US 11786596

I'll search for authoritative information on this specific patent number and any associated litigation.

Let me retrieve the claim language and confirm filing/priority chain details.

Let me confirm the exact claim set of the '596 patent itself.

US 11,786,596 B2 — Analyst Summary

1. Identification (as literally recorded)

Continuity chain (from the patent's own cross-reference section):

• Continuation of 15/894,085 (filed 2018-02-12) → issued as US 10,743,028 B2
• Which is a continuation of 14/682,462 (filed 2015-04-09) → US 9,900,622
• Which is a continuation of 13/354,465 (filed 2012-01-20) → US 9,008,184

Later family members include US 12,186,395 B2 (app. 18/201,171) and US 2025/0099592 A1 (app. 18/974,008), plus an EP family (e.g., EP 4258663 A3).

2. Abstract (verbatim)

"Methods of encoding and decoding for video data are described for encoding or decoding coefficients for a transform unit. In particular, the sign bits for the non-zero coefficients are encoded using sign bit hiding. Two or more sets of coefficients are defined for the transform unit and a sign bit may be hidden for each set, subject to satisfaction of a threshold test. The sets may correspond to coefficient groups that are otherwise used in multi-level significance map encoding and decoding."

3. Plain-language overview of the independent claims

The patent has three independent claims — a decoding method, a decoder, and a non-transitory processor-readable medium — sharing one core technique. The granted claim 1 (as published for this application in US 2020/0368265 A1 and for US 11,786,596) reads in substance:

• Claim 1 — Method of decoding. For each coefficient group in a transform unit that contains a first non-zero coefficient and a last non-zero coefficient:

1. determine whether the number of coefficients between the first non-zero coefficient and the last non-zero coefficient exceeds a threshold; and
2. if it does, determine a hidden sign bit by summing the absolute values of the non-zero coefficients in that group, and assign a sign to a coefficient in the group based on that sum (i.e., infer the sign from parity rather than reading a transmitted sign bit).

• Claim 4 — Decoder. A processor + memory + decoding application configured to perform the same sequence of operations as claim 1.

• Claim 7 — Non-transitory processor-readable medium. Stored instructions that, when executed, cause a processor to perform the same decoding operations as claim 1.

Dependent claims add refinements such as: assigning a negative sign when the sum is odd; identifying first/last non-zero coefficients in scan order and assigning the sign to the last non-zero coefficient in scan order; and (in the closely related family version) the hidden-sign syntax element optionally being a sign bit of a coefficient in another coefficient group.

Core inventive concept in plain terms: Instead of transmitting one sign bit per non-zero coefficient, the encoder "hides" a sign bit inside the transform-unit data by forcing the parity (even/odd) of the sum of the absolute coefficient values in a group to encode that sign. The decoder reverses this, but only when the group is "large enough" (meets a threshold test measuring the span of coefficients between the first and last non-zero coefficient), so the scheme is applied selectively. Multiple sign bits can be hidden per transform unit because the unit is split into multiple sets/coefficient groups, each potentially hiding its own sign. The encoder adjusts one coefficient level up or down by 1 (with rate-distortion considerations) to make parity match when needed.

4. Litigation / docket status (searched)

No CAFC 2026 appeal or district-court litigation involving US 11,786,596 specifically was found. Searching the highlighted results:

• Unified Patents, Inc. filed an ex parte reexamination against U.S. Patent 11,786,596 on 2026-09-18. Unified's own write-up (dated 2026-09-21) states that the '596 patent "generally relates to methods of encoding video data, and specifically encoding or decoding coefficients for a transform unit using sign bit hiding," and that it "has not been in litigation but is related to patents that have been asserted against streaming applications." Counsel noted: Jon Bowser and Dan Smith (Haynes Boone), and in-house counsel David Seastrunk and Vinu Raj. (Source: unifiedpatents.com/insights/2026/9/21/velos-hevc-video-codec-patent-challenged)

• Velos Media is an NPE whose patents are associated with the Avanci Video patent pool. Related Velos patents (e.g., US 8,964,849; 8,767,824; 9,100,634; 10,390,013; 10,257,520; 9,955,171) have been the subject of Unified IPRs/reexaminations, but those proceedings concern different patent numbers and are not dockets for '596.

⚠️ Date discrepancy to flag: The instructions state the current date is April 26, 2026, but the source describing the '596 reexamination filing is dated September 2026 (filing 2026-09-18; article 2026-09-21). Per the operating rule to prioritize search results over prior assumptions, I am reporting the reexamination as filed 2026-09-18, while noting the inconsistency with the stated "current date."

5. Uncertainty / data-quality notes

• Claim set is drawn from the published application (US 2020/0368265 A1) and the corresponding grant record. I could not confirm from an authoritative USPTO page the exact total claim count of the granted '596 patent; I found claims 1–9 reflected consistently across secondary sources (3 independent: method, decoder, medium; the rest dependent). Treat the exact granted claim numbering as moderately, not fully, certain and verify against the official USPTO PatentCenter face record if precision is required.
• Classification anomaly: The Google Patents/FPO classification listings for '596 include many unrelated herbal/medicinal classes (e.g., A61K31/12, A61K36/185, A61K36/9066, etc.) alongside the expected H04N19/* video-coding classes. These appear to be erroneous or artifact tags and are not relevant to the disclosure. Reported as listed, not corrected.
• I did not find any record of the '596 patent itself being asserted in district court, and no CAFC 2026 appeal docket naming it. Its issuance date (2023-10-17) post-dates the bulk of the Velos/Unified district filings, consistent with Unified's statement that it "has not been in litigation."

Primary sources: patents.google.com/patent/US11786596/en; freepatentsonline.com/11786596.html; patents.justia.com/patent/11786596; unifiedpatents.com (2026-09-21 article); patentimages US 10,743,028 B2 (family member).

I'll search litigation databases and patent portals for any cases involving US 11,786,596.

Let me search more specifically to confirm whether the '596 patent itself has been asserted in any district court case.

Let me confirm the reexamination control number and verify the patent lists in the two Velos district court cases.

Litigation Search — U.S. Patent No. 11,786,596

Bottom line

No district-court litigation asserting U.S. 11,786,596 exists (as of the search cutoff). The patent has never been named in a complaint, counterclaim, or appeal. The only adversarial proceeding against it is an ex parte reexamination at the USPTO filed by Unified Patents. This is an administrative validity challenge, not litigation.

Unified Patents states this directly in its own write-up: the '596 patent "has not been in litigation but is related to patents that have been asserted against streaming applications." (unifiedpatents.com, 2026-09-21)

1. Proceedings directly involving US 11,786,596

⚠️ The requested fields (plaintiff/defendant/jurisdiction/case number) do not map onto this proceeding — a reexamination has no plaintiff, defendant, or court. Reported literally rather than forced into those categories.

2. Litigation involving related Velos patents — NOT '596

These are the cases a researcher will surface when searching "Velos + sign bit hiding." None asserts '596. Two of them assert patents in the same family as '596 (U.S. 9,008,184 and U.S. 12,186,395 are continuations of the same priority chain), which is the likely source of confusion.

ByteDance case status: Answer/counterclaims filed (First Amended Counterclaims, Dkt. 21, 2025-11-13). Velos moved for an anti-interference injunction over parallel ByteDance actions in China (Chongqing No. One Intermediate People's Court, filed 2025-11-24) and Brazil (São Paulo, filed 2025-12-18); TikTok opposed (Dkt. 53). RAND/good-faith declaratory claims are also pleaded. Ongoing.

Disney case status: Defendants (Disney entities, Hulu, BAMTech, ESPN) filed a Motion to Dismiss on 2026-05-25, set for hearing 2026-08-05. Its supporting exhibits attach the file histories of U.S. 9,008,184; 12,186,395; and 11,627,338 — again, not '596. ESPN, Inc. was substituted by ESPN, LLC (order 2026-05-22). Ongoing.

Why the '184 / '395 overlap matters to you: the '596 patent is a continuation of the same 2012-01-20 priority chain as '184 (see the continuity section of the patent summary above). '395 is a later sibling continuation. So while the sign-bit-hiding technology is being litigated, the '596 patent itself is not.

3. Earlier Velos portfolio challenges (for context; none involve '596)

Unified's 2018–2020 IPR wave and later reexaminations targeted different numbers: 8,964,849 (IPR2019-00710); 8,767,824 (IPR2019-00635); 9,414,066 (IPR2020-00352); 9,338,449 (IPR2019-00194); 10,390,013; 10,257,520; 9,955,171; 12,088,843; EP 2 347 592 B1. None is '596. Velos Media's patents are associated with the Avanci Video pool.

CAFC: No appeal naming 11,786,596 was found.

4. Explicit caveats and a contradiction to flag

1. No '596 case number exists to report. Any number you see attached to "'596 litigation" in a secondary source most likely belongs to a sibling patent ('184, '395, '338, '843, '962, '849). Verify against the specific patent number, not the patent family.
2. No reexamination control number confirmed. Stated above; verify via USPTO PatentCenter.
3. Contradiction in the record — flagging per instructions. You state the current date is April 26, 2026, but the operative system date is 2026-09-24, and the search results themselves contain events dated through late September 2026: the Unified reexam filing (2026-09-18), the Unified article (2026-09-21), the Disney May 2026 docket entries, and a CRU grant on the '843 patent (2026-09-10). A search performed on April 26, 2026 could not return a September 18, 2026 filing. I have therefore reported the events by their literal source dates and treated the September 2026 developments as the current ground truth, consistent with the instruction to prefer search results over prior assumptions. If the true current date is April 26, 2026, the September 2026 items should be disregarded as not-yet-occurred — but the evidence does not support that date.
4. Update to the earlier-generated summary section: that section reported only the Unified reexamination and noted no district-court case. This search adds the Disney (2:26-cv-03052) and ByteDance (1:25-cv-00967) dockets as the current Velos enforcement activity. Neither contradicts the earlier finding that '596 itself is unasserted.
5. No litigation = current answer. If you need a negative-clearance opinion, the accurate statement is: "U.S. 11,786,596 has not been asserted in any litigation; it is the subject of a pending ex parte reexamination request filed by Unified Patents on September 18, 2026."

Primary sources: unifiedpatents.com/insights/2026/9/21/velos-hevc-video-codec-patent-challenged; dockets.justia.com docket 2:2026cv03052; dockets.justia.com docket 1:2025cv00967; docketalarm.com (Disney 2:26-cv-03052 docket); mondaq.com (Velos v. TikTok asserted-patent list); ai-lab.exparte.com complaint analyses for both cases.

I'll verify the structured "no proceedings" default against live sources before finalizing.

Let me check family members and the reexamination status, plus any IPR naming the '596 or its parents.

Verification note before the answer

I ran live searches to test the structured ODP default ("no AIA trial proceedings on file"). The searches confirm the default: I found no IPR, PGR, or CBM proceeding naming U.S. 11,786,596. The only USPTO adversarial challenge to the '596 patent is an ex parte reexamination filed by Unified Patents on 2026-09-18 — which is not an AIA trial and is already covered in the prior Litigation section, so I do not repeat it here except to keep it from being mistaken for one.

Proceedings overview

U.S. 11,786,596 is the subject of zero (0) PTAB AIA trial proceedings. Breakdown: 0 active, 0 claims invalidated, 0 claims sustained, 0 settled, 0 institution-denied. There is no Final Written Decision, no institution decision, no termination, and no Federal Circuit appeal on this patent to report — because no trial was ever instituted or even petitioned.

Bottom line for a defendant: the '596 patent is PTAB-unscathed, but that is not the same as PTAB-hardened. No tribunal has ever construed its claims, and no claim has been tested by an Article I panel. The patent is untested, not hardened. Contrast the sibling patents in the same Velos HEVC portfolio, which have been through full IPR trials (and in at least one case, e.g. the '849 patent, survived with all challenged claims intact). If you are being threatened with '596, you are not bound by any prior panel's reasoning on it — but you also cannot rely on any prior holding against it. Its validity is an open question, and IPR remains fully available to you.

(No AIA trial proceeding exists to profile)

Per the task format, each proceeding would normally be profiled most-impactful-first. There is nothing to profile. So that the absence is not confused with a missed search, here is the complete universe of what did not happen, keyed to the family it belongs to:

Family members there are untested too. The PTAB activity practitioners actually associate with "Velos + video coding" sits on different patent numbers — Unified's 2018–2020 campaign (IPR2019-00194 / '449 patent; IPR2019-00635 / '824 patent; IPR2019-00707; IPR2019-00710 / '849 patent; IPR2019-00883 / '634 patent; IPR2020-00352 / '066 patent) — none of which is '596 or any claim-chain relative of it. Do not let a search engine's "Velos + IPR" results pull you onto a sibling docket. ("Velos Media" is the common thread; the patent numbers are not interchangeable.)

Strategic summary

Claim status. All claims of 11,786,596 are UNTESTED before the Board. None is canceled, none is confirmed. Because no FWD issued, no claim has been narrowed by the PTAB, and there is no § 318(b) certificate of cancellation. Practically: claims 1, 4, and 7 (method / decoder / non-transitory medium) are intact as granted and, if Velos asserts '596, you would be litigating all of them from scratch — no free kills, no free passes.

Estoppel landscape. There is no § 315(e)(2) estoppel of any kind against anyone on this patent, because no IPR was instituted. That cuts in your favor: nothing prevents you from petitioning for IPR on any § 102/§ 103 ground based on patents or printed publications. It also means no one else's failed or abandoned challenge is available as a shortcut — but equally, no one has pre-empted your art. Note that the pending ex parte reexamination is not an AIA trial and carries no AIA estoppel (§ 315(e) applies to IPR/PGR, not reexam); a reexam requester is also not subject to the § 315(b) one-year-service bar, which is why Unified can file it despite no litigation. If you (as a real-party-in-interest privy to that reexam requester) later file an IPR, watch for § 325(d) discretionary denial arguments by Velos directed at art already before the CRU.

Procedural availability. The '596 patent claims priority to an application filed 2012-01-20, i.e. a pre-AIA effective filing date, so PGR is unavailable (and the nine-month PGR window from the 2023-10-17 grant closed 2024-07-17 in any event). CBM review is unavailable (sunset for new petitions passed 2020-09-16). IPR is the only remaining AIA trial vehicle — and it is fully open, with no § 315(b) bar currently running because '596 has never been asserted in a district-court complaint. That clock starts the day a complaint is served on you.

Pattern signals. The same defensive aggregator, Unified Patents, is in the chain — it petitioned IPRs against at least a half-dozen Velos patents in 2018–2020 and has now challenged '596 via reexam rather than IPR. That pattern (reexam, not IPR, on an unasserted patent) is consistent with a cost-driven validity attack on a patent Velos had not yet asserted, and it tells you the art exists but has not been through an AIA forum. Velos has historically defended PTAB challenges aggressively (extensive RPI discovery, motions to seal, sur-replies) and has pursued appeals; expect a vigorous defense if you file. No Federal Circuit appeal naming '596 exists.

Recommended next steps

• If you face assertion of '596: there is no FWD to lean on and no canceled claims to quote. Your affirmative defense path is a fresh IPR petition under § 311(b) (patents/printed publications only), filed within one year of service of any complaint on you or a privy (§ 315(b)). Because no prior IPR exists, you will not face General Plastic follow-on/abuse-of-discretion denial, and there is no § 315(e)(2) estoppel constraining your grounds.
• Coordinate with the reexam, do not collide with it. The Unified ex parte reexam (filed 2026-09-18) is pending; confirm its control number and CRU status on USPTO PatentCenter before filing, because art you plan to raise may already be of record there and could invite a § 325(d) argument. (The prior Litigation section already flagged that the reexam control number is unconfirmed — treat any number you see as unverified until pulled from PatentCenter.)
• If you are not yet a defendant: no § 315(b) bar is running, so you can file an IPR pre-emptively. But confirm your RPI position relative to Unified's reexam requester if you share counsel or funding, to avoid a § 315(b)/RPI trap.
• If you need the absence stated for a record or opinion: "As of 2026-09-24, U.S. 11,786,596 has never been the subject of an inter partes review, post-grant review, or covered business method review. No PTAB Final Written Decision, institution decision, or Federal Circuit appeal involving this patent exists. Its only adversarial USPTO proceeding is an ex parte reexamination requested by Unified Patents on 2026-09-18."
• Caveat. The ODP ingest did not surface a proceeding, and my live searches returned none; I cannot affirmatively prove a negative for a very recently filed, not-yet-indexed petition. If you find a proceeding number, verify it against the '596 patent number specifically — sibling/parent numbers ('395, '184, '338, '843, '962, '849, '824, '449, '634, '066) dominate Velos PTAB results and are the likeliest source of a misattribution.

Date note (consistency). Today's operative date is 2026-09-24. This is consistent with the September 2026 events recorded in the prior Litigation section (Unified reexam filed 2026-09-18; Unified article 2026-09-21). The earlier "April 26, 2026" date referenced in the previously generated sections remains the only inconsistency in the record and does not affect this analysis.

Primary sources: USPTO ODP structured block (canonical, "no AIA trial proceedings"); unifiedpatents.com/insights/2026/9/21/velos-hevc-video-codec-patent-challenged; docketalarm.com PTAB dockets for IPR2019-00194, IPR2019-00635, IPR2019-00710, IPR2019-00883, IPR2020-00352 (sibling Velos patents, not '596).

I'll research the ownership chain for US 11,786,596, starting with assignment records and the corporate genealogy.

The chain clearly runs BlackBerry → Velos Media. Let me pin down the specific assignment records (reel/frame, correspondent) for this patent family.

Let me pull the Google Patents legal-events records and look for the specific BlackBerry→Velos assignment reel/frame.

I encountered a tool-step limit mid-research, so the following is built strictly on what the retrieved sources document — with explicit flags where the Assignment Center reel/frame data could not be confirmed. I did not fabricate any reel/frame numbers.

Assignment & Ownership Analysis — US 11,786,596 B2

⚠️ Data-completeness caveat up front. I was unable to retrieve the specific reel/frame numbers or recording correspondents for the BlackBerry→Velos transfer from the sources returned. The USPTO Assignment Center index (assignmentcenter.uspto.gov) did not surface in the search results as a directly-quotable record for this patent, and the family transfer is documented in secondary sources (Unified Patents; PTAB filings; the TikTok complaint) rather than in a reel/frame citation I could read. Where a reel/frame or correspondent is absent below, that reflects a gap in my evidence, not a gap in the record — verify at the source. Everything I do assert is tied to a named source.

Inventors

Basis. The three inventors are named identically across the family ('596 and its parents '184 / '622 / '028). A companion Wang/Yu/He/Campbell application filed 19 Nov 2012 lists "Assignee: RESEARCH IN MOTION LTD" on its face (IEEE Xplore citation record), and RIM's registered address is 295 Phillip Street, Waterloo, Ontario N2L 3W8 — the Waterloo cluster that matches the inventors' residences.

Unusual-pattern check — not present. There is no evidence of inventorship departure preceding the transfer. These were salaried BlackBerry video-coding engineers; the assignee (BlackBerry) sold the portfolio, the inventors did not change. That distinguishes a corporate portfolio divestiture from a "founders-walk-out / IP fire-sale" pattern. Dake He in particular is a career BlackBerry video-compression researcher who appears as an inventor across multiple Velos-held HEVC families (e.g., the '744 sign-bit-hiding EP member below). No flag.

Original assignee

Two different entities answer to "original assignee," and the distinction matters for this continuation:

1. Originating assignee (the entity that owned the invention in 2012): Research In Motion Limited, renamed BlackBerry Limited (corporate name change filed ~2013). RIM/BlackBerry is an operating company — it shipped BlackBerry handsets and BB10/BlackBerry 10 devices and enterprise software that implement H.264/HEVC video codecs, and it was a genuine contributor to the HEVC standard. It also practiced in-role: BlackBerry asserted the sibling '849 patent against Avaya (BlackBerry Ltd. et al. v. Avaya Inc., No. 3:16-cv-2185, N.D. Tex., filed 27 Jul 2016) before transferring it (Unified IPR2019-00710 petition; Velos mandatory notices). Current status: operating; exited handsets (2016), now enterprise software/IoT/security; not in bankruptcy.

2. Applicant/assignee named on the issued '596 patent: Velos Media, LLC (Delaware; Plano/Dallas, TX). Google Patents records Velos Media LLC as both the original and current assignee of the '596. This is the tell that matters: the '596 is a continuation filed 2020-08-10, i.e., after the family had already been conveyed to Velos, so the continuing application was filed by Velos as applicant of record and issued directly to Velos — meaning there may be no separate recorded assignment against 16/989,674 itself. Velos Media is a non-practising licensing/assertion entity (see NPE signals below), formed in Delaware June 2016.

Assignment timeline

Important structural finding. For the '596 patent specifically, I found no recorded post-issuance assignment and no recorded change of ownership on the '596 itself. The ownership change is recorded upstream in the family (against the parents), and the '596 then issued to Velos as applicant. I could not verify the controlling reel/frame. Presenting the documented chain with explicit evidence gaps:

• 2020-08-10 (filing, not an assignment) — No reel/frame available

  • Conveyance: n/a — application filed directly by Velos Media, LLC as applicant (continuation of 15/894,085)
  • Assignor: n/a
  • Assignee/applicant: Velos Media, LLC, Dallas/Plano, TX
  • Correspondent: not determinable from retrieved sources — flag for verification. (Likely the same firm that prosecuted the family; I could not confirm.)
  • Context: standing record preserved — by filing the continuation in its own name, Velos kept a clean, unbroken chain of title to the granted '596, which is why the front page reads "Velos Media LLC" as both original and current assignee.

• ~2017 to ~2019 (execution date not confirmed) — Reel/frame NOT confirmed

  • Conveyance: Assignment (portfolio transfer, BlackBerry → Velos)
  • Assignor: BlackBerry Limited (successor by name change to Research In Motion Limited)
  • Assignee: Velos Media, LLC
  • Correspondent: not retrieved — this is the single most valuable missing datapoint for your task; pull it from the Assignment Center abstract of title on a parent patent ('184 / '622 / '028) rather than on the '596.
  • Context: transfer-to-asserter / licensing-pool consolidation. Documented in two independent places: (a) Unified's write-up — the '849 "was originally assigned to Blackberry before being transferred to Velos in 2017"; (b) Unified's IPR press release — "a family of patents that were originally assigned to Blackberry Ltd. and transferred to Velos Media in 2019" (Unified News & Views). Note the source inconsistency: "2017" vs "2019." Unified's own statements conflict on the BlackBerry transfer year; I am reporting both rather than picking one.

• 2016-06 (entity formation — not an assignment of this patent, but the assignee's origin)

  • Conveyance: n/a — corporate formation
  • Assignee (formed): Velos Media, LLC, Delaware; parent Velos Media Management, LLC (also Delaware, formed 2016-06-21) (Mondaq, 8 Aug 2025).
  • Context: licensing-vehicle formation ahead of the 2017–2019 acquisitions from BlackBerry, Panasonic, Qualcomm, Sony, Sharp.

• 2019-01-18 (commercial, not an assignment)

  • BlackBerry joined the Velos Media licensing platform, pooling its HEVC SEPs (Businesswire / TMCnet press release). Context: pool consolidation, corroborating that the BlackBerry→Velos conveyance is a portfolio deal, not an ad-hoc transfer.

• Downstream (post-2024) — Velos itself changes hands (documented at entity level, not by reel/frame)

  • 2024-12: TechStream, LLC formed in Delaware by Patrick J. Patnode (ex-GE Ventures). Velos Media's site now reads "A TechStream Company"; Velos Media Management, LLC remains the disclosed parent (Mondaq).
  • Context: ownership change of the asserting entity — the '596 is downstream of this reorganization. I found no reel/frame evidencing a patent-level assignment from this event; it appears to be a membership/equity change at the Velos entity level, which would not require a USPTO patent-by-patent recordation for the '596 itself. Flagged as unverified at the patent-record level.

Non-US sibling holdings (useful corroboration of the Velos corporate structure). The EP members of this exact family are held by Velos Media International Limited, not Velos Media LLC — e.g. EP 4 447 447 B1 ("Multiple sign bit hiding within a transform unit," priority 2012-01-20, inventors He/Wang/Yu) and EP 4 686 202 A2/A3, agent of record Grünecker Patent- und Rechtsanwälte PartG mbB (PatentGuru assignee listing). So the family is split across two Velos entities by geography — a standard tax/enforcement structuring tell.

Timeline diagram

timeline
    title Ownership of US 11786596
    2012 : Filed by RIM for inventors Wang Yu He
    2013 : RIM renamed BlackBerry Limited
    2015 : Parent patent 9008184 issued
    2016 : Velos Media LLC formed in Delaware
    2017 : BlackBerry transfers HEVC patents to Velos
    2019 : BlackBerry joins Velos licensing platform
    2020 : Continuation filed by Velos as applicant
    2023 : US 11786596 issued to Velos Media
    2025 : Velos becomes a TechStream company
         : Velos sues ByteDance on sibling patents
    2026 : Unified files ex parte reexam on 596

NPE / troll-pattern signals

1. Shell-entity transfer — PRESENT (with nuance).
The patent moved from an operating assignee (BlackBerry/RIM — handsets, enterprise software, real HEVC R&D) to Velos Media, LLC, a licensing-only entity with no manufacturing, formed in Delaware in June 2016 and operating out of Plano/Dallas, TX. Name suffix "Media/LLC," no products in commerce, single-purpose licensing vehicle. Evidence: Unified's characterization of Velos as an NPE and the documented BlackBerry transfer. Nuance: Velos is not an anonymous paper shell — it is the administrator of a SEP pool backed by Qualcomm, Ericsson, Sony, Panasonic, Sharp (and later BlackBerry), operating under RAND commitments. The transfer is real; the "shell" label overstates the anonymity. Signal met on the transfer fact; the "designed-to-look-unrelated LLC" sub-criterion is not supported.

2. Known asserter in the chain — PRESENT (strong).
Velos Media, LLC is publicly labelled an NPE by Unified Patents — "U.S. Patent 11,786,596, owned by Velos Media, an NPE" (Unified, 21 Sep 2026) — is the subject of a known enforcement program, and is associated with the Avanci Video pool. Velos has now asserted family patents: Velos Media v. ByteDance Ltd., W.D. Tex. 1:25-cv-00967-ADA (filed 2025-06-24; asserted patents include 9,008,184 and 12,186,395 — the '596's parent and sibling) and Velos Media v. The Walt Disney Co., C.D. Cal. 2:26-cv-03052-SRM-MBK (filed 2026-03-20) (Mondaq). It does not fit the enumerated "dirty dozen" (Acacia/Marathon/IV/etc.) by name, but it squarely fits the Unified/RPX high-frequency-assertion category.

3. Repeat correspondent across the chain — PARTIALLY PRESENT (one confirmed recurrence; the key occurrence is unverified).

• Confirmed recurrence (foreign): Grünecker Patent- und Rechtsanwälte PartG mbB (Munich) is the agent of record on multiple Velos Media International Limited EP members of this same family — EP 4 447 447, EP 4 686 202, plus Velos EP '157/'990 analytics and imaging families (PatentGuru). That is the same firm filing for two different Velos entities across many members — a genuine recurrence.
• Litigation/PTAB "correspondent" recurrences (separate from assignment correspondents, but probative of a coordinated program): Brent N. Bumgardner / Nelson Bumgardner Albritton P.C. filed repeatedly for Velos across the IPR wave (IPR2019-00635; IPR2020-00352), and McKool Smith, P.C. filed the TikTok complaint. Unified's side is stable too — Jon Bowser / Dan Smith, Haynes Boone on the '596 reexam.
• The assignment-correspondent recurrence you specifically asked for — UNVERIFIED. I could not retrieve the attorney/firm who recorded the BlackBerry→Velos assignment. This is the single most important item to pull from the Assignment Center (see below). I will not guess it.

4. Cascading transfers — NOT PRESENT within the '596 chain.
This is a single transfer (BlackBerry → Velos), not a chain of intermediate LLCs. No LLC-to-LLC hop, no serial re-recordation, no shared-principal cluster inside the '596's own title record. The downstream TechStream ownership change is at the equity level, not a recorded patent assignment. Signal not met. (Contrast the Acacia/Monarch example that surfaced in my searches, where a single patent runs through five reels in under 24 months — nothing like that here.)

5. Pre-litigation transfer — NOT PRESENT for the '596.
The '596 has never been asserted, so there is no "assignment within 6 months before suit" event. The family's enforcement is against the '184/'395/'338/'843/'962 — and the Velos acquisitions predate that litigation by roughly six years, which is the opposite of the classic last-minute venue/standing transfer.

6. Bankruptcy fire-sale — NOT PRESENT.
The originating assignee (BlackBerry/RIM) did not enter bankruptcy; the transfer to Velos was a voluntary portfolio monetization, not a Chapter 7/11 sale. (The Avaya bankruptcy in the '849 story is the defendant in a BlackBerry suit, not the assignor — do not conflate.)

7. Privateering — UNCLEAR / partially.
This is directionally a pool-monetization variant of privateering: BlackBerry pooled HEVC SEPs into Velos, and Velos now asserts against streaming platforms (TikTok, Disney). But classic privateering (opco funds an NPE to sue competitors) is not evidenced here — Velos is a multi-member pool (Qualcomm, Ericsson, Sony, Panasonic, Sharp, BlackBerry) operating under RAND obligations, which is SEP-pool licensing, a legally distinct posture. Marked unclear; do not overstate.

8. Defensive aggregator — NOT PRESENT.
The chain terminates at an asserter, not at RPX / AST / LOT / Unified / OIN. (Unified appears only as the challenger, post-hoc, via the 2026-09-18 reexam — the inverse of being a defensive end-owner.) Signal not met.

Verdict

NPE — high confidence.

The chain terminates at Velos Media, LLC, a licensing-only entity that (i) acquired this family from operating company BlackBerry/RIM, (ii) is publicly classified as an NPE by Unified Patents, and (iii) is currently asserting sibling patents from the same 2012 priority family against TikTok (W.D. Tex. 1:25-cv-00967, filed 2025-06-24) and Disney (C.D. Cal. 2:26-cv-03052, filed 2026-03-20). That presents two or more strong signals — a transfer from an operating company to a non-practising licensing LLC, and a known asserter occupying the end of the chain. Two integrity caveats that do not change the verdict but should accompany it: (a) the controlling reel/frame and recording correspondent were not verifiable in the sources I could reach, so I cannot cite a reel number the way the standard demands — this is the analysis's weakest link; and (b) Velos's RAND-committed, multi-member SEP-pool structure makes it a licensing entity rather than an anonymous shell, so "NPE" is accurate in the no-products sense while "patent troll" in the pejorative sense is not established by this record.

Two contradictions / gaps to carry forward

1. "2017" vs. "2019" for the BlackBerry→Velos transfer. Unified's own publications give both years for the same '849 family transfer; I report both rather than selecting one. The '596 may have a different effective transfer date from the '849, since the '596's continuation (15/894,085) was filed 2018-02-12 — implying the family was already in Velos's hands by early 2018, which favors 2017 over 2019 for the earliest BlackBerry→Velos conveyance.
2. The prior "Patent summary" section lists Original Assignee = Velos Media LLC. That is not a contradiction — it is literally what Google Patents records — but it is misleading if read as the invention's origin. The originating assignee is Research In Motion Limited / BlackBerry Limited; Velos Media is the applicant-of-record only because the '596 is a post-transfer continuation. Both are correct at different points in time.

Where to close the evidentiary gap (do these first)

1. USPTO Assignment Center — https://assignmentcenter.uspto.gov/ (same underlying records as https://assignment.uspto.gov/patent/index.html). Search US 9,008,184 and US 9,900,622 — the BlackBerry→Velos assignment will be recorded there with the reel/frame and recording correspondent that the '596 record likely omits. Then re-run against 11,786,596 to confirm whether any post-grant assignment exists.
2. Google Patents legal events for US11786596 — the "Landscapes"/legal-events block is where any post-issuance recording would appear.
3. SEC EDGAR — BlackBerry's 10-K/8-K disclosures around the 2016–2019 divestitures will corroborate the conveyance and dollar terms; a 10-K "portfolio divestiture" line would convert the secondary-source transfer into a primary-source one.

I'll search for the front-page citations and prior-art references for US 11,786,596 specifically.

Let me target the patent's own citation list more directly.

I found the actual front-page reference list from Justia and FreePatentsOnline. Let me verify the identities and dates of the most important cited references.

Prior-Art Citation Analysis — U.S. Patent No. 11,786,596 B2

Target: US 11,786,596 B2 ("Multiple Sign Bit Hiding Within a Transform Unit"), app. 16/989,674, filed 2020-08-10, granted 2023-10-17, earliest priority 2012-01-20 (pre-AIA), inventors Wang/Yu/He, assignee Velos Media, LLC.

Scope note. Per your instruction I searched for patent number 11,786,596 specifically and did not fold in sibling numbers. The reference list below is the patent's own front-page "References Cited" (Box 56) and its "Other References" (NPL) list, reproduced from the USPTO front-page reproductions carried by Justia and FreePatentsOnline for 11,786,596. I did not run a fresh independent prior-art search — only the cited-art analysis you asked for. Where I could not verify a cited document's interior content from a source I actually retrieved, I say so rather than guessing; those items are marked [content unverified].

Date note. The prior generated sections already flagged the April-26-2026 vs. 2026-09-24 inconsistency; that remains the only date conflict in the record and does not affect this analysis (all § 102 dates run off the 2012-01-20 priority date, not "today").

1. The critical-date framework (this is what makes or breaks each reference)

Claim 1 of '596 (as granted) recites, for each coefficient group of a plurality of coefficient groups in a transform unit that has a first and a last non-zero coefficient:

1. determining whether a quantity of coefficients between the first non-zero coefficient and the last non-zero coefficient exceeds a threshold; and
2. based on that determination, determining a hidden sign bit by summing absolute values of the non-zero coefficients in the group, and assigning a sign to a coefficient in the group based on the summing result.

⚑ Minor correction to the previously generated "Patent summary." That section paraphrased claim 1 as "the number of coefficients…" and as a positive "determine that… exceeds" limitation. The granted text as reproduced for 11,786,596 recites "whether a quantity of coefficients … exceeds a threshold" (a determining-whether, not a determining-that), and claims 4 and 7 use "quantity." This is not a substantive contradiction, but the literal terms differ ("quantity," "whether") and matter for claim-construction and for element-by-element prior-art mapping.

Because '596 is a continuation whose claims are entitled to the 2012-01-20 filing date of 13/354,465, and because no claim has an effective filing date on/after 2013-03-16, the pre-AIA § 102 regime governs. Practically:

Several cited references fall in the last row and are therefore conditional prior art. I flag each.

2. Front-page reference list (as literally recorded)

2a. U.S. Patent Documents

2b. U.S. Patent Application Publications

2c. Foreign Patent Documents

2d. Non-Patent Literature ("Other References") — the substantive art

2e. NOT prior art (procedural / file-wrapper items on the same list)

These appear in the "Other References" block but are prosecution documents, not printed publications under § 102: EP Examination Reports (12151973.0, 2013-10-24; Indian 133/CHE/2013, 2018-07-11), EESRs (12151973.0, 2012-07-09; 18159781.6, 2018-10-01; 16176581.3, 2016-10-11), JP Notices/Office Actions (2013-006987 / 2013-6987, Oct–Nov 2013), AU Office Action (2013200325, 2013-11-12), CN Office Action (201610370914.6, 2018-08-22), U.S. Office Actions and Notice of Allowance in 14/682,462 (2017-03-02, 2017-08-07, 2017-10-11). They can be cited for what they say about the art, but they are not themselves § 102 references.

3. Element-by-element anticipation verdict for the independent claims

4. The one open question that decides the whole picture

The '596 specification itself reproduces an HEVC-style syntax in which:

signHidden = lastNZPosInCG − firstNZPosInCG + 1 >= tsig
… followed by per-CG sumAbs = Σ |level| and a sumAbs % 2 sign flip on firstNZPosInCG.

That construct is exactly the claim-1 subject matter — and an equivalent construct appears in the HEVC draft residual-coding syntax (as quoted in Qualcomm's own later application material: signHidden = ( lastSigScanPos − firstSigScanPos > 3 && !cu_transquant_bypass_flag ), applied per 4×4 sub-block, with sumAbsLevel % 2 determining the sign of the first significant coefficient).

Consequence: the validity question for claims 1/4/7 is almost certainly a date question — at which JCT-VC meeting did per-coefficient-group parity sign hiding enter the working draft? If it was in a pre-2012-01-20 draft (JCTVC-G1103_d24, 7th meeting, Nov 2011 is the candidate), that draft anticipates, and Clare/Nguyen/§ 103 never need be reached. If it entered only later (8th meeting, Feb 2012, or 9th, Oct 2012), then the art is § 102(a) Clare + § 102(a) Nguyen/WD5 under § 103, which is still a strong obviousness case given both were before the 7th meeting and were cited on the face.

This is precisely the axis on which Unified's 2026-09-18 ex parte reexamination (noted in the earlier Litigation section) is likely to operate, and it is why the ">= tsig" vs. "exceeds a threshold" wording is worth pressing on: the granted claim says exceeds, while the applicant's own reproduced syntax is >= — a genuine claim-construction lever.

5. Ranked "most dangerous" cited references

1. Clare et al., "Sign Data Hiding," JCTVC-G271 (Nov 2011) — discloses parity-based sign hiding and the odd→negative mapping. Admitted on the face of the patent (cited in the background). § 102(a) art. Anticipates nothing alone on claims 1/4/7; directly implicates claims 2/5/8; the linchpin of any § 103 combination.
2. Any pre-2012-01-20 HEVC working draft carrying per-CG SDH (candidate: JCTVC-G1103_d24, Nov 2011) — the only plausible single-reference § 102 reference to claims 1/4/7. Verify the exact text.
3. Nguyen et al., JCTVC-G644 (Nov 2011) — supplies the "plurality of coefficient groups" element.
4. US 2013/0051459 A1 (Kirchhoffer et al., Technicolor) — significance-map / coefficient-block coding; conditional § 102(e) art (Technicolor 2011 provisionals).
5. US 2011/0268183 A1 (Sole et al., Qualcomm, 2011-11-03) — § 102(a) transform-coefficient coding.
6. US 2011/0310974 (Shand, 2011-12-22) — § 102(a); sign/coefficient signalling.
7. WD5/WD4 (Bross et al.), Bossen F900, Marpe CABAC — environment / secondary references.

Everything else on the list is field-of-the-art background (pre-2011 lossy-coding, watermarks, steganography) or procedural.

6. Bottom line (record-ready)

• 26 U.S. patent documents and 4 foreign documents are cited on the face of 11,786,596, plus a large "Other References" block. The U.S./foreign documents are general image/video compression and entropy-coding art; on the sources retrieved, none of them discloses the combination of (i) a per-coefficient-group threshold test on the span between the first and last non-zero coefficient and (ii) parity-of-summed-absolute-values sign inference, so none is shown to anticipate claims 1, 4, or 7.
• The substantive prior art is the NPL, not the patents: Clare JCTVC-G271 (parity sign hiding, Nov 2011) and Nguyen JCTVC-G644 (multi-level significance maps / 4×4 coefficient groups, Nov 2011), both § 102(a) art and both publicly cited. Together they are a strong § 103 case; a pre-2012-01-20 working draft containing the per-CG signHidden + sumAbsLevel % 2 construct would be a single-reference § 102 case.
• Claim 2/5/8 (negative sign when the sum is odd) is the claim most exposed to a direct § 102 attack on Clare alone.
• Five cited documents (2012/0163448, 2013/0051459, 2013/0114730, 2014/0301462, and the TU Berlin 2012 thesis) post-date the 2012-01-20 priority date and are available, if at all, only as § 102(e) art conditioned on their effective filing dates — verify each before relying on it.

7. Verification gaps I am flagging rather than papering over

1. The interior text of the cited U.S. patents/publications was not verified item-by-item. I hold citable bibliographic data (number, date, inventor) directly from the '596 front page; the short "brief description" entries are field-level characterisations, and the nine marked [content unverified] should be confirmed before filing anything that depends on them.
2. No primary-source confirmation of when per-CG parity sign hiding entered the HEVC working draft. § 4 is the highest-value follow-up.
3. I did not run an independent prior-art search beyond the patent's own citation list, as the task directed. A full § 102/§ 103 search would also reach Hyundai/Qualcomm's SDH work (US 2013/0272424 and its 61/671,166-family provisionals), which is not on this cite list but sits squarely on the same disclosure.
4. Source basis: the reference list was taken from the USPTO front-page reproduction for 11,786,596 as carried at patents.justia.com/patent/11786596 and freepatentsonline.com/11786596.html, corroborated against the patents.google.com/patent/US11786596/en record. I did not retrieve the PatentCenter face record itself; if a reel-precision cite is required, pull the PDF front page directly.