Have Claude read what humans don't.
Arbitrage the gap.

From the AI Archaeology series launch on 2026-05-01 (ep01-07 completed the same day) through Day 29 = ep100 on 2026-05-08, the series reached 100 episodes (31 excavation notes + 69 excavation memos) over 8 days end-to-end. The 100-episode plan portion (Day 1 starting 2026-05-06 through Day 29) was a 3-day, 29-session high-density operation; 'Day' here is a session number, not a calendar day (5/7 alone hosted Day 3 through Day 13, and 5/8 hosted Day 14 through Day 29). This synthesis note does not introduce individual episodes; it stocktakes what the 100 episodes excavated, organized along four axes. Axis 1 is 'DB reliability 6 forms', summarizing 57 cumulative DB corrections and 13 confirmations from Day 8 onward into six recurrent forms: (1) wrong-number swaps where the listed patent points to a completely different invention (cosmetics 5+ cases, food-health 1 case), (2) marketing-phrase misreadings where the patent number is real but the conventional 'Claim 1 subject' diverges from the verbatim claim (CS-009 P&G niacinamide, conventionally 'whitening' vs verbatim 'regulating mammalian skin pore size'), (3) information-wall via interactive search UI (CS-004 DPMA, CS-005 USPTO 1970s, CS-010 J-PlatPat as a three-form set), (4) absence (PH-004 Köhler-Milstein, CS-002 botulinum toxin, multiple SW absences), (5) eligibility wall (7 SW cases), and (6) information-wall via OCR failure (SW-007 Lapson, second inventor field garbled). Axis 2 is 'Eligibility Wall 4 forms', established Days 24-26 as a structural classification of why 1957-1990s US software inventions were not patented: (a) pre-judicial era with 4 sub-forms (FORTRAN/LISP/ALGOL/COBOL), (b) unsettled era (HyperCard), (c) government-contract forced disclosure (BBN IMP, Bell-LaPadula), (d) corporate-strategy voluntary disclosure (Smalltalk). Axis 3 is 'Cage Patents 9 forms', accumulated Days 19-28: a 'confine in order to use' design genealogy that spans 6 physical cage forms (electron / charge / static molecular / electrical / dynamic molecular / container / ion) and 3 logical cage forms (type / policy / capability). Axis 4 is the '12 sub-series excavation map' that grew from the original 4 sub-series (Patent / IR / Standard / Declassified) by adding Kitchen Health / Pharma / Cosmetic / Hardware-Energy / Internet-Crypto / Software-UI / AI-ML / Food-Health.

On December 23, 1983, Norman Hardy (architect of KeyKOS / GNOSIS) at Tymshare Inc in Cupertino, California filed 'Computer security system' in the United States as the assignor. It was granted as US4584639A on April 22, 1986. This patent fences in Claim 1 'the combination of domains (units of processing) / keys (units of authority) / kernel (the sole institution to resolve authority) / factories (mechanism to generate new domains) / non-sensory keys (authority that can extract information) in a capability-based data processing system,' historically important as a successful patenting example of a capability cage. The three-stage transfer history — 1984-02-06 assigned to Tymshare, 1985-08-01 to McDonnell Douglas Corporation, 1985-12-02 to Key Logic Inc — was traversed. This memo verifies the positioning of the patent as the origin of KeyKOS against six secondary sources: Wikipedia EN KeyKOS / Norman Hardy / Capability-based security articles, Mark S. Miller's Medium tribute 'Norm Hardy's Place in History,' Semantic Scholar's 'Security in KeyKOS' Rajunas/Hardy paper, Justia Patents Search Norman Hardy results, the Confused Deputy paper (Hardy 1988), and Google Patents US4584639A. As the Day 28 / Cage Patents axis SW Open closing memo, it completes the SW Cage three forms together with ep97 Yellin/Gosling type-system cage (confines via information consistency, patenting success) / ep98 Bell-LaPadula policy-model cage (confines the direction of information flow, patenting absence) / this memo Hardy capability cage (confines via physical distribution of authority, patenting success).

Day 28 / Cage Patents axis SW Open memo. In summer 1972, under contract from the United States Air Force Electronic Systems Division (USAF ESD, located at Hanscom Air Force Base), David Elliott Bell (then 27) and Leonard J. LaPadula at MITRE Corporation began the mathematical formalization of multi-level security for military use. In November 1973, it was released as MITRE Technical Report 2547 Volume I 'Secure Computer Systems: Mathematical Foundations,' with DTIC Accession Number AD-770768 and Air Force Report Number ESD-TR-73-278 for permanent archival. The Bell-LaPadula model (BLP) centers on three axioms: (1) Simple Security Property 'no read up' (a high-confidentiality object cannot be read by a low-clearance subject), (2) Star Property 'no write down' (a high-clearance subject cannot write to a low-confidentiality object), and (3) Discretionary Security Property (per-pair access-control matrix). This memo verifies, against five secondary sources — Wikipedia EN Bell-LaPadula model article / DTIC AD-770768 official PDF / Internet Archive file text / Bell's own 2005 ACSAC retrospective paper 'Looking Back at the Bell-La Padula Model' / Springer Encyclopedia of Cryptography and Security article — that the BLP model was published as a MITRE TR and a CACM 1976 paper without any reference to related patent numbers. Alongside Day 25 ep89 SW-003 BBN IMP's 'ARPA contract → 57-year continuous publication as Internet STD 39' form, this is the second instance of the 'eligibility wall (c) government-contract publication-mandate form.' On the Cage Patents axis SW, it is positioned as 'an information-theoretic policy model that was not patented (absence pattern),' placed at the center of ep97 Yellin/Gosling type-system cage success (US5740441A) and ep99 Hardy KeyKOS capability cage success (US4584639). The central memo of the SW Cage three forms (type cage success / policy cage absence / capability cage success).

On December 20, 1994, Frank Yellin and James A. Gosling at Sun Microsystems filed 'Bytecode program interpreter apparatus and method with pre-verification of data type restrictions and object initialization' in the United States. It was granted as US5740441A on April 14, 1998. This patent fenced the core of the Java Virtual Machine bytecode verifier, the origin of the 'sandbox by pre-execution type checking' design idea that runs continuously through modern JVM, Android Runtime, WebAssembly validators, and the .NET CLR. Against the six physical Cage forms consolidated by Day 27 (ep70 electron / ep71 charge / ep72 static molecular / ep94 electrical / ep95 dynamic molecular / ep96 container), this patent is positioned as the SW opening note for the logical Cage that 'uses type information itself as the wall of the cage.' Claim 1 verbatim leads 18 dependent claims and fences a four-step formal preprocessing procedure: (B1) maintaining data type snapshots for operand stack and registers, (B2A-D) propagating type information to successor instructions through virtual emulation, and (B3) iterating processing of marked instructions. Original Assignee Sun Microsystems Inc was acquired by Oracle Corporation on January 27, 2010, and the Current Assignee is Oracle America Inc. As the Day 28 Cage Patents axis SW Open opening note, this article excavates (1) reading Claim 1 verbatim as a type cage, (2) the contrast with the 'eligibility wall' forms of Day 25 ep88 SW-002 FORTRAN 1957 / ep89 BBN IMP 1969 / ep90 Smalltalk 1972 / ep91 LISP 1958, (3) Sun's two-tier patent strategy contemporaneous with the Java sandbox (the language itself was free, while only the verifier was patented), and (4) the three-form parallel with ep98 Bell-LaPadula policy-model cage absence and ep99 Hardy KeyKOS capability cage success.

On June 2, 1947, Earl Silas Tupper (1907–1983) of Leominster, Massachusetts sole-filed 'Open mouth container and nonsnap type of closure therefor' in the United States. It was granted as US2487400A on November 8, 1949 and expired by lifetime on November 8, 1966. Claim 1 fences a resilient and locally distortable polyethylene closure with a central wall, peripheral rim, and inner / outer walls forming a groove that receives the container's edge, plus an offset portion on the outer wall — when the rim is forced over the container's mouth, the outer wall flexes outward and creates a seal, and pressing the central wall produces a partial vacuum inside the container. This fenced an FH cage that confines food away from air with a polyethylene burping seal. Original / Current Assignee both Individual (no formal assignment to Tupperware Plastics Company / Tupperware Corporation on this patent's record). **Important DB correction:** the DB-registered patent number **US2613000A is in fact 'Towel or cloth holder' (Dudley E. Moore sole inventor, filed 1950-08-15, granted 1952-10-07, Original / Current Assignee Individual) — a complete mis-identification.** This memo detected the mis-identification through Day 27's WebFetch and confirmed the correct number US2487400A from the primary source. Day 27 / Cage Patents axis FH Open opening memo.

On June 5, 1972, Felix Theeuwes and Takeru Higuchi at Alza Corporation in Palo Alto, California co-filed 'Osmotic dispensing device for releasing beneficial agent' in the United States. It was granted as US3845770A on November 5, 1974. Claim 1 fences an osmotic device whose semipermeable wall holds the drug inside while allowing external fluid to enter — internal pressure rises by osmotic gradient, and the drug is pushed out at a constant rate through a laser-drilled orifice (described in the specification, not in Claim 1). On this patent's extension stand blockbuster controlled-release tablets such as Concerta (methylphenidate, ADHD), Procardia XL (nifedipine, hypertension), and Glucotrol XL (glipizide, diabetes). The DB consensus 'sole inventor Theeuwes' is **incorrect**: the patent's inventor field lists Theeuwes and Takeru Higuchi as **co-inventors** (Higuchi was a central figure in pharmacokinetics at the University of Kansas Pharmacy School, and was also involved as a consultant to Alza founder Alejandro Zaffaroni). This memo confirms the primary-source URL and Claim 1 — full specification unread — and organizes (1) the scope of the title and Claim 1, (2) the DB correction 'sole Theeuwes,' (3) the pharmacological-history meaning of Higuchi's contribution, and (4) the Cage-axis distance to modern microneedle patches / liposomal LNPs / mRNA-vaccine LNPs. Day 27 PH Open opening memo.

On July 30, 1959, Robert Norton Noyce, a Fairchild Semiconductor founding member (then 31 years old, vice president) filed 'Semiconductor device-and-lead structure' in the United States as the sole inventor. It was granted as US2981877A on April 25, 1961 and expired by lifetime on April 25, 1978. Claim 1 covers a structure in which a metal lead runs over a silicon-surface oxide layer (SiO2) and bridges a pn junction. This fenced the path to electrically isolating multiple devices on the same wafer. Kilby's US3138743A (covered in Day 17 / ep62) was filed 5 months earlier, but its Au flying-wires wiring scheme was unsuited to mass production; Noyce's oxide-isolated planar approach became the material substrate that connects continuously to modern CMOS. Original / Current Assignee both Fairchild Semiconductor Corp; Noyce is sole inventor, with an implicit dependence on Jean Hoerni's planar process patent US3025589A (Fairchild, contemporaneous, separate filing) for oxide passivation. In the 1969 patent litigation between Texas Instruments and Fairchild, the USPTO ruled that 'Noyce and Kilby are co-inventors,' and in 2000 Kilby received the Nobel Prize in Physics alone (Noyce, who died in 1990, was ineligible). As the Day 27 Cage Patents axis HW Open opening note, this article excavates (1) Claim 1 verbatim and reading the oxide isolation as an electrical cage, (2) the 5-month gap with Day 17 / ep62 Kilby and the divergence in mass-production wiring schemes, (3) the implicit dependence on Hoerni's planar process US3025589A (a separate filing), (4) positioning within the Cage axis material variations of Day 19 ep70 Masuoka flash, ep71 CCD, and ep72 HA gel, and (5) the continuity of the oxide-isolation step in modern semiconductor manufacturing — TSMC 2nm / Samsung GAA / Intel 18A in 2026.

Day 26 SW subseries' sixth memo (Eligibility Wall (a) pre-judicial era × (c) government-contract hybrid form). Verification: 1959-04-08 US DoD Charles Phillips (Director, Data System Research Staff) convened major computer vendors and government agencies at the Pentagon → 1959-06-04 CODASYL (Conference on Data Systems Languages, also Committee on Data Systems Languages) was founded → the Short-Range Committee (6 commercial vendors + 3 government agencies + Wegstein as chair) drafted the spec on top of Grace Hopper's FLOW-MATIC → 1960-01-08 the executive committee approved it → the Government Printing Office published 'COBOL 60' → 1960-08-17 first compile succeeded on the RCA 501 → 1960-12-06/07 RCA-Univac compatibility demo. Wikipedia EN COBOL / Grace Hopper / FLOW-MATIC, the Yale CS Hopper Story, Britannica Hopper, and gracehoppers.wordpress.com FLOW-MATIC explainer all carry no COBOL-related patent number. 1960 sits 12 years before Gottschalk v. Benson, in the pre-judicial era; the cooperative form of DoD government funding ($200M invested, 225 machines in service) + 6 commercial vendors + 3 government agencies stands as the hybrid of pre-judicial era (a) and government contract (c). Together with Day 25 ep88 SW-002 FORTRAN ((a-1) corporate-lab solo type), today's ep91 SW-009 LISP ((a-2) pure academic-disclosure type), and ep92 SW-008 ALGOL 60 ((a-3) international-committee cooperative type), this is the fourth sub-form.

Day 26 SW subseries' fifth memo (third instance of Eligibility Wall (a) pre-judicial era / international-committee cooperative sub-form). Verification of the ALGOL 60 spec, drafted at the Paris meeting on January 11-16, 1960 by a 13-member international committee: Wikipedia EN ALGOL 60, the dl.acm.org CACM 1960-05 paper (DOI 10.1145/367236.367262), CACM 1963 6(1):1-17 Revised Report (DOI 10.1145/366193.366201), the ACM Turing Award Naur 2005 page, the ACM Turing Award Perlis 1966 page, and Bauer's career profile all carry no patent number. 1960 sits 12 years before Gottschalk v. Benson, in the pre-judicial era; the cooperative-disclosure form across IFIP (International Federation for Information Processing) + ACM + GAMM (Gesellschaft für Angewandte Mathematik und Mechanik) supplies the (a-3) international-committee cooperative sub-form within pre-judicial era (a). Together with Day 25 ep88 SW-002 FORTRAN ((a-1) corporate-lab solo type), today's ep91 SW-009 LISP ((a-2) pure academic-disclosure type), and ep93 SW-010 COBOL ((a-4) government-contract hybrid type), this is the third sub-form.

Day 26 SW subseries' fourth note (third instance of Eligibility Wall (a) pre-judicial era). Starting from McCarthy's solo September 1958 MIT AI Memo No. 1, 'An Algebraic Language for the Manipulation of Symbolic Expressions,' the four-year LISP genesis runs through Steve Russell hand-coding eval on the IBM 704 in fall 1958, the April 1960 publication of Communications of the ACM 3(4):184-195 'Recursive Functions of Symbolic Expressions and Their Computation by Machine, Part I,' and Tim Hart + Mike Levin's first proper compiler in 1962. Verification: Wikipedia EN Lisp, the dl.acm.org CACM 1960-04 paper (DOI 10.1145/367177.367199), MIT AI Memo No. 1 via Stoyan 1984 ACM Symposium, McCarthy's own 'History of Lisp' PDF (jmc.stanford.edu/articles/lisp/lisp.pdf), and the ACM Turing Award Laureate McCarthy 1971 page all carry no LISP-related patent number. Together with Day 25 ep88 SW-002 FORTRAN (pre-judicial era #1) and today's ep92 SW-008 ALGOL 60 (international-committee form, #2) plus ep93 SW-010 COBOL (government-contract hybrid form, #4), this becomes the third instance in the SW subseries' Eligibility Wall (a) pre-judicial era form. Across McCarthy's 30-year career — FP language proposal, situation calculus, circumscription — every contribution was published as a paper with no patent record, and the 'shared academic-community culture' of contemporaries Backus / Hoare / Dijkstra / Knuth / Naur / Perlis underpinned the pure form of pre-judicial era (a). Set alongside today's ep92 ALGOL 60 international-committee form and ep93 COBOL government-contract hybrid form, this gives form (a) a three-sub-form resolution.

Day 25 SW subseries' fifth memo. Verification of Xerox PARC Smalltalk: from Alan Kay's Smalltalk-71 design (1971) → Smalltalk-72 (Dan Ingalls implements the first interpreter in about 700 lines of BASIC for the Data General Nova, October 1972) → April 1973 Xerox Alto port → Smalltalk-76 → November 1981 Smalltalk-80 Version 1 release (Adele Goldberg documents). Wikipedia EN Smalltalk, Wikipedia EN Alan Kay, Britannica Alan Kay, Lemelson-MIT Alan Kay, and ACM Turing Award Alan Kay 2003 page all carry no patent reference. Xerox PARC distributed Smalltalk-80 in 1981 to Tektronix / Hewlett-Packard / Apple Computer / DEC under 'unrestricted redistribution' terms, and made it fully public in the August 1981 ACM Computing Surveys 'Special Issue on Smalltalk.' Unlike the IMP's ARPA-mandated public disclosure, this was a voluntary disclosure under Xerox's own corporate judgment, positioning it as the eligibility wall's corporate-strategy form within the SW subseries.

Day 25 SW subseries' fourth memo. Verification of the IMP project: ARPA awarded a 4-node-network contract to BBN (Bolt, Beranek and Newman) in January 1969; Kahn solo-authored BBN Report 1822 'Specification for the Interconnection of a Host and an IMP' in April 1969; the first IMP was deployed at UCLA's Kleinrock lab on 1969-08-30; the first host-to-host connection was made between UCLA and SRI at 22:30 PST on 1969-10-29. Wikipedia EN Interface Message Processor, Wikipedia EN ARPANET, ACM Turing Award Robert E. Kahn 2004 page, and LivingInternet IMP article all contain no patent reference. Because ARPA government funding made the deliverables a public technical specification, BBN Report 1822 was later confirmed as Internet STD 39 and remains fully published at rfc-editor.org. Together with Day 25 ep88 SW-002 FORTRAN ('pre-judicial era' form), this is Eligibility Wall #3 in the form '(c) public disclosure mandated by government contract.'

Day 25 SW subseries' third note (Eligibility Wall #2). Verification of FORTRAN, with timeline: late-1953 Backus solo proposal, 1954-11 draft spec 'The IBM Mathematical Formula Translating System,' 1956-10-15 first reference manual 'The Fortran Automatic Coding System for the IBM 704,' April 1957 first compiler shipment. Google Patents query (inventor=Backus/assignee=IBM/priority 1953-1960) returns 0 results. ACM Turing Award Backus official page, Wikipedia EN John Backus, and Britannica Backus all contain no patent reference. 1957 predates US software-patent-eligibility case law by 15 years; Gottschalk v. Benson (1972) / Parker v. Flook (1978) / Diamond v. Diehr (1981) / State Street Bank (1998) / Alice Corp v. CLS Bank (2014) are the five later Supreme Court / Federal Circuit decisions that subsequently shaped 'pure-algorithm-invention eligibility.' FORTRAN sits 15 years before this case-law arc. IBM's three-pronged strategy — early manual disclosure (1956-10-15), free customer distribution, and source-code trade secrecy — overlaps with 2026 LLM foundation-model strategies (private weights / public APIs / public papers). Combined with Day 24 SW-005 HyperCard (1987 — eligibility-unsettled era post-Diamond v. Diehr), this completes Eligibility Wall forms #1 and #2 of the SW subseries. Today's ep89 SW-003 BBN IMP (1969 ARPA contract / public spec) and ep90 SW-004 Xerox PARC Smalltalk (1972 corporate lab / 4-licensee public release) round out the four-form lineup, raising resolution on 'eligibility wall' into four sub-forms.

Day 24 SW subseries' third memo. Verifying the patenting status of HyperCard, started by Atkinson alone in 1985 (codename WildCard), augmented with Dan Winkler's HyperTalk scripting language in 1986-87, announced at MacWorld Boston on 1987-08-11, and assigned to Apple under the 'bundle free with every Mac' condition. WebSearch, Wikipedia EN HyperCard page, Wikipedia EN Bill Atkinson page, Apple Wiki Fandom, History of Information, OSnews, and Science Museum Group all carry no HyperCard patent number. The only Atkinson-sole, Apple-assigned patent confirmed is US4622545 (image compression, 1986, treated in today's ep85). HyperCard's core elements (cards, stacks, HyperTalk, buttons, scripts, links) have no record of being claimed. Day 24's SW-001/SW-006/SW-007 (Engelbart Mouse / Atkinson region / Lapson + co-inventor cursor control) succeeded at patenting with **claims that include hardware elements**, while HyperCard, being a **pure-software, OS-application**, likely fell into the late-1980s US software-patent eligibility-unsettled-period wall. Recorded as the SW subseries' first 'Eligibility Wall' form.

Day 24 SW subseries' second memo. The Apple ball-type mouse patent US4464652, filed 1982-07-19, granted 1984-08-07, expired, pulled from patentimages.storage.googleapis.com PDF (14 pages, 1.08 MB). The first inventor field reads 'William F. Lapson, Cupertino,' but the second-inventor field is OCR-garbled in pdftotext -raw output as 'g-lfykihsm L°S Gatos' — the actual name not identifiable from the PDF. WebSearch returns 'William D. Atkinson' as co-inventor, but Wikipedia EN's Bill Atkinson page makes no reference to US4464652, and secondary sources (uspto.report, etc.) are blocked by Cloudflare so cross-check is unreachable. Claim 1 verbatim is ball-type (rotatable ball + domed portion + cut-outs + X-Y position indicating means + biasing means + lock cap/lock tabs/lock ridges removability mechanism), fundamentally different in detection principle from Engelbart Mouse US3541541's wheel type (position wheels + transducer + flexible conductor). Recorded as the SW subseries' first 'Information Wall — OCR Form.'

Day 24 launches the SW subseries' second note. The subject is the Atkinson image compression and manipulation patent US4622545, filed 1982-09-30 and granted 1986-11-11. (1) Inventor 'William D. Atkinson, Los Gatos, Calif.' — sole, assigned to Apple Computer, Inc., Cupertino, Calif. before the Mac launch. (2) Appl. No. 428,635. Filed Sep. 30, 1982. Granted Nov. 11, 1986. 4 years 1 month from filing to grant. (3) GATT-pre filing, so 17-years-from-grant rule applies — expired 2003-11-11. (4) Claim 1 verbatim has three means: display means (a plurality of selectively enabled display elements) + memory means (storing a plurality of inversion points, each having a coordinate that specifies orthogonal lines forming the boundaries of a contrasting area) + processing means (enabling display elements corresponding to inversion points and generating contrasting areas, with contrast as a function of previously displayed inversion-point coordinates). Per scanline, an ordered list of 'points where color changes' represents arbitrary closed regions. (5) Subsequent claims describe AND/OR/NOT/XOR logic operations between ordered lists of inversion points, and source-bitmap/destination-bitmap region transfer — the conceptual core of QuickDraw RgnHandle. (6) Modern SVG clip-path, HTML5 Canvas Path2D + clip(), iOS UIBezierPath bool ops, polygon-collider hit detection in game engines, and image-processing flood fill all share the problem-statement 'efficiently represent and compute arbitrary closed regions on screen.' Implementation differs (scanline-based vs. path-based), but the lineage is continuous for 40 years. With SW-001 Engelbart Mouse (filed 1967, 'point at a spatial position with the hand') as the input-side origin, this patent forms the display-side origin of the Apple Lisa/Mac GUI patent map.

Day 22 handover (a) CS-005 avobenzone origin patent USPTO direct search executed. Wikipedia en Avobenzone records only the single line 'patented in 1973' with no number, and PubChem CID 51040 / Pharmacompass / Smartskincare (smartskincare.com) / ScienceDirect Avobenzone overview / Univar Solutions / The Cosmetic Chemist / MakingCosmetics / CosDNA / Wikidoc / EWG Skin Deep / Incidecoder / Specialchem also lack the 1973 origin patent number. The Wikipedia patent citation is an incomplete OpenURL with 'rft.cc=US&rft.number=0' — the number field is empty. AI-generated candidate numbers (CH561168A5, CH570454A5, US3920734, DE2354187, likely Search AI Overview output) all return zero direct hits via WebSearch and patents.google.com is Captcha-blocked, so entity verification is impossible. The assignee transition is Roure Bertrand Dupont SA (Société Anonyme, France/Paris, possibly 1971-12-16 CH priority) → 1991 Givaudan and Roure merger → mid-1990s Givaudan-Roure → 2000 spinoff from Roche to DSM Nutritional Products → 2010s and later DSM holds the Parsol® 1789 trademark (currently dsm-firmenich.com manages BMI). FDA OTC approval was 1988 (US market entry 10 years after EU 1978 approval). USPTO Patent Public Search and Espacenet Worldwide Search both require interactive UIs; curl/WebFetch hits browser rendering, Captcha, or REST 403 access denial. Following Day 22 ep80 CS-010 Sansho Seiyaku kojic acid (J-PlatPat wall) and Day 23 ep83 CS-004 Nivea Lifschütz (DPMA wall), confirmed as cosmetic form-3 information wall excavation #3. This memo records the third instance of the finding 'origin patent number lies behind an information wall' and leaves USPTO Patent Public Search interactive search (assignee=Roure Bertrand Dupont, 1971-1974 priority) or DSM/Givaudan direct inquiry as the next action.

Day 22 handover (a) CS-004 Nivea/Eucerit origin patent DPMA direct search executed. Lifschütz's 1902 DRP number is not recorded in any of: (1) Beiersdorf official About Us page, (2) Beiersdorf official Chronicle 12 PDF (5.7 MB, pdftotext extracted body and captions), (3) Wikipedia en Eucerin / de Isaac Lifschütz, (4) German Chemie-Schule Eucerit/Lifschütz entries, (5) German PharmaWiki Wollwachsalkohole, (6) German Jüdische Allgemeine 'Der Doktor aus Pinsk,' (7) Beiersdorf official Eucerin Brand History. The Beiersdorf Chronicle 12 PDF contains a caption explainer '1) Patent number — This number precisely identifies the patent and allows it to be referenced at the patent office. Dr. Isaac Lifschütz laid the foundation for dermatological skin care with the patent for a method of producing highly water-absorbent ointment bases.' — but the actual number itself does not appear in body text. Early-1900s DRP records are outside DPMA DEPATISnet's electronic-coverage scope. Espacenet and Google Patents DE coverage centers on 1920+. AI-generated candidate numbers (DRP 132307 / DRP 154959 / DRP 171146, etc.) cannot be verified via WebSearch (patents.google.com 503/Captcha, Espacenet REST 403). Following Day 22 ep80 CS-010 Sansho Seiyaku kojic acid, confirmed as cosmetic-subseries form-3 information wall excavation #2. This memo records the finding 'origin patent number lies behind an information wall' for the second time, leaving DPMA paper-archive direct inquiry or Beiersdorf direct inquiry as the next action.

Day 23 launches a new AI Archaeology subseries: 'Software/UI Patents.' The first subject is the Engelbart mouse patent US3541541, 'X-Y POSITION INDICATOR FOR A DISPLAY SYSTEM,' filed 1967-06-21, granted 1970-11-17, expired 1987-11-17. Pulled from the Google Patents PDF archive at patentimages.storage.googleapis.com/7d/13/7e/3fd1de4c37ed12/US3541541.pdf, the front-matter inventor field reads 'Douglas C. Engelbart, Palo Alto, Calif., assignor to Stanford Research Institute, Menlo Park, Calif., a corporation of California' — a single inventor, not Bill English co-invention. Bill English was an SRI Augmentation Research Center implementation engineer who managed the 1968 Mother of All Demos demo run, but the patent's inventor field does not list him. Claim 1 verbatim claims four elements: (1) a housing, (2) first/second position wheels with rotation axes orthogonal and rims protruding from the housing to contact the surface, (3) transducer means generating digital position signals from wheel rotation, (4) flexible conductor means connecting the transducer to the computer while enabling unrestrained movement. Filed June 21, 1967, Ser. No. 647,872. Patented Nov. 17, 1970. 8 Claims total. The 1995 GATT predates this filing, so the 17-years-from-grant rule applies and the patent expired 1987-11-17. Modern mice (optical/laser), trackpads (capacitive), and Vision Pro hand-tracking (LiDAR) differ in detection principle from Engelbart's mechanical wheels, but the upper-layer design intent 'point at a spatial position with the hand → cursor follows' has been continuous for 60 years. SW subseries launch as a change of pace from the 5-day Phase 1 cosmetic streak (Day 18-22).

Day 22 attempted to excavate CS-002 Allergan Botox origin patent. The DB number US4932936 turned out to be University of Minnesota Twin Cities' 'Method and device for pharmacological control of spasticity' (Dennis D. Dykstra / Abraham A. Sidi co-invention, filed 1988-01-29, granted 1990-06-12) — completely unrelated. Further investigation across multiple secondary sources (Wikipedia Alan B. Scott, Medicine 2023 paper 'Early development history of Botox', ScienceDirect obituary 'Professor Alan B. Scott 1932-2021: The inventor of Botox') confirmed that the Allergan Botox origin patent itself does not exist. Scott's lab patent policy declared items / processes of possible patentability, but the lawyers determined that Scott's 1973 paper (Scott AB, Rosenbaum A, Collins CC. Pharmacologic weakening of extraocular muscles. Invest Ophthalmol Vis Sci 12: 924-927, 1973) had already publicly disclosed the invention's core — botulinum toxin A injection into extraocular muscles for strabismus correction — voiding patentability, so Scott decided not to file. Without a patent, no pharma company would take over manufacturing. Scott took out a home loan in the 1980s, founded Oculinum Inc., handled production and trials in-house, obtained FDA approval on 1989-12-29 (strabismus and blepharospasm), and sold to Allergan in 1991 (rebranded as Botox). FDA approved Botox Cosmetic for glabellar lines in 2002, growing into a multi-billion-dollar blockbuster. This memo records the second installment of treating 'patent absence' as a primary source in AI Archaeology, structurally identical to Day 10 ep42 PH-004 Köhler-Milstein monoclonal antibody (Nature 1975 paper voided patentability, MRC declined to file, Wistar Institute's derivative US4172124A was peripheral).

Day 22 attempted to excavate CS-010 Sansho Pharma kojic acid skin-whitening origin patent. The number JP S58-118507 handed off from Day 21 turned out to be Lion Corp's 'transparent toothpaste' patent (filed 1982-01-06, disclosed 1983-07-14, Uematsu / Ichikawa co-inventors), unrelated to Sansho Pharma. None of the five public resources we surveyed — Google Patents, Wikipedia, Sansho Pharma's official history page, Kose's research report (after redirect to koseholdings.co.jp), or the Dermed research report — list the origin patent number for Sansho Pharma kojic acid. Confirmed public-history facts: (1) Sansho Pharma discovered kojic acid's tyrosinase inhibition (melanin-synthesis enzyme suppression) in 1975, (2) after 13 years of R&D, in 1988 it became Japan's first quasi-drug whitening active approval, (3) Kose launched 'Kose Whitening Cream XX' in 1990, (4) MHLW issued a notice in March 2003 temporarily suspending kojic acid use in quasi-drugs, (5) on 2005-11-02 the suspension was lifted after additional safety tests showed no problem. J-PlatPat is JS-heavy and authenticated for WebFetch, so the origin patent number identification requires direct query to Sansho Pharma or interactive J-PlatPat search. This memo excavates the 'origin patent number is hard to identify from public web resources' information wall as the first installment.

We added CS-009 'P&G Niacinamide Skin-Whitening Patent US5939082' to candidates.tsv on Day 22 and pulled the primary source from Google Patents. Claim 1 verbatim turned out to be 'a method of regulating mammalian skin pore size,' the independent claim, while the term 'skin lightening agents' in the specification appears only as a list of optional auxiliary components — kojic acid, arbutin, ascorbic acid and derivatives. The widely circulated industry catchphrase 'P&G niacinamide skin-whitening patent' (1) ignores the patent's actual Claim 1 subject (pore size), and (2) confuses niacinamide with the parallel skin-lightening agents list in the specification. The patent's title is 'Methods of regulating skin appearance with vitamin B3 compound,' Original/Current Assignee is Procter and Gamble Co, and inventors are Donald Lynn Bissett / John Erich Oblong / Kimberly Ann Biedermann (the 'Bissett alone' folk wisdom needs correction). Priority 1995-11-06, Filing 1997-04-11, Grant 1999-08-17, expiration 2015-11-06. Day 21 ep76 promised the correct P&G niacinamide skin-whitening patent. The fact-check shows it is not a whitening patent at all — it is a pore-size regulation patent. Day 20-22 cumulative cosmetic-subseries DB corrections reach 49 entries with this note's three additions (inventor count, Claim 1 subject, catchphrase-misreading structural finding).

After ep76 confirmed the candidates.tsv CS-008 'Kojic acid whitening Ajinomoto US4767628' swap, this memo reads US4767628's reality — Imperial Chemical Industries Ltd (ICI, now AstraZeneca)'s Francis G. Hutchinson 'PLA / PLGA × polypeptide drug two-phase sustained-release pharmaceutical composition' patent — at Claim 1 verbatim level and translates it into pharmaceutical delivery context, organizing its position as the core patent of AstraZeneca's Zoladex® goserelin (LH-RH agonist 1-month / 3-month sustained-release implant) and its distant relation to Sculptra PLLA filler (medical aesthetics collagen biostimulator). Decomposes Claim 1 into six elements: (1) polylactide-system matrix (PLA homopolymer or PLA-PLGA copolymer; lactic acid in racemic or optically active form), (2) polylactide content 50% to 99.999% w/w, (3) polylactide viscosity range (benzene-soluble 0.093-0.5; benzene-insoluble 0.093-4), (4) pharmacologically active polypeptide content 0.001% to 50% w/w, (5) polypeptide molecular weight ≥ tetragastrin (four or more amino acid residues, not significantly hydrolyzed under use conditions), (6) two-phase sustained release in aqueous physiological environment (initial diffusion / late matrix degradation). Further organizes the ICI Pharmaceuticals → Zeneca → AstraZeneca lineage and the clinical importance of Zoladex® goserelin (1989 UK / 1990 US approval, currently maintaining $300-500M annual global sales for prostate, breast cancer, endometriosis), and the market scale of successor PLA-PLGA sustained-release products (Lupron Depot, Trelstar, Sandostatin LAR, Vivitrol, Bydureon, etc.). No direct cosmetic relevance, but Sculptra (PLLA filler, US Galderma, collagen biostimulator) widely used in medical aesthetics differs from this only in 'whether the same in vivo PLA / PLLA degradation reaction is used for drug release or collagen induction,' and is added to Day 22+ candidates for **medical aesthetics subseries spin-off** consideration. Patent presumed expired since 2008-08-30 (Grant + 20 years) (USPTO PAIR / litigation records need separate verification); PLA / PLGA × polypeptide sustained-release delivery forms can now be freely incorporated into new generics.

After ep76 confirmed the candidates.tsv CS-007 'Niacinamide whitening patent P&G US5851538' swap, this memo reads US5851538's reality — Advanced Polymer Systems Inc (APS, parent of Microsponge®)'s low-irritation retinol porous microsphere formulation patent — at Claim 1 verbatim level and translates it into cosmetic delivery context. Co-inventors Michael Froix / Masha Pukshansky / Sergio Nacht (Sergio Nacht is an APS co-founder with dermatology / pharmacy [PharmD] background, long responsible for Microsponge® application development). Original Assignee APS Inc, Application 1995-12-29, Grant 1998-12-22, US Patent US5,851,538 A. Claim 1 verbatim: 'A stable pharmaceutical composition for topical administration of retinol consisting of an oil-in-water emulsion comprising, suspended in the emulsion, solid water-insoluble microscopic particles containing a substantially continuous network of pores open to the exterior of the particles and a retinoid composition comprising retinol residing only in the pores, the composition causing lower irritancy when applied than a composition containing the same concentration of retinol in a non-(microscopic particle) formulation.' This memo decomposes Claim 1 verbatim into seven elements: (1) oil-in-water emulsion form, (2) microparticle solid / water-insoluble / microscopic size, (3) continuous pore network, (4) external pore opening, (5) retinol composes the retinoid composition, (6) retinol resides only within the pores, (7) lower irritancy compared to a non-microparticle formulation of the same concentration. It then traces the lineage of five major delivery products in the APS Microsponge® extension (Ortho-McNeil's Retin-A Micro 0.1% gel / Galderma's Differin Gel 0.1% / P&G/J&J Olay Total Effects / Johnson & Johnson RoC Retin-Ox / later 0.5-1.0% retinol face creams). Finally, if this 1998-Grant patent is fully expired (Grant + 20 years = 2018-12 expire), modern free-market retinol high-concentration skincare (The Ordinary Retinol 1%, La Roche-Posay Pure Retinol 0.3, Olay Regenerist Retinol 24, etc.) can freely incorporate Microsponge / Polytrap / Microcapsule delivery forms (other related patents need separate verification via J-PlatPat / USPTO PAIR). Adds to Day 22+ excavation candidates: positioning as cosmetic subseries Phase 2 'delivery patents' candidate, and survey of Microsponge / Polytrap / Microcapsule adoption in modern high-concentration retinol products.

Patents registered in candidates.tsv as CS-007 'Niacinamide (Vitamin B3) whitening patent (P&G) US5851538' and CS-008 'Kojic acid whitening patent (Ajinomoto) US4767628' were retrieved from Google Patents in Day 21 and found to be neither niacinamide nor kojic acid nor whitening patents. CS-007 US5851538 turned out to be 'Retinoid formulations in porous microspheres for reduced irritation and enhanced stability,' co-inventors Michael Froix / Masha Pukshansky / Sergio Nacht, Original Assignee Advanced Polymer Systems Inc (APS, the parent of Microsponge® technology), Application 1995-12-29, Grant 1998-12-22. Claim 1 verbatim covers 'a stable pharmaceutical composition for topical administration of retinol consisting of an oil-in-water emulsion comprising solid water-insoluble microscopic particles containing a substantially continuous network of pores open to the exterior of the particles and a retinoid composition comprising retinol residing only in the pores' — the lineage of Ortho-McNeil's Retin-A Micro 0.1% gel (FDA approval 1997). CS-008 US4767628 turned out to be 'Continuous release pharmaceutical compositions,' sole inventor Francis G. Hutchinson, Original Assignee Imperial Chemical Industries Ltd (ICI, later Zeneca, now AstraZeneca), Application 1987-07-01, Grant 1988-08-30. Claim 1 verbatim covers '50% to 99.999% polylactide (homopolymer of lactic acid or copolymer of lactic acid and glycolic acid) and 0.001% to 50% pharmacologically active polypeptide (molecular weight at least that of tetragastrin, four or more amino acid residues), uniformly dispersed, exhibiting two-phase release in aqueous physiological environment' — the lineage of AstraZeneca's prostate / breast cancer therapy Zoladex® (goserelin, LH-RH agonist 1-month / 3-month sustained-release implant). Neither patent has any connection with whitening, niacinamide, or kojic acid. DB corrections in this note: CS-007 patent number / ingredient / inventor / assignee / use / Claim 1 subject (six items, #36-#41), and CS-008 same six items (#42-#47), totaling twelve. Cosmetic subseries DB corrections across Days 20-21 now total 11 cases / 47 label discrepancies. The finding: cosmetic subseries DB contains a structural mis-pattern of pharmaceutical drug-delivery patents masquerading as cosmetic ingredient patents — a different class of error from Day 20 ep74-75 'cosmetic DB contains a non-cosmetic patent' swaps. Day 22+ tasks: (a) excavate the correct P&G niacinamide whitening patent (candidate US5939082 etc.), (b) excavate the correct kojic acid whitening patent (candidate JP S58-118507 from Sansho Pharmaceutical), (c) establish a number-validity check procedure for the cosmetic subseries DB.

When US3839566, registered in candidates.tsv as 'CS-005 Avobenzone (Parsol 1789) sunscreen patent,' was retrieved from Google Patents, the title turned out to be 'Compositions for topical application to animal tissue and method of enhancing penetration thereof,' with inventors Francis S. Kilmer MacMillan / Warren I. Lyness and assignee Procter & Gamble (P&G). Priority 1970-05-15, Grant 1974-10-01. Claim 1 covers 'a pharmacologically active steroid plus 0.1-10.0% of a C8-C12 aliphatic methyl sulfoxide selected from 8 species (octyl / nonyl / decyl / undecyl / dodecyl / 2-hydroxydecyl / 2-hydroxyundecyl / 2-hydroxydodecyl methyl sulfoxide).' No relation to Givaudan, Hoffmann-La Roche, or avobenzone. This patent is P&G's proprietary delivery technology — an extension of DMSO-class percutaneous penetration enhancers to higher-chain (C8-C12) aliphatic sulfoxides for medical and cosmetic topical formulations. The correct origin patent for avobenzone (butyl methoxydibenzoylmethane, Parsol 1789) is the **1973 Givaudan patent** (Wikipedia / industry reports), with the number unconfirmed at the time of writing. The next excavation step is direct USPTO Patent Public Search using 'Givaudan' / 'butyl methoxydibenzoylmethane' / '1973.' Day 20 cumulative DB corrections updated to 35th by this memo (CS-005 patent-number mix-up); cosmetic subseries accumulated 7 corrections within Day 20 alone (CS-006: 5, CS-004: 1, CS-005: 1).

When DE200619C, registered in candidates.tsv as 'CS-004 Nivea (oil-in-water emulsion) patent,' was retrieved from Google Patents, the title turned out to be 'Bienenkorbkühler' (honeycomb cooler — a motor radiator), the assignee was 'Firma C. Schnewindt in Neuenrade i.W.' (a German mechanical-engineering firm), and the grant date was 1907-01-16. Claim 1 covers a radiator structure of woven tubes and wires sealed at their joints with solder — a complete mechanical engineering patent unrelated to Beiersdorf, Lifschütz, Troplowitz, Nivea, or Eucerit. This memo records the find as a 'failed excavation log' in the same pattern as IR Archaeology #1. Lifschütz's Eucerit manufacturing process ('Eucerinum anhydricum,' 'highly water-absorbent salve bases') was filed in Germany in 1900 and granted as a Deutsches Reichspatent in 1902. Beiersdorf purchased the patent from Lifschütz in 1911, the same year Troplowitz, Unna, and Lifschütz launched Nivea Creme. The correct DRP number is unconfirmed at the time of writing; Beiersdorf's official history and the Eucerin brand history do not state the number, and direct search of the DPMA (German Patent and Trademark Office) archive is the next excavation step. Day 20 cumulative DB corrections: 29-33 from CS-006 (ep73) plus the 34th from this memo (CS-004 patent-number mix-up).

On March 7, 1975, Ruey J. Yu and Eugene J. Van Scott of Temple University Medical School Department of Dermatology filed US Patent 4105782 'Treatment of acne and dandruff.' Filing 1976-09-07, Grant 1978-08-08. Inventors: Yu (first) / Van Scott (co-inventor). Original Assignee: 'Individual' — no record of assignment to Temple University. Claim 1 requires (1) topical application to involved areas of acne or dandruff, (2) at least one of 16 listed α-hydroxy / α-keto / β-hydroxy acids (citric, glycolic, glucuronic, galacturonic, glucuronolactone, gluconolactone, α-hydroxybutyric, α-hydroxyisobutyric, malic, pyruvic, β-phenyllactic, β-phenylpyruvic, saccharic, mandelic, tartaric, tartronic, β-hydroxybutyric), and (3) reaction with ammonia or an organic amine (primary, secondary, tertiary alkylamine; alkanol amine; diamine; etc., C1-C8) to form 'amide and ammonium salts.' Abstract: 'addresses defective keratinization, 1-20% in solutions, lotions, creams, shampoos.' The patent became the claim-level origin of (1) NeoStrata's AHA skincare line (Yu / Van Scott co-founded NeoStrata in 1988), (2) chemical peels (30-70% glycolic acid in clinic), (3) LHA / PHA / gluconolactone derivatives, (4) US FDA's OTC AHA concentration and pH regulation discussions in the 1990s, (5) AHA-formulated quasi-drug products in Korea and Japan. Five DB corrections: (1) inventor order (DB 'Van Scott / Yu' → reality 'Yu first / Van Scott co-inventor'), (2) Original Assignee (DB 'Temple University license' → reality 'Individual / privately held'), (3) use (DB 'AHA peeling' → Claim 1 verbatim 'acne or dandruff treatment'), (4) compound scope (DB 'glycolic acid' → Claim 1 verbatim '16 α-hydroxy / α-keto / β-hydroxy acids'), (5) formulation requirement (DB 'topical application of acid' → Claim 1 verbatim 'reaction product of acid and base required'). Day 20 cumulative DB corrections updated to 29th / 30th / 31st / 32nd / 33rd by this single note.

On December 6, 1984, two researchers at Biomatrix Inc—Endre A. Balazs (Hungarian-born ophthalmologist and biomaterials scientist, founder of the Matrix Biology Institute and former Columbia University faculty) and Adolf Leshchiner—co-filed 'Cross-linked gels of hyaluronic acid and products containing such gels.' The US filing was 1985-03-08, granted as US4636524A on January 13, 1987. Claim 1 covers 'a delivery system for a substance having biological or pharmacological activity, said system comprising a molecular cage formed of a cross-linked gel of hyaluronic acid or a mixed cross-linked gel of hyaluronic acid and at least one other hydrophilic polymer co-polymerizable therewith and having dispersed therein a substance having biological or pharmacological activity and which is capable of being diffused therefrom in a controlled manner.' The Claim 1 verbatim writes 'molecular cage formed of a cross-linked gel of hyaluronic acid,' explicitly requiring the molecular cage of Day 19's 'Cage Patents' axis. The patent became the **patent-claim starting point** for (1) Restylane (Q-Med, 1996 EU approval / 2003 FDA approval), (2) Juvederm (Allergan, 2006 FDA approval), (3) joint-injection drugs Synvisc / Hyalgan (osteoarthritis treatment), (4) the ophthalmic Healon (cataract surgery), and (5) all crosslinked HA fillers used routinely in aesthetic dermatology. Two DB corrections: (1) DB 'Balazs (Columbia University) solo' → reality 'Balazs + Adolf Leshchiner two co-inventors,' and (2) DB 'Columbia University / Matrix Biology Institute assignee' → reality 'Original Assignee Biomatrix Inc / Current Assignee Sanofi Biosurgery Inc' (Biomatrix is the company Balazs co-founded in 1981, distinct from Columbia University). Day 19 cumulative DB corrections: 27th / 28th. This memo retrieves title, inventors, assignees, dates, and Claim 1 from Google Patents, but the full specification (details on crosslinking agents such as divinyl sulfone, formaldehyde, 1,4-butanediol diglycidyl ether; the relationship to Restylane / Juvederm derivative patents) is unread.

On April 19, 1973, two Bell Telephone Laboratories researchers—Willard Sterling Boyle and George Elwood Smith—co-filed 'Buried channel charge coupled devices' in the United States, granted as US3792322A on February 12, 1974. Claim 1 covers 'a charge coupled device with a planar charge storage medium and multiple discrete electrode field plates, capable of controllably moving charge between charge-storage sites by applying appropriate electrical bias, characterized by the storage medium being bounded on all sides by electrical barriers, depleted via ohmic contact at a limited area, with potential minima existing physically within the storage medium, interior of the boundary electrical barriers.' Claim 1 verbatim writes 'potential minima existing physically within the storage medium, interior of the boundary electrical barriers,' explicitly requiring the charge cage (confining charge in a potential well) of Day 19's 'Cage Patents' axis. The Original Assignee was 'Individual,' but a Certificate of Correction retroactively assigned it to Bell Telephone Laboratories Murray Hill NJ—an unusual transfer history. Inventors, Bell Labs assignment, filing date, and grant date all match the DB (13th match confirmed). In 2009 Boyle/Smith received the Nobel Prize in Physics for the CCD invention. This memo retrieves title, inventors, assignees, dates, and Claim 1 from Google Patents, but the full specification (relationship to the 1969 Boyle/Smith Bell System Technical Journal paper, and the generational shift to CMOS image sensors) is unread.

On December 20, 1980, two researchers at Tokyo Shibaura Electric Co Ltd (now Toshiba)—Fujio Masuoka and Hisakazu Iizuka—co-filed a 'Semiconductor memory device and method for manufacturing the same' in the United States, granted as US4531203A on July 23, 1985 (US filing 1981-11-13, Japan priority 1980-12-20). Claim 1 covers a memory device whose cells, arranged in matrix form, are built from a three-layer gate structure: erase gate, floating gate, and control gate, with the third insulating film thicker than the first. This is the precursor to modern NAND flash memory, and the entire infrastructure of SSDs, smartphone storage, microSD cards, SD cards, USB memory, and LLM weight-file storage rests on the principle of confining electrons inside a floating gate, as introduced here. The patent predates the term 'flash' by four years (Masuoka coined it in his 1984 IEDM paper). Claim 1's core is the **electrical isolation of the floating gate by three layers of gates with asymmetric insulating films**, enabling separate erase, write, and read operations. Two DB corrections: (1) DB 'Masuoka solo' → reality 'Masuoka + Iizuka two co-inventors' (same pattern as Day 17 ep64 Goodenough/Mizushima, Day 18 ep68 Nakamura/Mukai/Iwasa); (2) DB '1982 grant' → reality 'grant 1985-07-23' (priority 1980-12-20, filing 1981-11-13). Day 19's three-piece set 'Cage Patents' note slot, where Claim 1 explicitly requires a structure 'a portion of said floating gate overlapping a part of said erase gate with a first insulating film being interposed' that physically confines electrons in a cage.

On July 17, 1980, Ching W. Tang of Eastman Kodak Company solo-filed 'Organic electroluminescent cell' with the US Patent Office. The patent was granted as US4356429A on October 26, 1982. Claim 1 covers an improvement to an electroluminescent cell (anode, cathode, luminescent zone between the electrodes comprising an organic luminescent agent and a binder with breakdown field strength of at least about 10⁵ V/cm), the improvement being a hole-injecting zone of a porphyrinic compound layer between the luminescent zone and the anode. Both the Original Assignee and Current Assignee are Eastman Kodak Co. The inventor is **Ching W. Tang solo** — the popular industry narrative that 'Tang & Van Slyke co-invented OLED' does not match this patent's inventor field. Steven A. Van Slyke is a co-author of the 1987 Appl. Phys. Lett. 51:913 'Organic electroluminescent diodes' paper presenting the diamine HTL + Alq3 EML two-layer structure, but he is absent from this patent's inventor field. This is the same '**divergence between paper / textbook / award reporting and patent inventor names**' pattern as Day 11 ICI beta-blockers (James Black absent), Day 12 sildenafil (some researchers absent), and Day 9 PCR (Mullis solo award vs six co-inventors). Day 8–18 has accumulated **23 corrections** with **12 match confirmations**. This memo is at primary-URL-confirmed / Claim 1 retrieved / specification full text unread stage and organizes (1) the development from single-layer porphyrin to the 1987 two-layer structure, (2) Eastman Kodak's spinoff, bankruptcy, and OLED patent succession, (3) technology transfer to Samsung Display / LG Display / Chinese BOE / TCL CSOT OLED mass-production factories, and (4) connections to foldable OLEDs (Samsung Galaxy Z Fold/Flip, Huawei Mate X, Xiaomi Mix Fold) and OLED adoption from iPhone X onward.

On November 20, 1992, three Nichia Chemical Industries Ltd researchers (Anan, Tokushima)—Shuji Nakamura, Takashi Mukai, and Naruhito Iwasa—co-filed 'Light-emitting gallium nitride-based compound semiconductor device' with the Japan Patent Office. They filed a US continuation on November 17, 1993, granted as US5578839A on November 26, 1996. Claim 1 covers a double-heterostructure GaN-based light-emitting device comprising (a) an active layer with first and second major surfaces formed of a low-resistivity In_x Ga_(1-x) N (0 less than x less than 1) compound semiconductor doped with an impurity, (b) a first clad layer of n-type GaN-based compound semiconductor having a different composition from the active layer, joined to the first major surface, and (c) a second clad layer of low-resistivity p-type GaN-based compound semiconductor having a different composition from the active layer, joined to the second major surface. The Original Assignee is Nichia Chemical Industries Ltd (the company name at filing, headquartered in Anan, Tokushima); the Current Assignee is Nichia Corporation (after the 2003 corporate name change). Among the three inventors, Shuji Nakamura is widely recognized as the inventor of the patent's core—p-type GaN crystal growth technology and the InGaN quantum well structure—but **Takashi Mukai and Naruhito Iwasa also appear as co-inventors** in the patent, so the DB entry 'Shuji Nakamura solo' is corrected. Day 8–18 has accumulated **22 corrections** with **12 match confirmations** (the correction sequence resumes after the three consecutive matches at Day 17 ep65/66 / Day 18 ep67). The DB Original Assignee 'Nichia Corporation' is also incorrect; at filing the company name was **Nichia Chemical Industries Ltd** (used 1956–2003). This memo is at primary-URL-confirmed / Claim 1 retrieved / specification full text unread stage, and organizes (1) the meaning of the three-name co-invention structure, (2) Nakamura's compensation lawsuit against Nichia (filed 2001, Tokyo District Court ¥20 billion judgment 2004, Tokyo High Court ¥840M settlement 2005), (3) the inventor-field separation in the 2014 Nobel Prize in Physics (Akasaki / Amano / Nakamura), and (4) the connections to modern Samsung / LG / Chinese BOE OLED / Mini-LED TVs and smartphone white-LED light sources.

On March 5, 1954, three researchers at Bell Telephone Laboratories—Daryl M. Chapin (electrical engineer, Basking Ridge NJ), Calvin S. Fuller (chemist, Chatham NJ, semiconductor impurity diffusion specialist), and Gerald L. Pearson (physicist, Bernards Township NJ, semiconductor physicist)—co-filed 'Solar energy converting apparatus' with the US Patent Office. Three years later, the patent was granted as US2780765A on February 5, 1957. Claim 1 covers an arrangement for utilizing solar radiation to keep a storage battery charged, comprising (a) a storage battery to be charged, (b) at least one photosensitive element including a silicon body with an n-type zone contiguous with a p-type zone including a concentration of boron impurities (with the thickness of the p-type zone of the order of the diffusion length of electrons), and (c) a unilaterally-conductive element serially connected with the storage battery and photosensitive element, poled to pass charging currents developed by the photosensitive element and to block discharging currents from the battery through the photosensitive element. The Abstract claims efficiencies exceeding 5%, marking the first practical-line crossing over the prior Charles Fritts selenium cell (1883, under 1% efficiency) and Russell Ohl's silicon p-n junction discovery patent at Bell Labs (1941, US2402662A). The Original Assignee is Bell Telephone Laboratories, Incorporated (AT&T's research subsidiary), and the Current Assignee is AT&T Corp. All inventors, assignees, priority date, and grant date match the DB entry—the 12th match confirmation in the Day 8–17 correction sequence (which has accumulated 21 corrections), following the back-to-back HW-007/008 matches in Day 17. The Patent Family was filed in the Netherlands, Switzerland, France, Germany, Japan, and the United Kingdom—six countries, reflecting Bell Labs' ambitious global rollout strategy for solar cells in the 1950s. This note examines (1) the meaning of Claim 1's three-element configuration (battery + photosensitive element + unilaterally-conductive element), (2) the relationship to Russell Ohl's 1941 silicon p-n junction discovery patent (same Bell Labs but different lineage), (3) the April 25, 1954 public demonstration at AT&T's New York headquarters with radio and telephone driven by solar power, (4) the 1958 Vanguard 1 satellite installation (NASA / US ARPA, the first major commercialization in space applications), (5) the relationship to the 1956 AT&T consent decree and the royalty-free compulsory licensing of transistor patents, (6) technology transfer from Bell Labs to Japan (Sharp's research start in 1959, commercialization in 1963), Germany (Siemens), and US RCA, (7) the material design distance to modern monocrystalline Si (Tongwei, JinkoSolar, LONGi), PERC, TOPCon, HJT, and perovskite cells, and (8) the contemporary significance of China's ~80% dominance of the solar cell supply chain (IEA 2024) viewed through the lens of this 72-year-old material foundation.

On July 14, 1967, Robert H. Dennard at IBM's T.J. Watson Research Center solo-filed 'Field-effect transistor memory' in the United States, granted on June 4, 1968 as US3387286A. Claim 1 covers an integrated circuit memory comprising multiple memory cells coupled to word lines and bit lines, each cell comprising (a) an input field-effect transistor with a channel region and a gate electrode, (b) a storage device exhibiting capacitance with one electrode connected to the transistor and the other to a reference potential, (c) a word line connected to the gate electrode, (d) a bit line connected to one terminal of the transistor, (e) control means applying voltage signals to the word line, the bit line, and the reference potential to charge the capacitance, and (f) the bit-line signal being ineffective on the capacitance unless the word-line signal renders the transistor conductive. The Abstract states that each cell is formed using a single field-effect transistor and a single capacitor, with information storage occurring through capacitor charging and periodic regeneration necessary because charge leaks over time; the body notes that, in the worst case, refresh is needed 'every 200 microseconds,' consuming about 10% of memory operation time. This patent is **the core patent fencing the minimal 1T1C (1 Transistor – 1 Capacitor) DRAM cell as of 1968**; it expired on June 4, 1985 due to US patent term. The inventor, assignee, filing year, and grant year all match the DB record (the 10th match in the Day 8–16 correction streak, the second consecutive match in Day 17 after HW-007). It must be clearly distinguished from the Dennard scaling rule formalized in Dennard's 1974 co-authored paper 'Design of ion-implanted MOSFETs with very small physical dimensions' (IEEE J. Solid-State Circuits 9(5), 256–268)—the patent is a structural invention (1T1C cell), and the scaling paper is an empirical rule (constant power density when voltage, current, and dimensions are scaled at the same ratio). This memo retrieves Claim 1, the inventor, the assignee, the dates, and the '200 µs refresh' reference from Google Patents, but the full specification (correspondence with the IBM Type 369 memory product, patent relationships with 1970s competitors such as Intel 1103 and Mostek MK4096) has not been read.

On January 22, 1973, Federico Faggin, Marcian E. 'Ted' Hoff Jr., and Stanley Mazor at Intel Corporation jointly filed 'Memory system for a multi chip digital computer' in the United States, granted on June 28, 1974 as US3821715A. Claim 1 covers a general-purpose digital computer comprising (a) a central processor on a first semiconductor chip, (b) a plurality of bidirectional data bus lines, and (c) at least separate first and second semiconductor memory chips, each defining a memory and including a chip decoding circuit that recognizes a different predetermined code on the bidirectional data bus lines and activates a portion of its memory upon receipt of the predetermined code. The Abstract describes a configuration in which a group of MOS chips (RAMs and ROMs) couple to a CPU via common bidirectional data buses; the body uses 'central processing unit,' 'MOS,' and 'single MOS chip' repeatedly, but **does not mention** Busicom or 4004 by name. This patent covers the **bus architecture of the Intel MCS-4 family (4004 CPU / 4001 ROM / 4002 RAM / 4003 shift register), not the structural patent of the 4004 alone**. The three inventors match the DB record (the 9th match in the Day 8–16 correction streak). Federico Faggin's contribution extends beyond the bus configuration in this patent to the silicon-gate MOS process (US3597469A, filed 1968 while at Fairchild) that made the 4004 manufacturable, while Masatoshi Shima, sent by Busicom as a design collaborator and deeply involved in the 4004's logic design, is **absent from this patent's inventor field**. This memo confirms the primary URL, has retrieved Claim 1, but has not read the full specification. It positions the patent as a starting point for clarifying (1) the precise interpretation of the title, (2) the distinction between 'MCS-4 family patent' and '4004-alone patent,' (3) the absence of Shima from the inventor field, and (4) the distance to modern AI-chip bus architectures (Nvidia H100, Cerebras WSE-3, China's Cambricon, Google TPU).

On April 5, 1979, John B. Goodenough, professor at Oxford's Inorganic Chemistry Laboratory, filed a UK priority application for an electrochemical cell with new fast ion conductors together with Koichi Mizuchima (a visiting researcher from the University of Tokyo). The US continuation was filed on March 31, 1980 and granted on November 24, 1981 as US4302518A. Claim 1 covers an ion conductor of the formula AxMyO2 with α-NaCrO2 layered structure (A is Li, Na, or K; M is a transition metal; x < 1, y ≈ 1) whose A+ cation vacancies have been created by A+ cation extraction. This fences the LiCoO2 (lithium cobalt oxide) layered cathode as a material patent in the United States. The Original Assignee is 'Individual,' which was transferred to the United Kingdom Atomic Energy Authority (UKAEA) in 1984, then to AEA Technology PLC in 1997, and the current assignee is Ricardo AEA Ltd. The inventor field lists only Goodenough and Mizuchima; P.C. Jones and Philip J. Wiseman, who appear as co-authors on the parallel Mater. Res. Bull. 15(6), 783–789 (1980) paper, are absent from the patent's inventor field. Sony commercialized the world's first Li-ion battery in 1991 by combining Yoshino Akira's lithium-ion intercalation graphite anode (US4668595A, 1985 filing, Asahi Kasei) with Goodenough's LiCoO2 cathode, providing the material foundation for smartphones, laptops, data center UPS systems, and electric vehicles. The 2019 Nobel Prize in Chemistry was awarded jointly to Goodenough (age 97 at the time of the award, the oldest Nobel laureate in history), M. Stanley Whittingham (TiS2 system, Exxon), and Yoshino Akira (graphite anode, Asahi Kasei). This note examines (1) the correction of DB entries claiming 'Mizushima/Jones/Wiseman co-inventors' and 'University of Oxford assignee,' (2) the divergence between paper authors and patent inventors, (3) the meaning of the 1984 UKAEA transfer, (4) the material substrate of modern AI data centers, smartphones, and EVs, and (5) design divergence with LFP/NMC/all-solid-state systems.

On September 23, 1969, in Philadelphia, PA, University of Pennsylvania dermatology professor Albert M. Kligman filed US Patent US3729568 'Acne treatment' as sole inventor. Claim 1 requests a method of 'applying vitamin A acid topically to the affected area in a concentration effective for the treatment of acne and repeating such applications daily, whereby said applications result in the peeling and diffuse redness of the skin treated, and continuing said treatment until the acne has subsided.' The patent was granted April 24, 1973 and expired April 24, 1990. Both original and current assignee are Johnson & Johnson (New Brunswick, NJ). The invention's background includes Kligman and J.E. Fulton's late-1960s clinical research, including trials on inmates at Holmesburg Prison (now criticized as an ethics issue). In 1971, FDA approved Retin-A (prescription tretinoin) for acne vulgaris, backed by this patent. In 1986, Kligman et al. published 'Topical tretinoin for photoaged skin' in J Am Acad Dermatol (not NEJM), first describing histological improvement against photoaging. In 1995, FDA approved Renova (tretinoin 0.05% emollient cream) as the first photoaging-indication drug. Kligman coined 'cosmeceutical' in 1984, popularizing it as a concept that redefines the FDA cosmetic-vs-drug boundary (FDA does not legally recognize this category; tretinoin is treated as a drug because it changes structure/function, prescription only in the US). In Japan, tretinoin is treated as a drug under the Pharmaceutical Affairs Act, prohibited from cosmetic and quasi-drug formulations (only available via personal import). This memo confirms primary URL and Claim 1 retrieval, but the full specification body remains unread, and the Cosmetic Archaeology subseries (Kao / L'Oréal / Nivea sensitive-skin axis) differentiation axis is also addressed. DB corrections (2 items: inventorship, 1986 paper journal) follow the Day 8~15 19-correction series.

On February 6, 1959, in Dallas, Texas, the new Texas Instruments engineer Jack S. Kilby filed US Patent US3138743 'Miniaturized electronic circuits' as sole inventor. Claim 1 requests 'a plurality of junction transistors defined in a single-crystal semiconductor wafer (each transistor with thin opposite-conductivity layers on one major face providing base and emitter regions overlying a collector region, with the base-emitter and base-collector junctions both extending wholly to that major face) + a plurality of thin elongated regions in the wafer providing semiconductor resistors + conductive means selectively connecting them.' Background: Kilby joined TI in May 1958 and conceived the IC during the company-wide vacation period when, as a new hire, he had no vacation rights. On September 12, 1958, he assembled an oscillation circuit on a germanium mesa p-n-p transistor and confirmed continuous sine-wave output on an oscilloscope — the world's first working IC. The patent was granted June 23, 1964 and expired June 23, 1981. In parallel, Robert N. Noyce of Fairchild Semiconductor filed US2981877 'Semiconductor device-and-lead structure' on July 30, 1959 (about 5 months after Kilby), granted April 25, 1961. Noyce, building on Hoerni's planar process, achieved monolithic structure with on-chip Al wiring; whereas Kilby's chip used thin gold (Au) flying wires between elements, a hybrid structure that is, strictly speaking, not a monolithic IC. Commercial ICs followed the Noyce lineage. Kilby received half of the 2000 Nobel Physics Prize (the other half to Alferov and Kroemer for heterostructures); Noyce, having died in 1990, was ineligible. In his Nobel lecture, Kilby remarked, 'Bob ought to have been here.' This memo confirms primary source URL and Claim 1, but the full specification text remains unread. We position this as a starting point for organizing what's analogy and what's similarity in the modern China RISC-V vs. ARM vs. x86 instruction-set split.

On February 26, 1948, at Bell Telephone Laboratories in Murray Hill, NJ, physicist John Bardeen and experimentalist Walter H. Brattain filed US Serial No. 11,165 (later abandoned, parent of US2524035) titled 'three-electrode circuit element utilizing semiconductive materials.' The patent was refiled as a continuation-in-part on June 17, 1948 (Serial No. 33,466) and granted on October 3, 1950 as US2524035. The invention covers a circuit element comprising (a) a semiconductor block of one conductivity type, (b) a thin surface layer of opposite conductivity type, (c) an emitter electrode contacting the surface layer, (d) a collector electrode collecting current spreading from the emitter, and (e) a base electrode contacting the body. Claim 1 covers this 5-element combination. The design core is 'a 3-terminal semiconductor amplifying element,' which is the basic patent of the point-contact transistor. The absence of William Shockley from the inventor list reflects Bell Labs' internal division of labor: Shockley was not directly involved in the December 23, 1947 working demo, which was established by Bardeen's surface states theory and Brattain's experimental technique. Shockley filed a separate junction transistor patent US2569347 as sole inventor on June 26, 1948, granted September 25, 1951, distinct from the Bardeen-Brattain patent. The 1956 Nobel Physics Prize was awarded jointly to all three, but the patent inventorship remained at two. Furthermore, on January 24, 1956, the US v. Western Electric & AT&T antitrust consent decree placed all pre-decree Bell Labs semiconductor patents (including this one) under royalty-free compulsory licensing for all applicants (RCA/GE/Westinghouse excluded due to existing cross-licenses). This enabled Texas Instruments, IBM, Sony, Toshiba, NEC, Hitachi, Samsung, and TSMC to enter at low cost, becoming the institutional origin of global semiconductor industry diffusion (Watzinger, Fackler, Nagler & Schnitzer 2020 economic history paper rates this consent decree above Marshall Plan in economic contribution). This article excavates (1) the verbatim Claim 1 of US2524035, (2) the design split between point-contact and junction, (3) the circumstances of Shockley's absent name, (4) the institutional effect of the 1956 consent decree, (5) connection to the contemporary China-AI × Korea-Taiwan semiconductor niche, and (6) positioning as the inaugural episode of Week 4 'Hardware & Energy Patent' subseries.

Filed Sep 1, 1959; granted Aug 23, 1960. US patent US2950228A 'Enzymatic process'. Sole inventor Richard O. Marshall (Argo, Illinois); original assignee Corn Products Company (New York, N.Y.); current assignee Unilever Bestfoods North America (succession Corn Products → Bestfoods → Unilever). Claim 1 covers a process for converting dextrose (glucose) to levulose (fructose) by incubating dextrose at ≥ 0.2 molar with an enzyme preparation containing xylose isomerase. The genesis is 1957: Marshall and Earl R. Kooi published 'Enzymatic Conversion of D-Glucose to D-Fructose' in *Science* (Vol. 125, Issue 3249, pp. 648–649, April 5, 1957), reporting the surprising finding that an *Aeromonas hydrophila* xylose isomerase also acted on glucose at concentrations above ~0.2 M. Kooi co-authored the *Science* paper but is not on this patent's inventor list (the candidates.tsv 'Marshall & Kooi' framing is wrong). Commercialization came in 1967, when the Clinton Corn Processing Company implemented glucose isomerase at industrial scale; in 1968 it introduced batch fructose production using immobilized enzymes (42% fructose). Through the 1970s, HFCS-42 / HFCS-55 supplied US soft drinks at scale and became the principal sweetener in major brands like Coca-Cola and Pepsi — and the entry point of the HFCS-vs-obesity debate. Together with Day 13 ep51 (aspartame US3492131A) and Day 14 ep55 (Flavr Savr US4801540A), this is part of a 60-year food-and-health-patent lineage. URL confirmed; Claim 1 retrieved; full specification unread. The candidates.tsv entry US3379608 is a USGypsum mineral-wool building product (1968), wholly unrelated, corrected during Day 15. **Publication of this memo closes Week 3 (Food / Health + Pharma) — 22 entries across FH-001..012 and PH-001..010.**

Filed August 7, 1884; granted June 2, 1885. US patent US319082A 'Manufacture of Saccharine Compounds.' Sole inventor Constantin Fahlberg (then resident in New York; German-American chemist), with one-half assigned to Adolph List (Leipzig, Germany) — the patent cover page reads 'ASSIGNOR OF ONE-HALF TO ADOLPH LIST.' Claim covers an eight-step synthesis from toluene and coal-tar derivatives yielding a 'new sweet compound' (later named benzoic sulfinide / saccharin). The discovery dates to 1879, when Fahlberg, while working in Ira Remsen's lab at Johns Hopkins on o-toluenesulfonamide chemistry, noticed a sweet taste on his fingers at dinner. Fahlberg patented alone; Remsen claimed shared discovery credit but was kept off the inventor list. A follow-on patent US326281A (1886, improved saccharin compound) and a reissue USRE10667E followed. The patent connects to Theodore Roosevelt's 1907 intervention on saccharin regulation ('anyone who refuses saccharin is an idiot'), the 1977 FDA-proposed ban followed by a Congressional moratorium, the 2000 NTP delisting from carcinogen rosters, and the 2010s artificial-sweetener portfolio (aspartame, sucralose, etc.). Day 13 ep51 (aspartame US3492131A) cites this patent in its 'comparison with saccharin and cyclamate' passage. URL confirmed; Claim 1 retrieved; full specification unread. The candidates.tsv entry US284081 is a sawmill set-works patent — wholly unrelated — corrected during Day 15.

Filed July 18, 1938; granted October 30, 1945. Earl W. Flosdorf — the University of Pennsylvania microbiologist who, with Stuart Mudd in December 1933, achieved the first successful freeze-drying of human serum — and three F.J. Stokes Machine Co. co-inventors (Charles J. Westin, Francis Joseph Stokes Jr., Edith Wolfrom Westin) co-filed US patent US2388134A 'Biological apparatus, container, and method'. Claim 1 covers an apparatus consisting of a desiccant chamber, a solid-porous regenerable chemical desiccant, multiple outlets for container connection, valves at approximately 45° above horizontal, and vacuum-producing means — i.e., dehydration of biological material via sublimation plus chemical-desiccant water absorption. The context: the December 1933 Mudd–Flosdorf first human-serum lyophilization, the 1935 Flosdorf–Greaves first commercial lyophilizer, and the wartime plasma-drying program (Sharp & Dohme mass-producing dried blood plasma for forward field hospitals). URL confirmed; Claim 1 retrieved; full specification unread. The candidates.tsv entry — 'multiple inventors; multiple patents in 1920s–1930s' — is vague, so this memo positions US2388134A as a representative commercial-lyophilizer patent.

On January 5, 1983, Luca Comai, a plant molecular biologist at the Davis-based startup Calgene LLC, filed US patent US4535060A as sole inventor. Claim 1 covers a culture of cells containing a mutated 5-enolpyruvyl-3-phosphoshikimate synthetase (EPSPS, encoded by the aroA structural gene) whose inhibition by glyphosate (the active ingredient of Monsanto's Roundup) is less than about half that of the wild-type enzyme. The mutation is introduced in vitro at the aro locus. Granted August 13, 1985. The patent became one of Calgene's flagship IP assets. The commercial Roundup Ready soybean Monsanto launched in 1996 used a different lineage — the soil-bacterium-derived CP4 EPSPS (Agrobacterium sp.), covered by a separate patent (US5633435A) — so Comai's aroA mutant is not the direct ancestor of commercial Roundup Ready. The two are parallel solutions to the same problem: (a) Comai = mutate an existing plant EPSPS; (b) CP4 = bring in a naturally tolerant bacterial enzyme. After Monsanto's full acquisition of Calgene in 1997, this patent passed into the Monsanto portfolio. This note covers (1) Claim 1 verbatim from Google Patents; (2) the EPSPS / aroA / shikimate-pathway biochemistry; (3) the design split between Comai's mutant and CP4 EPSPS; (4) Calgene's two-shot IP structure together with Day 14 ep55 Flavr Savr (US4801540A); (5) the line through to commercial Roundup Ready soy (1996); and (6) corrections to candidates.tsv where 'Monsanto Company; filed around 1986' is wrong on multiple counts. A significant correction case in the Day 8–14 streak.

Filed Nov 1 1873, granted Dec 30 1873. Hippolyte Mège (also known as Mège-Mouriès, 1817–1880), a Paris-based pharmacist and food chemist, is sole inventor of US146012A 'Improvement in Treating Animal Fats'. The work originated in 1869 when Napoleon III opened a public call for an affordable, shelf-stable butter substitute for the army and the working class. Mège-Mouriès won the French patent on July 15, 1869 (the world's first margarine patent), and re-filed in the US in 1873. The claim covers a nine-step process for treating animal fats: (1) neutralizing ferments, (2) crushing fat cells, (3) concentrated digestion using artificial gastric juice, (4) crystallization separating stearin from oleomargarine, (5) centrifugal separation, (6) conversion of oleomargarine into a butter-like material via mammary pepsin, and (7) processing stearin into superior stearic acid for candles. By the time of the 1873 US filing, four years had passed since the French patent, but US law accepted it as a fresh filing. URL confirmed; Claim 1 retrieved; full specification unread. **DB correction: the candidates.tsv-listed number US119428 is actually John F. Taylor's 1871 'feed-water heater for steam-boilers' patent, completely unrelated. The substantive primary source is US146012A.**

Priority Oct 17 1986; US filing Jan 2 1987; granted Jan 31 1989. Five plant-molecular-biology researchers at Calgene LLC (Davis, California) — William R. Hiatt, Raymond E. Sheehy, Christine K. Shewmaker, Jean C. Kridl, Vic Knauf — co-invent US4801540A 'PG gene and its use in plants'. The claim covers a DNA sequence encoding tomato polygalacturonase (PG) flanked at the 5′ or 3′ end by at least one non-wild-type sequence. Expressed in antisense orientation, the construct suppresses PG, slows pectin hydrolysis, and yields a 'shelf-stable tomato.' In May 1994 Flavr Savr received FDA clearance and went on sale as the world's first commercial GMO food. Production ended around 1996 due to commercial failure. Calgene was fully acquired by Monsanto in 1997 and the patent transferred. URL confirmed; Claim 1 retrieved; full specification unread.

On May 30, 1978, three pharmacists at Bristol-Myers Co. (Edmund S. Granatek, Gerald M. Ziemba, Frederick L. Grab) filed US patent US4310515A 'Pharmaceutical compositions of cisplatin'. The compound itself — cis-diamminedichloridoplatinum(II), cis-Pt(NH₃)₂Cl₂ — had been synthesized in 1845 by Michele Peyrone and described in Liebigs Annalen der Chemie 51 as 'Peyrone's salt'. In 1893 Alfred Werner deduced its octahedral coordination structure. In 1965 Barnett Rosenberg at Michigan State University accidentally discovered, during electrolysis experiments, that the soluble platinum complex inhibited binary fission in E. coli, and in 1969 he reported antitumor activity in Nature 222 (385–386). Clinical trials began in 1971. But by 1978 the compound was 133 years old — long-public art, ineligible under the novelty requirement of 35 U.S.C. § 102. Bristol-Myers' Claim 1 instead covered an 'injectable, stable, sterile aqueous solution of cisplatin in a unit dosage form in a sealed container, at a concentration of 0.1 to 1.0 mg/mL and a pH in the range of 2.0 to 3.0.' This bypassed prior lyophilization-and-refrigeration practice and gave Bristol-Myers a room-temperature ready-to-use product. FDA approved the drug as Platinol on December 19, 1978 for testicular and ovarian cancers, and Bristol-Myers locked the 1980s cisplatin market through this formulation patent. This piece walks the five-layer IP stack — Peyrone 1845 / Werner 1893 / Rosenberg 1965 & 1969 / Research Corp's malonato platinum patent US4140707A (Cleare/Hoeschele/Rosenberg/Van Camp, priority 1972) / Bristol-Myers' formulation patent US4310515A (1978) — and records a heavy DB correction: the candidates.tsv number US4169726 actually points to General Electric's casting-alloy patent (Norman P. Fairbanks, 1977), entirely unrelated to cisplatin.

Filed May 2, 1911 by Kikunae Ikeda and Saburosuke Suzuki, US1015891A claims a process for separating the products of albumin hydrolysis into basic, acidic, and neutral groups by electrolysis. This is the commercial-manufacturing US patent paired with Ikeda's 1908 Japanese discovery patent (JP14805, partially covered in ep10). The candidates.tsv URL (JP1908000009380) returns 404; US1015891A is adopted as the substantive primary source. Patent URL confirmed; Claim 1 retrieved; full specification unread.

Filed July 1, 1939 by Roy J. Plunkett, US2230654A claims polymerized tetrafluoroethylene (PTFE) as a substance — Claim 1 is just three words: 'Polymerized tetrafluoroethylene.' The original assignee is Kinetic Chemicals Inc., a DuPont/GM joint venture, not DuPont directly (database correction). Patent URL confirmed; Claim 1 retrieved; full specification and litigation history unread.

On April 18, 1966, James M. Schlatter at G.D. Searle & Co. (Chicago) filed US3492131A 'Peptide Sweetening Agents.' Claim 1 covers five esters of aspartylphenylalanine (methyl, ethyl, n-propyl, isopropyl, tert-butyl) in four stereochemical configurations (LL / DLDL / DL-L / L-DL), describing them as '100–200 times as sweet as sucrose' and free of the unpleasant aftertaste of saccharin or cyclamate. The discovery was accidental: in December 1965, Schlatter was synthesizing an intermediate for the anti-ulcer drug tetragastrin, got crystals on his finger, turned a page of his notebook, later licked the finger, and noticed strong sweetness. The patent was reassigned to NutraSweet Company in 1986. The FDA approval history is unusual: initial approval 1974, suspended 1975 over animal-test data integrity (Bressler Report), public hearing 1980, reapproved July 1981 for dry foods, expanded to carbonated beverages July 1983 — about sixteen years end to end. Donald Rumsfeld, who served as Searle CEO from 1977 to 1985, led the reapproval push. This note records (1) Claim 1 verbatim from Google Patents, (2) the meaning of the LL/DLDL/DL-L/L-DL stereo requirement, (3) the 1986 NutraSweet reassignment, (4) the FDA timeline, (5) the post-1983-expiration generic market, and (6) connections to today's artificial sweetener portfolio (aspartame / sucralose / acesulfame K / steviol glycosides / advantame). One database correction: candidates.tsv lists the patent year as '1965,' but 1965 is the lab discovery year — the actual filing date is April 18, 1966.

On May 11, 1945, microbiologist Andrew J. Moyer at the **USDA Northern Regional Research Laboratory** in Peoria, Illinois, filed U.S. Patent US2442141A 'Method for production of penicillin' as sole inventor, granted on May 25, 1948. The assignee is '**United States of America, as represented by the Secretary of Agriculture**'—the U.S. government (USDA representative). Claim 1 claims a method 'incubating a penicillin-producing mold in contact with an aqueous nutrient medium containing an assimilable carbon source 5–150 g/L and degraded proteinaceous material 5.0+ g/L, with carbon source added stepwise during the incubation period.' The specification **explicitly applies to both surface culture in Fernbach flasks and submerged state in tank fermenters**, and the application to agitated, aerated submerged culture in particular has historical significance as the WW2-era mass production technology. Alexander Fleming discovered the antibacterial action of *Penicillium notatum* in 1928, and Howard Florey/Ernst Chain/Norman Heatley established purification and therapeutic application at Oxford University in 1939–1941. But Florey sought U.S. support due to insufficient British production capacity, and visited the Peoria USDA Northern Lab with Heatley in July 1941. With Heatley's collaboration, Moyer (a) introduced **corn steep liquor**, a Peoria industrial byproduct of the time, as a nitrogen and growth-factor source, (b) used lactose as a carbon source, and (c) established submerged fermentation, increasing *Penicillium chrysogenum* productivity by **several hundred-fold**. This enabled the U.S. Army to secure 2.3 million doses of penicillin for the Normandy landings (D-Day) in June 1944, beginning the full-scale era of antibiotic medicine. This patent should be read as a **manufacturing-method patent** alongside the story 'Fleming/Florey/Chain did not patent the discovery,' and the British side criticized after the war that 'the U.S. took the patent and put us in a position of paying royalties' (in fact, USDA = U.S. government, not Moyer personally). This article excavates: (1) Claim 1 verbatim from US2442141A as obtained on Google Patents, (2) Florey/Heatley's Peoria visit in July 1941, (3) the three-element breakthrough of corn steep liquor + lactose + submerged fermentation, (4) the mass supply for D-Day in 1944, (5) the contrast between the Nobel Prize (1945 Fleming/Florey/Chain) and the patent, (6) the resonance with modern recombinant fermentation, antibody pharmaceuticals, and bioreactors. The DB (PH-009) description (USDA, Andrew J. Moyer, 1945 filing/1948 grant, Fleming did not patent but the manufacturing process was patented) is consistent with the primary source—the **3rd 'DB-agreement-confirmed' case** (after Day 11 aspirin and Day 12 norethindrone).

On June 20, 1990, three chemists at Pfizer Central Research (Sandwich research site) in the UK—Andrew S. Bell, David Brown, and Nicholas K. Terrett—established the priority date for U.S. Patent US5250534A 'Pyrazolopyrimidinone antianginal agents.' The patent claims a class of compounds with a **pyrazolo[4,3-d]pyrimidinone scaffold** as cGMP-PDE selective inhibitors. The specification includes as a specific example the compound later named sildenafil (UK-92,480, Viagra brand). The original target indications were **cardiovascular diseases such as angina pectoris, hypertension, and heart failure**, with the specification explicitly listing 'stable, unstable and variant (Prinzmetal) angina.' Within the cGMP-PDE inhibitor exploration program ongoing at Pfizer Sandwich since 1989, Nicholas Terrett led molecular design, and UK-92,480 was synthesized in 1989, with limited efficacy for angina found in initial clinical trials in 1991-1992. Side-effect observations (penile erection) in a small Phase I trial in 1992 began the indication switch to ED treatment, and on March 27, 1998, the FDA approved Viagra (sildenafil 25/50/100mg) as a treatment for **erectile dysfunction (ED)**. The DB row PH-006 in `~/ai-archaeology/db/candidates.tsv` lists 'Pfizer Inc., Peter Dunn / Albert Wood' / '1992' but Google Patents primary source confirms three errors: (a) Original Assignee is **Pfizer Corp SRL** (Pfizer's UK subsidiary), (b) inventors are **Andrew S. Bell / David Brown / Nicholas K. Terrett three co-inventors**, with Peter Dunn and Albert Wood **not listed in the inventor field of this patent**, (c) Priority date is **1990-06-20** (DB '1992' is the year of U.S. filing date 1992-05-14 only). Peter Dunn was a medicinal chemist at Pfizer Sandwich and may be listed as inventor on separate patents (later use patents or synthesis process improvement patents), but is absent from the body of US5250534A. Albert Wood was likely affiliated with the Pfizer Sandwich analytical chemistry team but is absent from the inventor field of this patent. In the consecutive DB-error correction series of Days 8/9/10/11, this is **the 10th case of DB-error discovery and correction** (after IC-009 Bluetooth, IC-011 CDMA, IC-012 RFID on Day 8; PH-003 PCR, PH-001 statin on Day 9; PH-004 monoclonal antibody, PH-002 insulin on Day 10; PH-007 captopril, PH-005 propranolol on Day 11).

On November 22, 1951, three researchers at Mexico's Syntex SA—Carl Djerassi (Austrian-born, emigrated to the U.S. under Nazi rule), Luis E. Miramontes (Mexican chemistry student, performed the actual synthesis), and George Rosenkranz (Hungarian-born, head of research at Syntex)—established the priority date for U.S. Patent US2744122A, 'delta 4-19-nor-17alpha-ethinylandrosten-17beta-ol-3-one and process.' Claim 1 encloses both the compound (a 17α-ethynyl-17β-ol-3-one in the 19-norandrostene series) and a six-step process: reduction of an estrone lower alkyl ether with alkali metal in liquid ammonia → mineral acid hydrolysis → chromic acid oxidation → selective formation of a 3-enol ether → treatment with acetylene in the presence of an alkali metal alkoxide → mineral acid hydrolysis. The core of this patent is **17α-ethynyl-19-nortestosterone (later named norethindrone / norethisterone)**, the first synthetic progestin to retain activity by oral administration. Russell Marker's 1944 establishment of mass progesterone synthesis from diosgenin obtained from wild yam *Dioscorea* in Veracruz, Mexico (later called the Marker degradation) created the research foundation for Syntex's founding (1944). Djerassi moved from the University of Wisconsin to Syntex in 1949 and led a steroid molecular-modification program. Miramontes recorded the first synthesis in his lab notebook on October 15, 1951 (now held by UNAM), consistent with the patent's priority date of 1951-11-22. In 1957 the FDA approved norethindrone for menstrual disorders; in 1960 Searle's norethynodrel (Frank Colton 1953 synthesis, Enovid brand) received FDA approval as an oral contraceptive, with Ortho-Novum (norethindrone-based) following. This article excavates: (1) Claim 1 verbatim from US2744122A as obtained on Google Patents, (2) the historical path from Marker degradation through the founding of Syntex SA and the Mexican steroid industry, (3) consistency between Miramontes's lab notebook of 1951-10-15 and the priority date of 1951-11-22, (4) the Djerassi vs. Miramontes credit problem (the patent cleanly lists three co-inventors, but the popular narrative tilts toward Djerassi), (5) parallel running of Searle's Frank Colton norethynodrel patent US2725389A, and (6) the resonance with modern low-dose pills, hormonal IUDs, ART, and hormone replacement therapy. After the consecutive DB error corrections in Days 8/9/10/11, this is the **rare case in which the DB description (Djerassi/Miramontes/Rosenkranz three co-inventors, Syntex Mexico City, 1951 synthesis, 1956 grant) matches the primary source**—the second 'DB-agreement-confirmed' case after Day 11's aspirin US644077A.

On August 1, 1898, German chemist Felix Hoffmann of Farbenfabriken vorm. Friedr. Bayer & Co. (later Bayer AG) filed and the U.S. granted on February 27, 1900, US Patent US644077A 'Acetyl salicylic acid.' As the title indicates, this patent enclosed acetyl salicylic acid (the chemical name of aspirin) **as a substance itself** as a 'new article of manufacture,' a substance patent. Claim 1 defines acetyl salicylic acid by physicochemical properties: white needle crystals from chloroform recrystallization, soluble in benzene/alcohol, slightly soluble in cold water, melting point ~135°C, hydrolysis to acetic acid and salicylic acid in hot water. The specification records that Hoffmann established a synthesis using acetic anhydride (higher purity than the existing acetyl chloride method) in 1897. Germany's patent office rejected the patent for 'lack of novelty,' while the U.S. and U.K. granted substance patents. Read this as the precursor to the inventor dispute (in 1949, former colleague Arthur Eichengrün claimed 'I directed the actual synthesis; Hoffmann only executed the experiment'), the post-WWI U.S. seizure of Bayer's American assets (December 1918, sold to Sterling Products Co.; the U.S. 'Bayer' and 'Aspirin' trademarks remained on the U.S. side until 1994), the 1918 Spanish flu pandemic over-dosing records, and the 120-year reckoning to modern 100mg antiplatelet therapy (cardiovascular prevention from the 1980s onward) and OTC market expansion (after the 1917 U.S. patent expiration). Unlike Day 11's PH-007 / PH-005 DB-number/inventor errors, this is the rare case where the DB description (Felix Hoffmann sole inventor, Bayer assignee) matches the primary source.

From the early 1960s, Imperial Chemical Industries (ICI) systematically explored β-adrenergic receptor blockers under James W. Black's research direction. After the first clinical candidate pronethalol (Alderlin, approved 1962, withdrawn 1963 for carcinogenicity), they reached propranolol (Inderal brand, UK approval 1965, US approval 1967). Propranolol's core patent is **US3337628A '3-naphthyloxy-2-hydroxypropylamines'** (priority 1962-11-23, filed 1963-11-12, granted 1967-08-22), inventors **Albert Frederick Crowther and Leslie Harold Smith** (assignee ICI). The DB-registered US3408387A 'Amidoaroxyalkanolamines' (priority 1964-09-30, filed 1965-09-17, granted 1968-10-29) was invented by **Ralph Howe and Leslie Harold Smith** and covers a **derivative series** within ICI's β-blocker exploration program. The DB row PH-005 lists 'Inventor: James W. Black,' but Black himself is **not named** on either patent — Black directed research as ICI's Research Director, while patent-filing work was handled by the chemists Crowther / Howe / Smith in a divided-labor structure. Black received the 1988 Nobel Prize in Physiology or Medicine ('important principles for drug treatment,' jointly evaluated with cimetidine H2 blocker work). Inventor-field corrections have occurred for four consecutive days (Day 8/9/10/11).

Squibb's Miguel A. Ondetti and David W. Cushman pursued angiotensin-converting enzyme (ACE) inhibitor molecular design from the early 1970s. The starting point was the bradykinin potentiating factor (BPF) peptides isolated by Brazilian São Paulo pharmacologist Sérgio H. Ferreira from the venom of Bothrops jararaca (jararaca pit viper) between 1965 and 1970. Squibb adopted a strategy to design orally active small-molecule ACE inhibitors using BPF peptides as templates, executing rational design in three stages: (1) Byers & Wolfenden's 1973 carboxypeptidase A model (zinc enzyme, substrate-structure analogy), (2) peptidomimetic design, and (3) sulfhydryl group (-SH) zinc coordination — arriving at SQ 14225 (later named captopril). The candidates DB row PH-007 lists 'Captopril patent US4046889A' with 'inventors Ondetti/Rubin/Cushman three co-inventors,' but the primary sources confirmed on Google Patents show: (a) US4046889A is 'Azetidine-2-carboxylic acid derivatives' with two inventors (Ondetti / Cushman), and (b) captopril's chemical backbone (proline + 3-mercaptopropanoyl linkage) is covered by US4105776A 'Proline derivatives and related compounds' (priority 1976-06-21, filed 1976-12-22, granted 1978-08-08, Ondetti/Cushman, Squibb). DB corrections have occurred for four consecutive days (Day 8/9/10/11). Read this as the 50-year reckoning between 1981 captopril FDA approval, 1985 Merck enalapril (second-generation ACE inhibitor), and modern AI-driven molecular design (AlphaFold / RoseTTAFold / structure-based drug design).

U.S. filing October 8, 1945; granted January 24, 1950; expired January 24, 1967. Raytheon Manufacturing Company assignee, Percy L. Spencer **sole inventor** for the foundational microwave-oven patent US2495429A 'Method of Treating Foodstuffs.' Claim 1 defines a method that generates electromagnetic wave energy in the microwave region, concentrates and guides it within a restricted region of space, and exposes the foodstuff to the energy long enough to cook it to a predetermined degree. Spencer's biographical account: while standing near a magnetron during radar work, the chocolate bar in his pocket melted — that was the trigger. Spencer received only **a $2 lump-sum bonus** from Raytheon for this invention (long-standing biographical anecdote, primary source unverified). The 1947 first commercial unit Radarange weighed about 340 kg, cost about $5,000, and required cooling plumbing — strictly an industrial appliance. The first home model came from Tappan in 1955 at $1,295. Amana's $495 Radarange Counter Top in 1967 accelerated household adoption. Read this as the precursor to convenience-store bento reheating, frozen-food thawing, microwave recipe books, and the 1-minute instant-food culture. Title, full Claim 1, sole inventor, filing/grant dates, original/current assignees retrieved from Google Patents.

U.S. priority/filing January 12, 1923; granted October 9, 1923. Inventors: Frederick G. Banting, Charles H. Best, **and James B. Collip** as three co-inventors (the DB record listing only Banting/Best is incorrect). Original Assignee: University of Toronto **and** University of Alberta as joint assignees (the DB record listing UoT alone is also incorrect — Collip's share went to Alberta to reflect his university affiliation). Title: 'Extract obtainable from the mammalian pancreas or from the related glands in fishes, useful in the treatment of diabetes mellitus, and a method of preparing it.' Claim 1 defines a substance prepared from fresh pancreatic or related glands containing in concentrated form the extractive from the ductless portion of the gland, suitably free from injurious substances for repeated administration, with the physiological effect of reducing blood sugar useful for treating diabetes mellitus. Read this as the precursor to the 100-year reckoning between the legendary '$1 transfer' and the modern insulin pricing problem (US ~$300/vial vs Canada ~$30/vial, driven by recombinant analog insulins and PBM distribution structures, not by the 1923 patent). Title, full Claim 1, three inventors, filing/grant dates, and original/current assignees retrieved from Google Patents.

On August 7, 1975, Georges J. F. Köhler and César Milstein published 'Continuous cultures of fused cells secreting antibody of predefined specificity' in Nature (volume 256, issue 5517, pages 495-497), demonstrating monoclonal antibody production via continuous culture using the hybridoma method. Affiliation: **MRC Laboratory of Molecular Biology, Cambridge** — a Medical Research Council-operated research institute that is *not* part of Cambridge University (the DB record listing 'Cambridge University' is structurally inaccurate). Milstein wrote to the MRC proposing patenting, and the MRC replied: 'it is not immediately obvious what patentable features are at present disclosed in the Nature paper.' The UK NRDC (National Research Development Corporation) did not pursue patenting. On April 28, 1978, Hilary Koprowski and Carlo M. Croce of the Wistar Institute of Anatomy and Biology in the U.S. filed the **application patent US4172124A 'Method of producing tumor antibodies,'** granted October 23, 1979. The core invention (Nature paper) remained in the public domain; the application patent (tumor antibody production) was enclosed in the U.S. The 1980s UK industrial-policy debate, the 1984 Nobel Prize in Physiology or Medicine awarded to Jerne / Köhler / Milstein, and the 50-year reckoning with today's monoclonal antibody pharmaceutical market (over $200 billion annually as of 2024 — Herceptin, Keytruda, Humira, etc.).

U.S. priority July 13, 1926; filed June 18, 1927; granted August 12, 1930. Frosted Foods Co Inc obtained the quick-freeze food packaging core patent US1773079A 'Method of preparing food products,' invented by **Clarence Birdseye alone**. Claim 1 defines a packaging-and-freezing process: pack food into the container in which it will be marketed, then freeze it under pressure applied to substantial surface areas of the packed container. In 1929, General Foods Corporation reportedly acquired Frosted Foods Co and Clarence Birdseye–related patents for $22 million (primary source unverified — consensus in later business-history books). Read this as the precursor to grocery-aisle frozen food, frozen dumplings, frozen pizza, frozen bakery, IQF (Individual Quick Freezing) seafood, and Uber Eats frozen delivery. Title, full Claim 1 text, sole inventor, filing/priority/grant dates, original and current assignees retrieved from Google Patents. **The DB record (candidates.tsv) listing '1927 patent' actually refers to the filing year; priority is 1926-07-13, and Birdseye-related patents number 20+.**

Japanese priority June 7, 1974; U.S. filed December 4, 1975; granted September 20, 1977. Akira Endo, Masao Kuroda, Akira Terahara, Yoshio Tsujita, and Chihiro Tamura at Sankyo Co Ltd **jointly** invented the statin core patent US4049495A 'Physiologically active substances and fermentative process for producing the same.' The patent describes a fermentation process to obtain HMG-CoA reductase inhibitors ML-236A, ML-236B (later compactin / mevastatin), and ML-236C from the genus *Penicillium*. Verbatim Claim 1 text, title, five inventors, filing/priority/grant dates, and assignee retrieved from Google Patents. **The DB record (candidates.tsv) listing patent number US4231938 is wrong.** US4231938A is Merck's Lovastatin patent (MSD803, Monaghan/Alberts/Hoffman/Albers-Schonberg, filed 1979, granted 1980) — a different compound. The DB entries 'sole inventor,' 'discovered 1973, filed 1976' are also wrong. Read this as the precursor to the statin class (Mevacor / Pravachol / Zocor / Lipitor / Crestor) and modern coronary-disease prevention.

Filed in the U.S. on March 28, 1985; granted July 28, 1987. Norman Arnheim, Henry A. Erlich, Glenn T. Horn, Kary B. Mullis, Randall K. Saiki, and Stephen J. Scharf at Cetus Corporation **jointly** invented the PCR (Polymerase Chain Reaction) core patent US4683195A 'Process for amplifying, detecting, and/or-cloning nucleic acid sequences.' Mullis's standalone core-idea patent is the separate US4683202A; the Taq DNA polymerase implementation is US4965188A — three patents form one family. Roche acquired the Cetus PCR portfolio in 1991 (publicly reported around $300M initially, with subsequent licensing earning billions). Read this as the precursor to COVID-19 PCR testing, qPCR, digital PCR, CRISPR diagnostics (SHERLOCK / DETECTR), forensics, prenatal screening, and ancient DNA analysis. Title, six inventors, filing/grant dates, and assignee retrieved from Google Patents and WebSearch sources. Full text of Claim 1 and Description not yet verified verbatim. **The DB record (candidates.tsv) attributing this invention to 'Mullis himself' is inaccurate** — Mullis is one of six co-inventors; his sole-inventor patent is the separate US4683202A.

Filed in the U.S. on May 21, 1970, granted on January 23, 1973, expired on January 23, 1990 (Lifetime). Mario W. Cardullo and William L. Parks III, working at Communications Services Corporation (Rockville, Maryland), jointly invented the foundational RFID patent US3713148A 'Transponder Apparatus and System.' Claim 1 defines a 'self-contained transponder with writable memory powered by the interrogation signal itself.' Claim 4 extends the carrier wave to light frequency, Claim 5 to acoustic frequency. The Description explicitly cites 'automatic highway toll device for motor vehicles' as a sample application — the conceptual ancestor of ETC, written 31 years before Japan's ETC launched in 2001. 185 forward citations. Read this as the precursor to apartment key fobs, employee badges (HID Prox), Walmart inventory tags, pet ID chips, and the NFC tag inside an Apple TechWoven Case. Title, full Claim 1, inventors, filing/grant/expiration dates, Abstract, and forward citation count retrieved from Google Patents. **The DB record (candidates.tsv) attributing this to 'Charles Walton' is wrong** — Walton was a separate RFID pioneer; he is not an inventor on this patent.

Filed in the U.S. on May 21, 1970; granted January 23, 1973. Communications Services Corporation obtained the RFID (radio frequency identification) ancestor patent US3713148A, 'Transponder apparatus and system.' Inventors are Mario W. Cardullo and William L. Parks III — **two co-inventors** (the 'Charles Walton' entry in the local DB belongs to a separate patent line and is not on this patent). Claim 1 describes a transponder system with 'changeable memory,' 'self-powering from the interrogation signal,' and 'read/write control.' More than 50 years after grant in 1973, read this as the precursor to Suica, PASMO, NFC payment, electronic passports, retail anti-theft tags, and livestock microchips. Basic info, Claim 1 summary, inventors, filing/grant dates, and Abstract retrieved from Google Patents. Full description and comparison to Charles Walton's patents not yet checked.

Filed in the U.S. on October 17, 1986; granted February 13, 1990. Qualcomm Inc obtained the CDMA core patent US4901307A, 'Spread spectrum multiple access communication system using satellite or terrestrial repeaters.' Inventors are Klein S. Gilhousen, Irwin M. Jacobs, and Lindsay A. Weaver, Jr. — **three co-inventors** (Andrew Viterbi, listed in the local DB, is not on this patent). Claim 1 defines a 'spread spectrum multiple access communication system through satellite or terrestrial repeaters' and lays the design skeleton that connects to IS-95 (cdmaOne) → CDMA2000 → 3G WCDMA. Read this as the precursor to 3G/4G/5G. Basic info, Claim 1 summary, inventors, filing/grant dates, and Abstract retrieved from Google Patents. Full description and Viterbi-related patents not yet checked.

Filed in the U.S. on September 17, 1997; granted July 8, 2003; expired August 15, 2018. Jaap Haartsen (Jacobus Cornelis Haartsen) of Ericsson is registered as the **sole inventor** on the Bluetooth core patent US6590928B1, 'Frequency hopping piconets in an uncoordinated wireless multi-user system.' Claim 1 defines a wireless network of master and slave units, in which a virtual frequency hopping channel is derived from the master address and master clock. Multiple piconets coexist without synchronizing with each other (uncoordinated). Read this as the precursor to AirPods, Magic Mouse, hearing aids, Apple Watch, fitness bands, and smart locks. Title, full Claim 1, inventor, filing date, priority date, grant date, abstract, and current assignee retrieved from Google Patents.

Filed August 1995 by Netscape Communications, US5657390A 'Secure socket layer application program apparatus and method' was invented by Taher Elgamal and Kipp E. B. Hickman. It describes a socket API that places a security protocol between the application layer and the transport layer. The body uses 'Secure Sockets Layer,' 'SSL library,' and 'SSL protocol' multiple times. Expired in the U.S. (Lifetime). Current Assignee is Meta Platforms Inc — a curious twist of corporate history. Read this as the precursor to HTTPS, TLS 1.3, HSTS, and QUIC/HTTP3. Title, Abstract, Claim 1 outline, inventors, dates, and Legal Status retrieved from Google Patents; full text and the chain of correspondences with later TLS RFCs not yet read.

Filed October 1986, US4698672A 'Coding system for reducing redundancy' was issued to Compression Labs (CLI), invented by Wen-Hsiung Chen and Daniel J. Klenke. The text says it is 'particularly useful in video compression systems' and never once uses the word 'JPEG.' Yet in 2002, Forgent Networks — which had inherited the patent through corporate transfers — claimed JPEG images infringed it and demanded license fees from companies across the web. Expired in the U.S. on October 27, 2006. Read this as a precursor to WebP, AVIF, HEIC, and to the rise of patent-pool governance. Title, Abstract, Claim 1 outline, inventors, dates, and Legal Status retrieved from Google Patents; full text and Forgent litigation primary sources not yet read.

Filed in the U.S. on March 14, 1995, with priority from Japanese application JP4258794A (March 14, 1994), Toyota Central R&D Labs and NipponDenso jointly received US5726435A 'Optically readable two-dimensional code and method and apparatus using the same.' Three position-detecting symbols (finder patterns) are placed so that any scan line passing through a symbol's center returns the same black:white:black:white:black = 1:1:3:1:1 ratio, regardless of direction. Expired in the U.S. on March 14, 2015. Not Hara's solo invention — five co-inventors. The decisive design choice was Denso Wave's later policy: hold the patent but make the spec royalty-free. Read this as the precursor to PayPay, airline boarding QRs, and vaccination certificates. Title, Claim 1, inventors, filing/priority dates, and Abstract retrieved from Google Patents.

Filed November 1993, US5487069 'Wireless LAN' by O'Sullivan and four co-inventors at Australia's CSIRO describes multi-subchannel modulation (effectively OFDM) for practical data transmission in indoor multipath environments at frequencies above 10 GHz. Later asserted as a standard-essential patent on IEEE 802.11a/g/n/ac, drawing more than $430M in cumulative settlements from 14 Wi-Fi equipment makers worldwide. Title, gist of Claim 1, inventors, filing date, and Legal Status retrieved from Google Patents. Full Description and litigation primary sources not yet read.

Filed September 1977, US4200770A 'Cryptographic apparatus and method' by Hellman, Diffie, and Merkle at Stanford. Built on the computational hardness of the discrete logarithm problem, it patented a way for two parties to derive a shared secret key over an insecure channel without any prior shared secret. Expired in the U.S. in 1997. The ancestor of TLS/HTTPS, Signal, WireGuard, and SSH handshakes. Title, Claim 1, and the Y₁^X₂ mod q = Y₂^X₁ mod q description retrieved from Google Patents. Full Description text and related litigation primary sources not yet read.

Filed January 1995 in the U.S. with priority from German DE3912605 (April 1989), Fraunhofer's US5579430 'Digital encoding process' captures the heart of MP3: spectral transform, variable-precision quantization, and a variable-length code where 'the more frequent the spectral coefficient, the shorter the code word.' Not the single MP3 patent — one core piece of ISO/IEC 11172-3 Layer 3. U.S. expired 2013; Fraunhofer's licensing program ended in 2017. A precursor to Spotify, Apple Music, Podcasts, and Discord audio. Title, Claim 1, inventors, filing/priority dates, and Abstract retrieved from Google Patents.

Filed March 1975, US4063220A by Metcalfe, Boggs, Thacker, and Lampson at Xerox PARC describes 3 Mbps packet communication over coaxial cable using CSMA/CD (carrier sensing + collision detection + randomized backoff). A precursor to modern Ethernet, Wi-Fi, data center networks, and cloud infrastructure. Primary source URL confirmed; Claim 1, CSMA/CD details, and 3 Mbps speed retrieved; full text not yet read.

Filed October 1995, US5774670A by Lou Montulli at Netscape describes 'persistent client state' — what we now call cookies — as the answer to HTTP statelessness. The patent specifies the Set-Cookie header syntax, domain/path matching, and the secure flag. Later assigned to Meta Platforms via Netscape's asset acquisition. A reading aid for understanding modern Cookie regulations, third-party Cookie deprecation, and the design difference vs. JWT/OAuth. Primary source URL confirmed; Claim 1 retrieved; full text not yet read.

Filed October 1949, US2612994A by Woodland and Silver describes a 'classifying apparatus' that identifies articles using concentric circular light-reflective lines and photoelectric scanning. Not linear, not laser-based — the original barcode was circular and read by a photocell. A precursor to modern UPC, QR codes, RFID, and supply-chain identification. Expired 1969. Title, Claim 1, inventors, and filing data retrieved from Google Patents.

Filed June 1983, US4558302A by Terry A. Welch describes an adaptive dictionary compression algorithm (LZW) implemented with high-speed hashing. CompuServe adopted it in the GIF format in 1987. In 1994 Unisys began demanding license fees, triggering the GIF/PNG controversy. GIF was freed when the US patent expired in 2003. Primary source URL confirmed; full text not yet read.

Filed December 1977, US4405829A by Rivest, Shamir, and Adleman encodes public-key cryptography based on the computational difficulty of factoring the product of two primes. The mathematical foundation of TLS, HTTPS, and digital signatures. Expired September 2000. Primary source URL confirmed; full text not yet read.

Filed September 1997, US5960411A by Bezos and three co-inventors describes a method for completing a purchase order with a single user action, drawing on pre-stored purchaser information. 1,630 forward citations. A prior example of the 'zero-friction payment' design intent shared by Apple Pay, Shop Pay, and PayPay. Expired September 2017. Full text retrieved from Google Patents.

Filed in 1989, US5197005A describes a rule-based system for querying databases using natural language — a knowledge base plus expert system approach to the problem that LLMs now solve with probabilistic generation. The question is the same; the method is fundamentally different. Primary source URL confirmed; full text not yet read.

Filed in 1991, US5253329A by NASA researchers Villarreal and Shelton replaces scalar synaptic weights with adaptive digital filters, encoding temporal dependencies directly into network connections. A prior example of the same problem intent that later led to LSTMs and Transformers. Primary source URL confirmed; full text not yet read.

Filed in 1993, US5517598A describes a hardware architecture where the same processor structure handles both forward inference and error backpropagation, using transpose matrices to route error signals. The problem it posed — running learning on dedicated hardware — overlaps in intent with modern GPU and TPU design, though the underlying architecture is fundamentally different. Full text retrieved from Google Patents.

Filed 1992 by LeCun and Wu at AT&T Bell Labs, US5337372A describes a symbol recognition system that uses low-resolution feature arrays to narrow candidates, then confirms matches at higher resolution — reducing computation. Patent URL confirmed; full text unread.

Filed 1994 by LeCun, Denker, Simard, and Victorri at Lucent Technologies, US5572628A describes a method for making neural networks invariant to transformations like rotation, scaling, and translation — using tangent vectors instead of data augmentation. Patent URL and abstract confirmed; full text unread.

Filed November 1989 by John S. Denker, Richard E. Howard, Lawrence D. Jackel, and Yann LeCun at AT&T Bell Laboratories, US5067164A describes a layered network for optical character recognition using weight sharing: 90,000 connections represented with approximately 2,600 free parameters (97% reduction). Google Patents full text retrieved. Comparing the design problem to modern CNNs 36 years later.

Filed 1991 by Peter F. Brown, John Cocke, Frederick Jelinek, and colleagues at IBM Research, US5477451A describes learning translation probabilities from parallel corpora. Patent URL confirmed; full text unread. This is a record of what the design was trying to do — not a claim that it leads to LLM translation.

Filed 1997 by Lawrence Page, US6285999B1 describes a method for ranking nodes in a linked database by computing how often a random link-follower would land on each page. Patent URL confirmed; full text unread. Exploring connections to LLM search and graph-based RAG.

Filed 1998 by Gregory Linden, Jennifer Jacobi, and Eric Benson, US6266649B1 describes a recommendation system that precomputes item-to-item similarity tables offline and serves personalized recommendations at low latency by looking up and combining those tables in real time. Expired 2018. Distinct from embedding-based recommendation but shares the same design problem.

I read expired patent US6562355B1, filed by a single L'Oréal researcher in 1999 and expired in 2020. It describes six formulation examples — including an 'Eye Bag-Masking Gel' and a concealing emulsion made with La Roche-Posay spring water — built around the synergistic combination of dextran sulfate and escin. By 2026, both ingredients appear in commercial eye care. A 2018 peer-reviewed study confirmed the underlying mechanism.

In 1878, a chemist came home from his Baltimore lab without washing his hands, bit into his dinner bread, and noticed it tasted sweet. That accident became saccharin — and then sparked a 145-year battle through presidential corridors, FDA hearing rooms, and a congressional moratorium that one million citizens forced into existence. I read the primary documents.

Pharma Archaeology #1. I had Claude read the official press release from Daiichi Sankyo Healthcare, the Yakuji Nippo trade-press articles, and the related minutes of MHLW's Pharmaceutical Affairs and Food Sanitation Council, and reconstructed the 25-year gap between the prescription Loxonin (Sankyo, 1986) and the over-the-counter Loxonin S (2011). What does 'Switch-OTC' actually mean, why is the OTC dose 68.1 mg per tablet, and why is the same active ingredient subject to a different age limit on the OTC side? Read straight from the contents of a Japanese household medicine cabinet.

Kitchen Health Archaeology #1. In 1907, Kikunae Ikeda — professor at the College of Science, Tokyo Imperial University — isolated 'umami' from kombu broth. I made Claude read his 1909 paper in the Journal of the Chemical Society of Japan (Vol. 30, pp. 820–836) alongside Japanese Patent No. 14805 filed on April 24, 1908. From there, the long reconciliation: 1908 patent → 1968 'Chinese Restaurant Syndrome' controversy → 2002 Nature paper on the TAS1R1/R3 receptors. A 94-year gap before the global scientific community formally accepted the fifth basic taste. Reread the origin of the seasoning bottle in your fridge through one Meiji-era patent.

In 1956 the NSA launched Project Lightning, pouring millions into IBM, Sperry-Rand, RCA, Philco, and GE to build a 1-gigahertz logic circuit and the next-generation Harvest computer for Cold War cryptanalysis. The cryotron failed; Harvest was retired in 1976. But the machine they aimed at — running at GHz, a hundred times faster than commercial systems — was implemented 67 years later as the NVIDIA H100. I re-read NSA Project Lightning in the LLM era.

Patent Archaeology #2. I made Claude read the 4-page specification + 4 sheets of drawings of US381968, the AC motor patent Nikola Tesla filed in October 1887. The rotating magnetic field, the commutator-less design, the polyphase alternating current — every core idea of today's EV drivetrain is already there. Tesla's logo is the copper rotor, the company name is in tribute to Nikola Tesla, and yet from Model 3 onwards Tesla shifted its main rear motor to the permanent-magnet IPM-SynRM. I had the LLM translate that distance too.

The final post of the series. Every prompt I actually used in posts 1-5 (candidate selection, content extraction, modern translation, grading, pitfall checks), the complete pipeline diagram, the tool stack, and the reproduction steps — all collapsed into one article. The goal: by the end, you can start mining forgotten long-form documents in your own field today.

When you use an LLM to mine forgotten long-form documents, you will almost certainly hit three traps: fabrication (fake citations), cost explosion (503s, model deprecations, token blowups), misreading (language barriers, term-of-art confusion). This post discloses every real incident from posts 1–5 of this series, with the prompts and operating rules that prevent each one.

In 1966 the US National Academies' ALPAC (Automatic Language Processing Advisory Committee) issued a report — Language and Machines — that cut off machine-translation funding and triggered the first AI winter. Chair: John R. Pierce (Bell Labs). Sixty years later, with LLMs effectively solving machine translation, I grade which of ALPAC's claims were right and which were wrong.

Token Ring, standardized as IEEE 802.5 in 1989, was destroyed by Ethernet's probabilistic access and effectively retired. But its core idea — deterministic access control — has fully returned, under different names: InfiniBand, NVLink, RDMA, Ultra Ethernet. Token Ring didn't lose. It was just 30 years too early.

Samsung's official IR has no Annual Report from before 2008. SEC EDGAR returns 403. TSMC IR also returns 403. The Wayback Machine blocks the fetch. With every primary source on the other side of a wall, I dig out — via Wikipedia — the fact that Samsung built the world's first 1Gb DRAM in 1996, and re-evaluate it as the direct ancestor of today's HBM4.

Patent Archaeology #1. I dug a 1990s neural network chip patent out of Google Patents. IBM and Guy Paillet's ZISC (Zero Instruction Set Computer): 36 independent neurons, Manhattan distance, daisy chain. The design philosophy and problem framing overlap with modern NPU/TPU in ways that are striking.

A 3M-view tweet by @gippp69 revealed something bigger than Amazon arbitrage: a meta-method for using LLMs to mine documents nobody reads anymore. Patents, IR archives, decommissioned standards, declassified reports — Claude can read them in a single night. Humans cannot. This article opens a 7-part series on doing exactly that.