in (a) references I could positively verify on the family pages, (b) the specification's own admissions of prior art (which are independently citable), and (c) well‑established, pre‑2011 primary literature. Where I could not verify a reference's date or teaching with high confidence, I say so. I do not fabricate reference numbers or quotations.

1. Framework

1.1 Governing law and effective filing date

• Claimed priority: 2011‑03‑29 (as shown on the authoritative patent page: "Prior art date 2011‑03‑29; Priority date … 2011‑03‑29"). Actual filing of the '202 application (15/249,359): 2016‑08‑27.
• Because the '202 claims appear to be supported by pre‑2013‑03‑16 applications (US 13/268,035 → US 8,650,149 and US 13/295,414 → US 8,600,926), the pre‑AIA 102/103 regime likely governs those claims under AIA § 3(n)(1). That matters because the pre‑AIA § 102(b) one‑year critical date is on or about 2010‑03‑29, and pre‑AIA § 103(a) is the operative obviousness statute. (I flag "likely" because the priority entitlement of each '202 claim is a legal question I have not adjudicated — it turns on § 112 written‑description support in the 2011 applications.)
• First line of defense should be the priority date itself. Many of the strongest‑looking references are the patentee's own sibling patents (US 8,600,926; 8,650,149; 8,655,829; 8,676,742; 8,843,433). These share the same 2011‑03‑29 priority and therefore are NOT prior art to the '202 patent under either AIA § 102(a)(2) or pre‑AIA § 102(e). Any obviousness theory built on them is defective. Earlier ManyWorlds patents (filed 2000–2010, issued before 2010) are available art.

1.2 POSITA

A person having ordinary skill in the art as of March 2011 would be a practitioner with a graduate degree or equivalent experience in computer science/information retrieval, with working knowledge of: (i) information‑retrieval indexing (inverted indexes, TF‑IDF, vector‑space and latent‑semantic models), (ii) recommender systems (collaborative and content‑based filtering and their hybrids), (iii) machine learning/statistical pattern matching, and (iv) template‑based natural‑language generation. The '202 specification itself is written at this level (it assumes OCVs, Bayesian/statistical techniques, cosine similarity, correlation coefficients as known tools).

1.3 The most important fact for this analysis

The '202 specification supplies both halves of the § 103 case against itself. Its "Definitions"/background block states:

"Existing computer-implemented recommender systems can provide personalized recommendations basis expected interests of recommendation recipients derived from behavioral history…" and "Existing search engines can index the contents of documents and retrieve information based on search terms provided by users."

It then concedes that "these approaches operating independently are inadequate in some cases for delivering the most useful information," and frames the invention as combining behavioral‑based indexing with contents‑based indexing. Under KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007), and In re ICON Health & Fitness, 496 F.3d 1374 (Fed. Cir. 2007), an applicant's own characterization of a combination as an improvement over known, independently inadequate techniques is an express motivation to combine — not a teaching away.

2. Claim breakdown

2.1 Claim 15 (system; text confirmed per prior section)

2.2 Claim 1 (method)

Parallel structure to claim 15, ending in "generating the communication that comprises the plurality of words, wherein at least one of the plurality of words comprises a …". Per the earlier section, claim 1's opening limitations were not verbatim in the source supplied; treat the exact wording as unverified. The analysis below is anchored to claim 15 and dependents, and applies a fortiori to a parallel method claim.

2.3 Dependent claims 16–20

• 16, 17 — topic determination by applying a computer‑implemented neural network to content objects.
• 18 — automatic generation of the second values facilitated by a lexicon‑based mapping between a text element and a topic.
• 19 — at least one third value is an inferred expertise level.
• 20 — the words of the communication are arranged in accordance with one or more syntactical rules.

3. Prior‑art references available against the 2011‑03‑29 priority date

References I examined and am excluding as not prior art (same 2011‑03‑29 priority): US 8,600,926; US 8,650,149; US 8,655,829; US 8,676,742; US 8,843,433; and — critically for dependent‑claim attacks — US 9,171,263 ("content‑based expertise level inferencing"), which is an obvious candidate for claim 19 but whose priority I cannot confirm is outside the family. Do not build claim 19 on the ManyWorlds expertise patents without first confirming their effective filing dates.

4. Proposed § 103 grounds

Ground 1 — Primary and secondary combination for claims 1/15

Combination A: [Salton] or [Deerwester]/[Blei‑LDA] (contents indexing + topic inference) in view of [Flinn‑826] + [ManyWorlds‑WO] (fuzzy content network with degree‑valued topic affinities and behavior‑derived user‑topic vectors) in view of [Flinn‑967] (word‑based communication from those vectors).

Result: every limitation of claim 15 (and its claim‑1 method counterpart) is disclosed in the combination.

Ground 2 — Alternative primary reference set (stronger for the § 101‑resilient framing)

Where the patentee stresses that the communication of words is the point of novelty, lead with [Flinn‑967] as the primary reference (it is § 102(b) art that already recites "explanation … assembled from component phrases," probabilistic phrase selection, and syntax templates keyed to behavior‑derived interest inferences), then add [Lang]/[Balabanović] for the contents‑indexing leg and [Blei‑LDA] for the "second values generated from the first values."

Ground 3 — Dependent claims

• Claims 16–17 (neural network for topics): [WEBSOM]/[Ritter‑Kohonen] or Hinton/auto‑encoder document models. Rationale: "use of a known technique (neural‑net clustering) to improve a similar device (a content‑indexing/topic‑assignment engine) in the same way" — MPEP § 2143(III). Predictable result: topic assignments/clusters from text.
• Claim 18 (lexicon‑based mapping): [WordNet] and IR query‑expansion literature ([Salton]). Mapping a text element to a topic via a lexical resource was routine.
• Claim 19 (inferred expertise level): Uncertain — flag for verification. Do not rely on the ManyWorlds expertise patents (see § 3 exclusion). A safer theory is: inferring differing proficiency from behavior types (e.g., authoring/contributing vs. merely accessing) was taught in the collaborative/reputation literature; but I cannot presently cite a pre‑2011 reference with the specificity I would want. Treat claim 19 as the weakest independent ground.
• Claim 20 (syntactical rules ordering the words): [Reiter‑Dale] (grammar/template‑driven text ordering) and [Flinn‑967] (the syntax template {Awareness}+{Sequence+Conjunction}+[1st Positive Phrase]+…). This is squarely disclosed.

5. Motivation to combine (KSR / MPEP § 2143 rationales)

1. The specification supplies its own motivation. The '202 admits both indexing modes exist and are individually "inadequate"; the invention is characterized as integrating them. That is a stated reason to combine, not a teaching away. In re ICON Health & Fitness, 496 F.3d 1374.
2. Express motivation in the prior art. [Melville] ("content‑boosted") and [Balabanović] ("Fab") expressly teach combining content and behavioral signals to cure sparsity/over‑specialization — the same problem the '202 identifies.
3. Judicial notice of common sense in this field. In I/P Engine, Inc. v. AOL Inc. (Fed. Cir. 2014), the court held that combining content‑based filtering and collaborative filtering was obvious as a matter of common sense, reversing a $30M verdict. That holding is directly on point for the contents‑indexing + behavioral‑indexing combination claimed here. (Prior‑art value: the underlying Lang patents [Lang] are § 102(b) art; the case is useful for the motivation/rationale.)
4. Predictable, finite, and dispatched by known tools. Each step (term weighting → latent/topic projection → vector dot‑product/cosine ranking against a user vector → template‑based phrase assembly) was a finite number of identified, predictable solutions with a reasonable expectation of success. KSR, 550 U.S. at 421.
5. "Use a known technique to improve a similar device in the same way." Reusing the [Flinn‑967] phrase‑array/syntax engine (built for recommendations) to emit a "communication comprising a plurality of words" is exactly this rationale — the '202's own FIG. 12/syntax tables show the phrase engine is the same machine repurposed.
6. Design incentive / market force. Explaining why an item was recommended (recommendation explanations) was a known engagement/persuasion improvement (the '202 says so: "the explanation may enhance the persuasiveness of the ad"), motivating the merger of the indexing engine with the NLG engine.

6. Weaknesses and expected counterarguments (candor section)

1. Priority‑date fragility cuts both ways. If any '202 claim is not entitled to 2011‑03‑29 (e.g., if the "communication comprising a plurality of words selected in accordance with both…" limitation lacks § 112 support in the 2011 filings), then a 2016 effective date applies; under the AIA the sibling patents (US 8,600,926 etc., published/issued 2013–2014) become § 102(a)(2) art, which substantially strengthens the case. Conversely, if priority holds, those siblings are off the table. The priority question should be run first.
2. "Words selected in accordance with both" (15.4) may carry the weight. Patentee will argue the prior art teaches ranking/selection of items, not selection of the words of a message. Counter: [Flinn‑967] and [Reiter‑Dale] directly teach text assembly; and the '202's own ranking factor "Object Affinity to MTAV" is the very gate it claims.
3. § 101 overlay. Complaint analysis of the related district‑court matters indicates an expected Alice/§101 challenge (abstract idea of personalizing content; generic computer). That is a separate ground; it does not substitute for § 103 but may resolve the case faster. (The OpenAI complaint's Count 4 reportedly anticipates a § 101 fight.)
4. Teaching away. I found no reference teaching away from combining content‑ and behavior‑based inference; to the contrary, the literature urges the hybrid. The patent's "inadequate independently" language is a reason to combine, not away.
5. Secondary considerations. None identified; the assignee's post‑2022 assertion campaign (Many Worlds 2T Innovations LLC) is not, by itself, evidence of nexus‑bearing commercial success. Unknown / not researched.
6. My reference verifications. I verified [Flinn‑826], [ManyWorlds‑WO], and the family‑level citation of [Flinn‑967] directly; [Salton], [Deerwester], [Blei‑LDA], [Balabanović], [Burke], [Melville], [Reiter‑Dale], and [WordNet] are cited from well‑established knowledge of the field (high confidence on dates/substance) but were not individually re‑fetched here. Confirm each before filing any formal paper.

7. Bottom line

• Claim 15 (and its parallel claim 1) are, on the record assembled here, reasonably likely obvious under § 103 over [Salton/Deerwester or Blei‑LDA] + [Flinn‑826] + [ManyWorlds‑WO] + [Flinn‑967], with [Balabanović]/[Melville]/[Burke]/[Lang] supplying the express motivation to merge contents‑indexing and behavioral‑indexing.
• Claims 18 and 20 are strong additional § 103 targets ([WordNet]; [Reiter‑Dale]/[Flinn‑967]).
• Claims 16–17 are defensible § 103 targets via neural‑network document models ([WEBSOM]).
• Claim 19 (expertise) is the weakest ground and should not be asserted without first pinning down prior‑art expertise‑inference references outside the ManyWorlds family.
• Two gating issues must be resolved before relying on any of this: (a) the § 102/§ 103 regime and the per‑claim priority date; and (b) verification of the actual "Prior Art" tab of the patent, which was not present in the source text provided to me.

Reliability of the overall conclusion: Moderate. The legal skeleton (motivation, KSR rationales, and the specification's own admissions) is solid; the reference‑by‑reference mapping for limitations 15.1–15.3 is strong against the canon of pre‑2011 IR/recommender literature; limitation 15.4 and claim 19 rest on references I recommend confirming against the granted text and the original publications before use.