US 6594654

I'll search for litigation and docket information on this specific patent number.

Let me check for litigation involving the assignee and retrieve the full claim set.

Let me try to retrieve the full claim set to identify all independent claims.

US Patent 6,594,654 — Analyst Summary

Caveat up front: My searches of publicly indexed USPTO/Google Patents data and 2026 Federal Circuit materials returned no litigation, IPR, or CAFC 2026 docket activity for patent number 6594654. I cannot confirm any 2026 appellate docket for this patent, and I found nothing to contradict that. Likewise, I could not retrieve the complete, authoritative claim set for this patent, so my claim-level analysis below covers the independent claim I could verify and flags what I could not.

Bibliographic Data (per Google Patents, fetched 2026-09-24)

Assignment chain: Filed by VEDASOURCE LLC (2000-03-03); assignment of inventors' interest recorded 2000-07-06; second assignment recorded 2003-12-08.

One discrepancy to flag: Google Patents lists the priority date as 2000-03-03, while the Unified Patents portal lists US-6594654-B1 with a priority date of 2000-03-02. I am reporting both literally rather than reconciling them. The application number 09/518,862 and the 2000-03-03 filing date are consistent across the primary sources.

Related commercial context (not claim data): VedaSource LLC owned the marks KARNAK, GOFER, and VEDA. A contemporaneous trade newsletter ("Web Farming Newsletter," Jan. 2000) describes "Karnak" as a fee-based VedaSource research service using the library-bookshelf analogy with weekly e-mail updates — consistent with the patent's described commercial embodiment at karnak.com.

Abstract (verbatim)

"Systems and methods for accumulating and displaying information items obtained via a computer network. A user having access to the Internet is provided, via a web site, with a plurality of selectable expert topics. Each expert topic comprises one or more network computer accessible sources of information. The user enters a user search request, selects one of the expert topics, and enters update schedule information. The user search request, a selected expert topic, and update schedule information is stored at a server site. In accordance with the update schedule information, the user search request is provided to the information sources in the selected expert topic. Raw search results from the information sources are received, stored, and processed to eliminate dead links and duplicate items. The processed raw search results are stored as search results comprising a plurality of stored search items. A predetermined subset of the stored search items is selected and communicated to the user."

Plain-Language Overview of the Independent Claim(s)

Claim 1 — Method (the only independent claim I could verify). It is a method for accumulating and displaying information items obtained over a computer network, and it requires all of the following steps tied together:

1. Provide "expert topics" — multiple selectable subject groupings, each containing one or more network-accessible information sources; the items are located at addresses identifiable by resource-locator (URL-type) links.
2. Receive three things from the user: a search request, a selection of one of the expert topics, and update schedule information.
3. Store all three at the server.
4. In accordance with the schedule, send the same search request out to the sources inside the selected expert topic.
5. Receive and store the raw search results.
6. Process the raw results to eliminate dead resource-locator links and duplicate items (the specification's "Error 404"/link-rot and duplication problem).
7. Store the cleaned results as a persistent set of stored search items.
8. Select a predetermined subset of those stored items for delivery.
9. Communicate that subset to the user, with the subset including user-selectable resource-locator links.

The conceptual core: it is not a one-shot search engine. It is a persistent, scheduled, revisiting research service that curates a retained result set per user and hands back only a manageable "chunk," with dead links and duplicates stripped before display.

Dependent claims I could verify:

• Claim 2 adds a "hide" function — the user can command that selected result items be hidden; hidden items are not displayed, but they remain stored as stored search results (so they don't reappear as "new" on the next update).
• Claim 3 adds automatic updating of the stored results via a new search run on the stored schedule.

Uncertainty on the full claim set: I could not retrieve an authoritative, complete listing of all claims for US 6,594,654. I therefore cannot confirm the total number of claims, nor whether additional independent claims exist (e.g., a system/apparatus claim, a computer-readable-medium claim, or claims to the source-ranking / new-source-discovery / interface-probing features described in the specification, such as the "interface agent," "new source finder," and "source optimization" processes). Any statement that claim 1 is the sole independent claim would be an inference, not a verified fact. I recommend confirming via the USPTO Patent Center / PatentCenter full-text images or the printed patent's claims column before relying on claim scope.

Notable Disclosed Subject Matter (specification, not confirmed as claimed)

The specification describes several features that a reader should not assume are in the claims absent verification: the "knowledge engine" model; a mistake-detecting agent that auto-generates query-improvement rules; metatag-avoiding relevance re-evaluation by re-searching each retrieved page with the user's own string; source quality statistics and source ranking used to order displays; candidate new-source discovery from aggregate result statistics; and an interface agent that probes a source's search syntax to auto-derive interface parameters. These may be claimed in dependent claims I could not retrieve.

Litigation / Docket Check — Negative Result

I searched for CAFC 2026 dockets and district-court litigation naming this patent and found no matches. Searches returned only: (a) the patent itself, (b) third-party patents that cite US 6,594,654 as prior art (e.g., Google's news-ranking family, US 9,477,714 and related), and (c) VedaSource trademark records. Notably, the Unified Patents portal page for US-6594654-B1 shows no litigation or PTAB linkage in the retrieved snippet. Conclusion: I have no authoritative evidence of any 2026 CAFC or district-court proceeding involving US 6,594,654. If you need certainty, the definitive check is the USPTO Patent Center "Litigation" tab / assignment records and the Federal Circuit's own docket, which I could not access directly here.

I'll research litigation involving US patent 6594654 across patent litigation databases.

Let me search the specific litigation databases directly.

Litigation Search Report — U.S. Patent No. 6,594,654

Bottom line

I found no evidence of any litigation (district court, ITC, or Federal Circuit appeal) involving U.S. Patent No. 6,594,654. I searched and cannot confirm any case, past or pending, in which this patent was asserted.

I want to be explicit about the confidence level here: this is a negative finding based on the searches I was able to run, not a certified "no litigation ever existed" conclusion. Patent assertion data is imperfect, and I could not complete exhaustive queries against every database before hitting my tool-call limit.

What the patent is (for identification purposes)

• Patent: U.S. 6,594,654 B1 — "Systems and methods for continuously accumulating research information via a computer network"
• Application: 09/518,862; Filed: 2000-03-03; Granted: 2003-07-15
• Inventors: Aly A. Salam; Derek W. Ziglar
• Original/current assignee: VEDASOURCE, LLC
• Legal status per Google Patents: Expired – Fee Related; anticipated expiration 2020-03-03

An expired, fee-abandoned patent owned by a small entity is consistent with a low-to-zero litigation history — but the absence of an obvious plaintiff does not by itself prove none existed.

Sources consulted and results

Important caveat — do not conflate with similarly numbered patents

My searches repeatedly surfaced cases involving other patents whose shorthand designations can be confused with this one. None of these involve U.S. 6,594,654:

• PVT v. Apple (N.D. Cal., No. 4:13-cv-01710) — the "'659 patent" referenced in that Siri/SRI litigation is a different patent (the SRI speech-recognition family), not 6,594,654.
• Pantech v. OnePlus — references a "'654 Patent," which is U.S. 9,063,654, a telecommunications patent — not 6,594,654.
• Various other "654" / "659" shorthands appeared across unrelated telecom, pharma, and mechanical cases.

This is exactly the auto-correction trap the assignment warns against. I did not attribute any of these to your patent.

Recommended next steps to firm up the negative finding

Because I could not fully exhaust the databases, if you need a defensible clearance/opinion-grade answer, verify directly with:

1. PACER / CourtListener RECAP — full-text docket search for "6,594,654" and "6594654."
2. Lex Machina / Docket Navigator / Unified Patents litigation portal — patent-number-keyed case lists (these are the authoritative assertion databases).
3. USPTO Patent Center / Assignment records — confirm chain of title to VEDASOURCE, LLC and check for any recorded licenses or judgments.
4. Federal Circuit — search reported and unreported decisions citing the patent number.

Summary

• Known litigation: none identified.
• Patent status: Expired (fee related), anticipated expiration 2020-03-03.
• Confidence: Moderate — no hits across the sources I reached, but not an exhaustive negative clearance.

If you'd like, I can draft a short verification checklist or attempt narrower queries against a specific database if you can lift the tool limit.

I'll start by verifying the canonical list against web sources before drafting anything.

Let me run targeted searches to confirm whether any AIA trial exists for this specific patent number.

No hits yet on this patent. Let me check the defensive-aggregator and litigation angles explicitly.

Proceedings overview

Zero AIA trial proceedings are on file for US 6,594,654 ("Systems and methods for continuously accumulating research information via a computer network," Salam & Ziglar, filed 2000-03-03, granted 2003-07-15). Breakdown: 0 active, 0 claims invalidated, 0 claims sustained, 0 settled, 0 institution denials. This is not a "hardened patent" signal and it is not a "claims canceled" signal — it is an untested patent whose term already ran out: Google Patents records the status as "Expired – Fee Related," with an anticipated expiration of 2020-03-03. A defendant's best leverage here is the maintenance-fee lapse and the pre-AIA/PGR-unavailable posture, not any Board ruling.

Proceedings

No AIA trial proceedings to itemize

The USPTO ODP block enclosed with this task is the canonical list and shows no AIA trial proceedings. My independent web searches (PTAB public petition/decision documents, aggregate portals, Federal Circuit and CourtListener material, and litigation reporting) surfaced no IPR, PGR, or CBM naming US 6,594,654 as the challenged patent. I therefore have no proceeding number, petitioner, panel, institution decision, Final Written Decision, settlement, or appeal to report for this patent. I will not manufacture one.

Number-similarity false positives — do not let opposing counsel or a vendor's docket tool conflate these:

Note also that US 6,594,654's own U.S. claim text (as published) presents claim 1 as a method for accumulating and displaying information items via a computer network (expert topics → store search request + expert-topic selection + update schedule → query the topic's sources on schedule → eliminate dead links and duplicates → store → select a predetermined subset → communicate selectable resource-locator links), with claim 2 (user "hide" commands, items retained in storage) and claim 3 (scheduled updating) dependent on it. No PTAB panel has ever construed any of these terms.

Strategic summary

Claim status: every claim is UNTESTED — none canceled, none sustained. There is no IPR-narrowed claim set to work with. Because no petition was ever instituted, the inter partes estoppel machinery of 35 U.S.C. § 315(e)(2) was never triggered — no petitioner and no privy is estopped from anything. Conversely, the clock that normally gives a defendant leverage has run: the patent's term ended 2020-03-03 on a fee lapse, so the only damages exposure that could remain is back damages reaching back six years under 35 U.S.C. § 286, pinned to claims whose validity has never been administratively tested.

Trial-vehicle landscape is unusually narrow. Because the effective filing date is 2000-03-03 with no earlier priority claim of record (the Google Patents family note lists no family members), the patent is pre-AIA:

• PGR is unavailable — post-grant review only reaches patents subject to first-inventor-to-file, i.e., effective filing on or after 2013-03-16.
• CBM is unavailable — the transitional CBM window closed to new petitions on 2020-09-16, and in any event a web-search/research method would have faced the "technological invention" carve-out and the threshold requirement of a prior infringement suit or charge.
• IPR is the only AIA vehicle, and it is limited to pre-AIA § 102(a)/(b)/(e) and § 103 grounds on patents and printed publications. § 112 is not available in an IPR. IPR of an expired patent is legally permissible where past damages are at stake, so an IPR is not categorically foreclosed by the lapse — it is merely unattractive unless real money is on the table.

Pattern signals: none exist. There is no repeat petitioner, no patent-owner PTAB appeal practice, no Federal Circuit mandate, and no defensive aggregator in the chain that I could verify. The assignee of record is VedaSource, LLC (Smyrna, Georgia), the entity behind the "Karnak" research service (trademark serial 75665088, cancelled under § 8 as of 2008-04-25). The absence of even one IPR against a patent that has been on the books since 2003 is itself a signal: this patent appears never to have been asserted in a campaign large enough to attract a validity challenge.

Recommended next steps

1. If you received a demand letter citing US 6,594,654, verify current legal status before anything else. The patent lapsed 2020-03-03 for failure to pay a maintenance fee. Confirm on USPTO Patent Center whether any petition to revive under 37 C.F.R. § 1.378 has been filed or granted. I am not opining on revival deadlines — treat that as a diligence item, not a settled point. If it has not been revived, the patent is not enforceable for prospective conduct and the demand letter's injunction-style rhetoric is hollow.
2. Do not accept an assertion of claims 1–3 at face value as "settled law." No FWD exists, so there is no claim construction, no institution-stage claim-scope finding, and no estoppel to lean on. If past damages are genuinely at issue, an IPR on pre-AIA § 102(b)/§ 103 art (the 2000-03-03 priority date gives you a full year of prior art grace — art dated before 1999-03-03 is § 102(b) for printed publications and patents) is the cleanest lever, and IPRs can still be filed against expired patents where damages are in play.
3. Run your own confirmatory check rather than relying on this memo alone, since my conclusion is a negative finding from public search: PTAB E2E / the Board's public information system (https://ptacts.uspto.gov/ptacts/public-informations/), the USPTO Open Data Portal API for this application number (09/518,862), CourtListener (https://www.courtlistener.com/) and the Federal Circuit docket for any appeal naming "6,594,654," and district court dockets for VedaSource as plaintiff. If a recently filed proceeding exists that the ODP ingest has not yet picked up, that search — not this memo — will be what finds it.
4. Trap for the unwary: if any outside docket vendor or the other side serves a PTAB exhibit or decision referencing "the '654 patent," check the digits. The live '654 in PTAB practice is US 6,462,654 (IPR2016-00161), an entirely different patent and owner. Citing that decision as to US 6,594,654 would be a misrepresentation to the court or the Board.

I'll research the assignment chain for US 6,594,654. Let me run several targeted searches.

Let me dig into the inventors' affiliations and any recorded assignment data.

Let me look for the specific recorded assignment documents and the Unified Patents portal page.

Assignment & Ownership Analysis — US 6,594,654 B1

Data-limitation notice, up front. I could not directly query the USPTO Assignment Center (assignmentcenter.uspto.gov / assignment.uspto.gov) from this environment within the available steps, and neither the Google Patents legal-events list nor the indexed record surfaced reel/frame numbers or the recording correspondent for this patent. The two assignment events below are documented; the reel/frame and correspondent fields are not retrieved and are expressly marked as such rather than estimated. Per the operating rules, I will not fabricate reel/frame numbers or attorney names. Treat every "unretrieved" field as an open item to confirm in the Assignment Center.

Inventors

Patterns worth noting (flagged, not asserted):

• Both named inventors are also the only assignors on both recorded assignments, conveying to the same assignee — i.e., no inventor split off to a different party.
• The affiliated corporation VedaSource Acquisition Corp. was incorporated on 2000-02-14 — 18 days before the 2000-03-03 filing date. An "Acquisition Corp." created immediately before filing is a corporate-reorganization/roll-up signal surrounding the filing, though the assignment records name VedaSource, LLC (not the Acquisition Corp.) as assignee, so I cannot show the Acquisition Corp. ever held title.
• I found no evidence of inventors departing the assignee within 12 months of filing, and no data either way on their later employment. This is "unclear," not "clean."
• Contradiction to carry forward: the previously generated summary flagged that Google Patents lists priority 2000-03-03 while Unified Patents lists 2000-03-02. I am carrying both literally; the assignment analysis is unaffected either way.

Original assignee

VedaSource, LLC — a Georgia limited liability company, address of record Smyrna, GA 30080 (per the KARNAK trademark registration records).

• Primary line of business: a fee-based Internet research service — the "knowledge engine" commercialized as KARNAK. The same entity also filed the marks KARNAK, GOFER, and VEDA (per the earlier summary and trademark records).
• Did they ship a product embodying the claims? — Yes (evidence-based). The January 2000 Web Farming Newsletter describes Karnak as "a new fee-based service by VedaSource in Atlanta," using the library-bookshelf analogy, sending a weekly email summarizing any new, changed, or deleted material, at $10–$50/month by number of research topics. That description tracks the asserted claim-1 architecture (expert topics, stored search items, scheduled updates, notification of new/changed/unavailable items) and the "bookshelf" UI. This is a real operating product, not a licensing shingle.
• Trademark corroboration of operating status: KARNAK Serial No. 75665088, filed 1999-03-22, first use in commerce 1999-03-09, registered 2001-07-24, cancelled §8 on 2008-04-25 (non-renewal).
• Current status: Defunct / ceased operating. Direct Georgia registration status for "VedaSource, LLC" was not retrieved, but the surrounding entities are dead: VedaSource Acquisition Corp. dissolved 2005-07-09 (last annual registration 2001), and the KARNAK mark was cancelled in 2008. No evidence of an operating successor surfaced.
• Registered-agent detail on the affiliated corp (literal): VedaSource Acquisition Corp. registered agent Paul A. Quiros, address 191 Peachtree St., Ste. 4600, Atlanta, GA 30303 — a multi-tenant office-tower suite address (commonly a law-firm address). I cannot name the firm from the data retrieved, and I am not treating this as a correspondent finding.

Assignment timeline

Two assignment events appear in the Google Patents legal-events record. Both are "Assignment of assignors' interest (see document for details)"; both name the same assignors (Salam and Ziglar) and the same assignee (VedaSource, LLC).

• Executed — not retrieved / recorded 2000-07-06 — Reel unretrieved/unretrieved

  • Conveyance: Assignment (of assignors' interest)
  • Assignor: SALAM, ALY A.; ZIGLAR, DEREK W.
  • Assignee: VEDASOURCE, LLC
  • Correspondent: not retrieved (Assignment Center not queried). No recurrence finding possible — single retrieved appearance only.
  • Context: Internal/original acquisition — founders transfer the application to their own operating company ~4 months after the 2000-03-03 filing (standard employee/founder-to-company assignment).

• Executed — not retrieved / recorded 2003-12-08 — Reel unretrieved/unretrieved

  • Conveyance: Assignment (of assignors' interest)
  • Assignor: SALAM, ALY A.; ZIGLAR, DEREK W.
  • Assignee: VEDASOURCE, LLC
  • Correspondent: not retrieved. No recurrence finding possible.
  • Context: Confirmatory / duplicate recording only — no change of ownership. This recording lands ~5 months after issuance (2003-07-15) and, on the face of the record, conveys the identical interest from the identical inventors to the identical assignee as the 2000-07-06 entry. Two recordings of the same inventors→VedaSource transfer is the definition of a confirmatory filing, not a new transaction.

Key structural finding: the ownership chain never leaves the original operating assignee, VedaSource, LLC. There is no post-issuance transfer to any third party, LLC, or aggregator anywhere in the retrieved record.

Timeline diagram

timeline
    title Ownership of US 6594654
    1999 : VedaSource files KARNAK trademark
    2000 : Patent application filed
         : VedaSource Acquisition Corp incorporated
         : Founders assign to VedaSource LLC
    2001 : KARNAK trademark registered
    2003 : Patent issued
         : Confirmatory assignment recorded
    2005 : VedaSource Acquisition Corp dissolved
    2008 : KARNAK trademark cancelled
    2020 : Patent term expired

NPE / troll-pattern signals

Aggravating structural note (not a scoring factor above): the patent shows legal status "Expired – Fee Related," anticipated expiration 2020-03-03 — i.e., the full statutory 20-year term from the 2000-03-03 filing ran out with no owner ever asserting the patent in court or before the PTAB.

Verdict

Insufficient data — and, on the retrieved record, no NPE pattern is present.

The only recorded transactions are the founders' inventors→VedaSource, LLC assignments recorded 2000-07-06 and 2003-12-08, both conveying to the same original operating assignee that shipped an embodying fee-based product (Karnak). There is no post-issuance transfer to any third party, licensing LLC, or aggregator, the patent reached its anticipated term expiration on 2020-03-03, and no litigation or PTAB proceeding was found — so signals 1, 2, 4, 5, 6, 7, and 8 are each not present, while signal 3 is unresolvable because the reel/frame and correspondent fields were not retrievable here. Because the record shows only the original assignment(s), the governing category is Insufficient data, not any NPE tier.

Verify at: USPTO Assignment Center — https://assignmentcenter.uspto.gov/ and https://assignment.uspto.gov/patent/index.html (search by patent number 6594654). To close the open items, pull: (a) the reel/frame and correspondent of record for the 2000-07-06 and 2003-12-08 recordings, and (b) whether any unrecorded interest in VedaSource, LLC was later conveyed during the 2005 Acquisition Corp. dissolution — the Assignment Center will not show unrecorded transfers.

I'll research the prior-art citations for US 6,594,654 directly from patent databases.

Let me retrieve the cited-references tables directly from the patent's own pages.

Prior-Art Analysis — U.S. Patent No. 6,594,654 B1

Analyst note on sources and method. I could not reach USPTO Patent Public Search / PatentCenter directly this session (tool limit reached after the queries below). What I did retrieve are two mirrors of the USPTO front page: FreePatentsOnline's reproduction of the printed patent front page (FreePatentsOnline.com/6594654.html), which contains the face-of-patent "U.S. Patent References" list, and the Google Patents record (patents.google.com/patent/US6594654B1, supplied in the authoritative full text). I therefore treat the FPO list as the best available rendering of the examiner/applicant-cited references, and I flag where I could not independently confirm it. I did not retrieve the complete "Patent Citations" table from the Google Patents page itself, so I cannot 100% rule out that FPO's list is abridged.

Identifier discipline. All numbers below are reported literally. Note that a search for "6594654" surfaced an unrelated Chinese utility-model certificate number 6594654 (ZL 2017 2 0213078.0, 十堰深思科技有限公司) — excluded, not your patent. Likewise the "654 patent" in Pantech v. OnePlus and the "'659 patent" in PVT v. Apple remain excluded, as flagged in the litigation section.

1. Correction to the previously generated sections (flagging a contradiction)

The earlier Patent summary and Litigation summary sections correctly cautioned that only claim 1 could be verified and that the number of claims and existence of additional independent claims were unknown. That gap is now closed, and the earlier caveat is superseded:

• US 6,594,654 B1 has 16 claims.
• Three independent claims: 1, 14, and 15. Claims 2–13 depend (directly or through claim 3) on claim 1; claim 16 depends on claim 15. There is no system/apparatus claim and no computer-readable-medium claim — the patent is method-only.
• Claim 1 (method: expert topics → user request + topic selection + update schedule → store → schedule-driven submission → receive/store raw results → eliminate dead links and duplicates → store processed set → select subset → communicate with user-selectable links); claim 14 (method: arrange sources into expert topics; search per prestored source ranking; display high-quality sources first; re-rank sources; identify potential new sources off-line); claim 15 (method: sources in expert-topic categories; apply search on user input; receive/store first result set; process to eliminate dead links → second set; present third-set subset; receive user "hide" input); claim 16 adds a quality threshold to claim 15.

Secondary discrepancy to flag: FreePatentsOnline's front page lists the assignee as "SALAM ALY A.; ZIGLAR DEREK W." (the inventors), whereas Google Patents lists VEDASOURCE LLC as original and current assignee. Both are reported literally; the assignment records (2000-07-06 and 2003-12-08) support the VedaSource ownership, so the FPO assignee field appears to be a pre-assignment data artifact.

2. References cited on the face of U.S. 6,594,654 (the "US Patent References" list)

Per FreePatentsOnline's reproduction of the patent front page (https://FreePatentsOnline.com/6594654.html):

No foreign patent documents or non-patent literature appear on the face of the patent in the retrieved rendering.

Two anomalies you should verify against the printed patent:

1. US 408,296 (Doane, 1889; class 96/377). Class 96 is photography; an 1889 photographic-plate reference has no discernible bearing on networked information retrieval. This is very likely a miscitation or a mirror-site data artifact (possibly a stray citation carried over from another record). I report it literally rather than "correcting" it, per the identifier rule.
2. US 6,483,702 (Adar, "thermal spreader plate"). A mechanical/thermal-hardware patent on a search-engine method patent is likewise anomalous. Its issue date (2002-11-19) post-dates the 2000-03-03 filing date, so to have any §102(e)/§102(a) effect its own filing date must precede 2000-03-03 — which I could not verify. Treat it as a probable erroneous or formal citation unless the printed front page shows otherwise.

3. Per-reference analysis and §102 assessment

Legal standard applied: anticipation under 35 U.S.C. § 102 requires that a single reference disclose each and every element of the claim, arranged as in the claim. For pre-AIA § 102(e) references (patents/applications), the reference's effective filing date must pre-date 2000-03-03 (the §102(b) critical date is 1999-03-03).

Reference 1 — US 6,483,702 B1, Adar et al.

• Full citation: U.S. Patent 6,483,702 B1, "Apparatus and methods for attaching thermal spreader plate to an electronic card," issued 2002-11-19.
• Brief description: Mechanical/thermal-management subject matter (attaching a heat-spreader plate to an electronic card). No information-retrieval, search, scheduling, or network-research disclosure.
• §102 assessment: Anticipates no claim. It cannot disclose any element of claims 1, 14, or 15 (no expert topics, no search, no scheduling, no link processing). Also likely fails the date test as an anticipatory reference. §102: none.

Reference 2 — US 6,282,533 B1, Ramaswamy et al.

• Full citation: U.S. Patent 6,282,533 B1, "System and method for performing I/O-efficient join processing," issued 2001-08-28.
• Brief description: Database engine technique for performing memory/IO-efficient join operations — i.e., merging/combining data sets. Its only conceivable bearing is on the de-duplication element (merging result sets without duplicate records).
• §102 assessment: At most, it touches the "eliminate … duplicate items" element of claim 1 and the corresponding processing step in claim 15 — and even then, join processing is a generic database operation, not a disclosure of de-duplicating search result items from multiple network sources. No anticipation of claim 1 or claim 15 (all remaining elements absent). §102: none; possible §103 background on de-duplication.

Reference 3 — US 5,941,944, Messerly — closest reference on link-integrity

• Full citation: U.S. Patent 5,941,944, "Method for providing a substitute for a requested inaccessible object by identifying substantially similar objects using weights corresponding to object features," issued 1999-08-24 (class 709/203). Pre-dates the critical date (1999-03-03), so it qualifies as §102(b)/§102(a) art.
• Brief description: Handles inaccessible/broken links by substituting a substantially similar object, selected using feature weights — a "link rot" mitigation technique, but its solution is substitution, not elimination.
• §102 assessment: This is the reference most likely to have been cited against the "processing the raw search results to eliminate dead resource locator links" element of claim 1 and the parallel element in claim 15. However, Messerly does not disclose "eliminating" dead links — it supplies a substitute — and it discloses none of the expert-topic, update-schedule, subset-selection, or hide-command elements. It therefore does not anticipate claim 1, claim 15, or any dependent claim; it is properly a §103 reference if combined with a scheduling/multi-source teaching. §102: none.

Reference 4 — US 5,848,410, Walls et al. — closest reference on indexing

• Full citation: U.S. Patent 5,848,410, "System and method for selective and continuous index generation," issued 1998-12-08 (707/4). Qualifies as prior art (pre-1999-03-03).
• Brief description: Continuous/selective generation of an index over changing content — conceptually adjacent to the patent's "dynamic indexing" narrative and to storing a maintained result/index set.
• §102 assessment: Could be cited against the storing/maintenance aspects (claim 1 steps "receiving and storing raw search results" / "storing the processed raw search results"; arguably supporting claim 3's updating concept). But an index-generation system discloses no user search request + expert-topic selection + update schedule + subset delivery + hide function. Does not anticipate claim 1, 3, 14, or 15. §102: none; §103 candidate.

Reference 5 — US 5,708,825, Sotomayor — closest reference on result presentation

• Full citation: U.S. Patent 5,708,825, "Automatic summary page creation and hyperlink generation," issued 1998-01-13 (707/501.1). Prior art.
• Brief description: Automatically generates summary pages and hyperlinks — i.e., automated construction of a linked listing/summary presentation.
• §102 assessment: Bears on the "communicating … including resource locator links selectable by the user" element of claim 1, the collection/annotation ideas behind claim 4 (contextual display) and claim 11 (results grouped by site). But it discloses no scheduled multi-source search, no dead-link/duplicate elimination, no hide function, and no expert topics. Does not anticipate claim 1, 4, or 11. §102: none; §103 candidate on the presentation elements.

Reference 6 — US 5,566,330, Sheffield — closest reference on source interfacing

• Full citation: U.S. Patent 5,566,330, "Method for forming a reusable and modifiable database interface object," issued 1996-10-15 (707/4). Prior art.
• Brief description: Creating reusable/modifiable database interface objects — a general mechanism for adapting a program to heterogeneous database interfaces.
• §102 assessment: Relevant conceptually to the specification's later-claimed/described interface-determination subject matter (dependent claim 12 "determining the interface parameters associated with the potential new information source," and the "interface agent" of Fig. 9). It discloses no automated probing of a network search source to derive interface parameters, and no other claim element. Does not anticipate claim 12 or any other claim. §102: none; §103 candidate on claim 12 only if combined with network-source probing art.

Reference 7 — US 408,296, Doane

• Full citation: U.S. Patent 408,296 (title listed as "N/A"), issued 1889-08-06, class 96/377, inventor Doane.
• Brief description: Class 96 indicates photography/chemistry. No discernible relevance to networked search or database research.
• §102 assessment: Anticipates no claim. Almost certainly a miscitation or mirror-site artifact; verify against the printed front page. §102: none.

4. Bottom line on the cited prior art

Characterization: The seven face-of-patent references are background/general-art citations, not anticipatory art. If they were used in prosecution at all, they most plausibly supported §103 positions (Messerly on link handling; Walls on continuous indexing; Sotomayor on summary/link presentation; Sheffield on reusable interfaces; Ramaswamy on join/de-dup). No reference on the face of US 6,594,654 anticipates any claim of the patent.

Important scoping caveat: the strongest §102 candidates for this patent's subject matter are almost certainly non-patent prior art that does not appear on the face of the patent — e.g., contemporaneous commercial meta-search services (DOGPILE.COM, METACRAWLER.COM — referenced by name in the patent's own Background), "push"/personalized-news services of the 1996–1999 era, and academic metasearch work (the Hal-hosted thesis hal.science/tel-04697087v1 reproduces result-merging and "current news metasearcher" references, e.g., Rasolofo et al., Result merging strategies for a current news metasearcher, but those are 2003 and therefore post-date this patent). Because I reached my tool limit, I did not retrieve the prose "Other Publications" / non-patent-literature list that may also appear on the front page — if one exists there, it should be checked before any invalidity conclusion is drawn.

5. Forward citations (NOT prior art — listed only to prevent confusion)

For completeness: US 6,594,654 is cited as prior art in many later patents/applications, e.g. US 8,489,607 B2, US 2002/0129144 A1, US 8,676,868, US 9,165,039, US 7,627,621, and US 11,443,611 (per patents.justia.com/patent/11443611 and patents.justia.com/patent/8676868). These are "cited by" references and carry no anticipatory weight against US 6,594,654.

6. Recommended verification steps

1. Confirm the front-page reference list from the USPTO PatentCenter/Patent Public Search full-text image of US 6,594,654 (column 1) — this is the authoritative "References Cited," and I could not reach it directly. Specifically confirm (a) whether the list is exactly seven U.S. references, (b) whether US 408,296 (Doane) and US 6,483,702 (Adar) truly appear, and (c) whether any foreign references or "Other Publications" are listed.
2. Pull filing dates for US 6,483,702 and US 6,282,533 to test §102(e)/§102(a) eligibility against the 2000-03-03 filing date.
3. Retrieve the file wrapper (prosecution history) — the examiner's reasons for allowance and any §103 rejections would reveal which of these references actually mattered, and whether any reference now relied on was considered and overcome.
4. If invalidity is at issue, search pre-1999 non-patent literature (metasearch, "push" services, personal research organizers) — the highest-value §102 candidates are not on this patent's face.

Confidence: High that the 16-claim structure with independent claims 1, 14, and 15 is correct (two independent mirrors agree on the full claim text); moderate-to-high that the seven-reference list is the complete face-of-patent U.S. reference list, given I could not confirm it against the printed patent image; high that none of those references anticipates any claim.

I'll first note a discrepancy in the task setup, then gather the prior art I need to ground the analysis.

⚠️ Flagging a conflict with the task instruction: The task says "Use the results from the Prior Art section of this page." No Prior Art section has been generated in this analysis. The two prior sections are a Patent Summary and a Litigation Summary. There is nothing to build on for prior art, so I cannot claim to be "using" a section that does not exist. Rather than fabricate a prior-art section, I'll derive the art from (a) the admissions in the patent's own Background, and (b) live searches for pre-2000 references. Let me pull real references.

MetaCrawler is confirmed pre-1995 art. Now let me pull the actual references cited on the patent's face and the other key pre-2000 references in the meta-search / recurring-search space.

§103 Obviousness Analysis — U.S. Patent No. 6,594,654 B1

0. Threshold notes and conflicts to flag

1. No "Prior Art" section exists on this page. The task instructed me to "use the results from the Prior Art section of this page." There is no such section — only the Patent Summary and Litigation Summary reproduced above. I am therefore building the prior-art set myself from (a) admissions in the patent's own Background, and (b) live searches. This is a deviation from the instruction and should be recorded as such.

2. Claim-set uncertainty carries forward. The earlier section could not verify the complete claim set and expressly warned against assuming Claim 1 is the sole independent claim. That limitation applies directly here: a §103 analysis is only as good as the claims it addresses. Everything below is keyed to the Claim 1–3 limitations paraphrased in the prior section, and is flagged where the paraphrase (not verbatim claim text) is doing the work.

3. Priority-date discrepancy persists. Google Patents: 2000-03-03. Unified Patents: 2000-03-02. Immaterial to the outcome, but it shifts the pre-AIA §102(b) critical date to 1999-03-02 or 1999-03-03. I use 1999-03-03 below and note it does not change any reference's status.

4. Tool budget exhausted. My last retrieval call returned a hard stop ("maximum number of steps"). I could not pull the face-of-patent "(56) References Cited" list from the US6594654 PDF. I therefore do not assert what the examiner cited; I assert only what the art itself discloses.

1. Governing framework

• Pre-AIA §103(a) applies. Application 09/518,862 was filed 2000-03-03, well before the AIA first-to-file change (16 March 2013). The test is Graham v. John Deere + KSR Int'l v. Teleflex, 550 U.S. 398 (2007): scope/content of prior art; differences; PHOSITA level; and secondary considerations.
• PHOSITA (my construction): a software engineer with ~2 years' experience building Web information-retrieval systems, familiar with HTTP, HTML forms, CGI/proxy architectures, IR ranking, and the then-current meta-search literature. This art was unusually well documented in the open literature (not secret), which matters for both the level of skill and the motivation analysis.
• Because KSR governs, the patentee cannot defeat obviousness by showing the references lack an express "teaching, suggestion, or motivation" to combine where the combination is a predictable use of prior-art elements according to known methods, a mere substitution of one known element for another, or a "finite number of identified, predictable solutions."

2. The prior-art set (all verified in this session)

URLs for the key references:

• A1 architecture paper: https://homes.cs.washington.edu/~etzioni/papers/ieee-metacrawler.pdf ; WWW4 paper: https://archives.iw3c2.org/www4/Papers/169/
• A2: https://www.cs.colostate.edu/TechReports/Reports/1996/tr96-111.pdf ; https://dlnext.acm.org/doi/epdf/10.1145/[256163](/patent/256163).[256164](/patent/256164)
• A3: https://SumoBrain.com/patents/us/Method-apparatus-searching-information-in/[5715443](/patent/5715443).html
• A4/A6 (as cited of record in a later patent): https://patentimages.storage.googleapis.com/36/ef/05/1c0b2d3589689e/[US9002934](/patent/US9002934).pdf
• A5 (as cited in the same IDS): same document

3. Claim 1 — element-by-element mapping

Claim 1's verified limitations (per the prior section) map as follows. Bold = the reference that most cleanly meets it.

Result: every limitation of Claim 1 is disclosed in the pre-1999 art, and limitations 6 and 8 are met by express disclosures, not inference.

4. Primary combination — the one I would assert

Combination A: A1 (MetaCrawler) + A2 (SavvySearch) + A3 (US 5,715,443), optionally + A5 (Liu/Pu/Tang) for the "continuously accumulating" framing.

4.1 What each brings

• A1 supplies the multi-source aggregation engine, duplicate elimination, availability/quality checking by fetching each hit, per-service query reformatting via the "Harness"/wrapper, and ranked collated output.
• A2 supplies the dispatch agent / interface agent / display agent triad, source selection learned from prior performance (the "meta-index" tracking each engine's success per query term), reformatting user queries into per-engine syntax, and the deliberate delivery of a limited result subset (<30).
• A3 supplies the missing piece the patent frames as its own contribution — persisting the user's query and a scheduling parameter at a server and automatically re-running the query at the scheduled time and reporting in summary/email form. A3's own background states the problem in terms nearly identical to the patent's: "services have been available to periodically perform searches … and notify the requester of the results."

4.2 Motivation to combine — this is unusually strong

1. Express cross-reference in the art itself. The MetaCrawler IEEE paper's Future Work section states verbatim: "One possibility is to use a learning-based approach similar to SavvySearch[3], where MetaCrawler learns which services to use based on prior experience." This is not an inferred motivation — it is the prior art telling the PHOSITA to make the combination. Under KSR this is dispositive for the source-ranking and source-selection features.
2. Reciprocal citation. SavvySearch's own papers name MetaCrawler as the other true meta-search engine and position the two as the known design space ("These shortcomings motivate true meta-search engines, such as SavvySearch … and MetaCrawler"). Combining the two known architectures — or selecting between them per query — is the textbook "finite number of identified, predictable solutions" scenario.
3. Same field, same problem, same solution space. All of A1–A3 target the identical problem the patent recites as unsolved: too many irrelevant results, stale indexes, dangling links.
4. A3's mechanism is field-agnostic and mechanically compatible. Scheduling a stored query and reporting results is a routine server-side timer/queue function (A3 FIG. 7b "search scheduling window"); A8 shows the same timer-control + reserved-search + auto-email architecture independently arrived at, confirming it was a known technique by 1998. Applying a known scheduling/notification technique to a known meta-search dispatch engine is precisely KSR's "use of a known technique to improve similar devices in the same way."
5. Market/practical pressure. The patent's own Background recites the demand ("users … are often inundated with too many results"; "search engines provide only one tool for actually conducting a research project"). Pressure supplying a recognized need weighs toward obviousness when the solution is a predictable re-arrangement of known components.
6. Obvious-to-try roadmap. A1's Future Work section is effectively a published project plan listing the very modifications the patent claims as inventive (learned service selection, alternative ranking, clustering), and A5 (continual queries) supplies the persistent/subscription model.

I assess Claim 1 as strongly obvious over Combination A. If forced to pick a single reference for a §102 challenge, A3 alone anticipates the scheduling/persistence/subset/report core but lacks the Web "expert topic + multi-source + dead-link/duplicate" elements; A1 alone supplies the Web aggregation core but lacks persistence and scheduling. Hence §102 is weaker than §103, but §103 is robust.

5. Alternative combinations worth pleading in the alternative

Combination B: A3 (scheduled search reports) as primary + A1/A2 as secondary. This is the cleaner framing if the examiner or a court reads "expert topic" narrowly as curated subject groups rather than any source list — A2's categorical approach is the bridge.

Combination C: A2 + A5. A2 gives learned source selection + per-source interface adaptation + limited result sets; A5 gives continual/subscription queries with event-driven delivery over Internet-scale sources. Together they cover "continuously accumulating" plus "predetermined subset + notification."

Combination D: A1 + A4 (Inquirus) + A3. A4 is the cleanest mapping for the specification's "metatag-avoiding relevance" feature: Inquirus downloads and analyzes each retrieved document to compute relevance rather than trusting source-supplied metadata — exactly A1's stated advantage ("An added benefit of retrieving the documents is the ability to apply a sophisticated query language"). Use D if the metatag/relevance limitation is actually claimed.

6. Dependent claims

Claim 3 (automatic update on the stored schedule) — strongly obvious. This is met expressly by A3: the stored request is re-executed "at the scheduled search time" / "as necessary" when new documents are indexed, generating a fresh report. Add A8 as a redundant secondary (timer-controlled reserved search + automatic email of update results). Claim 3 adds essentially nothing over Claim 1's combination. Highest confidence of the three.

Claim 2 (hide selected items; hidden items remain stored and are not displayed) — the weakest claim, and I flag it as such. The prior art clearly teaches filtering/pruning and excluding results (A1: "excluding documents containing other phrases"; A2: boolean NOT and resource-limit-based pruning), and A3/A5 teach retaining a stored result list across runs. What I did not find in this session is a reference expressly teaching the specific combination of (i) a user-invoked per-item hide toggle, (ii) non-deletion of the hidden item from the persistent store, and (iii) suppression of that item from subsequent update deliveries. I would not overstate this. My §103 position on Claim 2:

• First argument: the claim is an obvious soft-delete / flag-field implementation. Storing a record with a boolean "displayed" flag set to false, while keeping the row, is one of the most predictable database implementation choices available, and KSR permits attacking a claim where the only difference is an obvious and commonly used implementation of an already-disclosed function. The prior art already discloses both the filter function and the persistent store; the flag merely joins them.
• Second argument: the motivating problem is admitted in the patent's own text — results must not "reappear by being 're-found' when research is updated," and the patent itself analogizes hiding to not throwing away index cards. That is a stated design need that flows directly from combining a persistent store (A3, A5) with a filter (A1, A2); it is a predictable result, not an inventive step.
• Third argument (needs verification): user-maintained persistent suppress-lists were a well-established interaction pattern outside this field before 2000 — most obviously Usenet/newsreader "kill files", mail filters, and news-filtering agents (the filters referenced in the GroupLens/Usenet literature surfaced in my search as background references). I did not verify a specific pre-1999 kill-file reference in this session and am not citing one as a §103 reference. If Claim 2 matters to the outcome, this is the first gap to close.

Bottom line on Claim 2: obviousness is defensible but materially weaker than for Claims 1 and 3, and I would want one additional secondary reference (persistent user-maintained suppression list) before asserting it with confidence.

7. Specification features — status if they turn out to be claimed

The prior section warned these may sit in claims I could not retrieve. My preliminary read:

8. Secondary considerations (Graham factor 4)

I have no evidence in the record to support a non-obviousness finding:

• Commercial success: the earlier section notes VedaSource's KARNAK service — a fee-based small-entity offering, and the patent expired for failure to pay maintenance fees with an anticipated expiration of 2020-03-03. Neither fact establishes commercial success, and there is no demonstrated nexus between any commercial product and the specific claim limitations.
• Long-felt need: the need is real and the patent recites it — but the "need" is also recited by A3's own background, which undercuts any claim that the problem went unsolved.
• Unexpected results: none apparent. The claim's benefit (fewer, fresher, de-duplicated links) is precisely the benefit A1 and A2 advertise ("MetaCrawler is very successful at pruning irrelevant results"; SavvySearch's users "are better served with a small set of results").
• Teaching away: none found.

Given KSR, absent objective evidence with nexus, the secondary-considerations factor does not rescue Claims 1 or 3.

9. Opinion and confidence levels

Explicit uncertainties:

1. I could not retrieve the verbatim claim text or the complete claim set; the mapping above rests on the paraphrase in the prior section. A verbatim claim chart must be run before any of this is relied on.
2. My tool budget was exhausted before I could pull US6594654's face-of-patent citations or verify A8's Japanese publication date/status independently; A8 is 102(a) art at best and possibly not prior art at all.
3. The Task's instruction to "use the Prior Art section of this page" was followed to the extent possible — but note the conflict flagged in §0.1: that section does not exist, so I substituted self-generated art.

If you can lift the tool limit or supply the printed patent's claims column, the highest-value next steps are (a) a verbatim element-by-element chart against Combination A, and (b) a hunt for a pre-1999 persistent user-maintained suppression list to shore up Claim 2.