1995 Advanced Polymer Systems Microsponge retinol patent US5851538 and 1987 Imperial Chemical Industries Zoladex peptide sustained-release patent US4767628 — both registered in cosmetic DB as 'CS-007 niacinamide whitening' and 'CS-008 kojic acid whitening' turned out to be neither niacinamide nor kojic acid nor whitening agents, but pharmaceutical drug delivery patents (excavation tale #4)

Conclusion first

Two entries registered in the cosmetic subseries DB (candidates.tsv) — CS-007 'Niacinamide (Vitamin B3) whitening patent P&G US5851538' and CS-008 'Kojic acid (kojic acid) whitening patent Ajinomoto US4767628' — were retrieved from Google Patents on Day 21. Both patents turned out to be completely unrelated to the DB claims, and both turned out to be pharmaceutical drug-delivery patents.

• CS-007 US5851538 is Advanced Polymer Systems Inc's (APS, the parent of Microsponge®) low-irritation topical formulation of retinol in porous microspheres, filed December 1995, granted December 22, 1998, by co-inventors Michael Froix / Masha Pukshansky / Sergio Nacht. This is the lineage of Ortho-McNeil's Retin-A Micro 0.1% gel (FDA approval 1997) core delivery patents.
• CS-008 US4767628 is Imperial Chemical Industries Ltd's (ICI, later Zeneca, now AstraZeneca) two-phase sustained-release composition of polylactide (PLA) / poly(lactide-co-glycolide) (PLGA) with uniformly dispersed polypeptide drug, filed July 1987, granted August 30, 1988, by sole inventor Francis G. Hutchinson. This is the lineage of AstraZeneca's Zoladex® (goserelin) — an LH-RH agonist 1-month / 3-month sustained-release implant for prostate and breast cancer.

In other words, "the cosmetic subseries DB contains two pharmaceutical drug-delivery patents that are neither cosmetic nor whitening nor niacinamide nor kojic acid, each being the core patent of a separate pharmaceutical program." This is a different class of structural error from the Day 20 ep74-75 "cosmetic DB contains a non-cosmetic patent" swaps. Day 21's two cases reveal a deeper pattern: the boundary between cosmetic DBs and pharmaceutical DBs gets crossed when delivery technology (porous microspheres, PLA-PLGA sustained release) is shared across both fields.

This note (1) cross-checks Claim 1 verbatim with primary sources for the CS-007 and CS-008 mismatches, (2) translates the realities of the two patents (APS Microsponge retinol formulation and ICI Zoladex-line PLA-PLGA peptide sustained release) into cosmetic context, (3) attempts a structural explanation of why pharmaceutical delivery patents got mixed into the cosmetic DB, and (4) lays out Day 22+ excavation steps for the correct niacinamide and kojic acid origin patents. Day 21 updates DB corrections #36 through #47 consecutively (twelve) in this note (CS-007 six items and CS-008 six items); cumulative cosmetic subseries DB corrections across Days 20-21 reach 11 cases / 47 label discrepancies (ep73 CS-006 5 / ep74 CS-004 1 / ep75 CS-005 1 / this note CS-007 6 + CS-008 6).

What happened

By the end of Day 20, candidates.tsv held three cosmetic Phase 1 entries verified in primary sources (CS-001 / CS-003 / CS-006, marked Published), one Source Found (CS-002, number obtained but Claim 1 unread), and two 'Source Not Confirmed' entries with consecutive number-swap discoveries (CS-005 → P&G transdermal penetration enhancer US3839566 in Day 20 ep75; CS-004 → ambiguous DRP number for Nivea / Eucerit lineage in Day 20 ep74). The remaining 'Source Not Confirmed' entries were CS-007 (niacinamide US5851538) and CS-008 (kojic acid US4767628) — the "three unverified items" handed off into Day 21 (CS-007, CS-008, plus CS-002 Claim 1).

Day 21 morning, I first checked the validity of the CS-007 and CS-008 patent numbers via Google Patents direct retrieval.

curl -sL https://patents.google.com/patent/US5851538 → meta DC.title
  → 'Retinoid formulations in porous microspheres for reduced irritation and enhanced stability'
curl -sL https://patents.google.com/patent/US4767628 → meta DC.title
  → 'Continuous release pharmaceutical compositions'

The DB claims were 'niacinamide whitening' and 'kojic acid whitening.' The primary-source titles are 'retinoid formulations in porous microspheres' and 'continuous release pharmaceutical compositions.' At this point, both entries were confirmed to be entirely mismatched with their DB claims. Subsequent extraction of DC.contributor / inventor itemprop / assigneeOriginal showed CS-007's assignee as APS (Advanced Polymer Systems Inc) and CS-008's as ICI (Imperial Chemical Industries Ltd) — not the DB's 'P&G' or 'Ajinomoto.'

The structure of this note was set: as an extension of Day 20, document the consecutive CS-007 / CS-008 swaps as an excavation note, translate the realities into the cosmetic context, and structure the boundary mixing between cosmetic DBs and pharmaceutical delivery patents.

Six DB corrections for CS-007 US5851538

The following six points are verbatim mismatches between the DB entry and primary source.

1. The patent number itself is not wrong (US5851538 exists), but this number is not a P&G niacinamide whitening patent. The number-to-content binding is wrong (Day 21 cumulative #36).
2. Assignee: DB 'P&G (Procter & Gamble)' → reality 'Advanced Polymer Systems Inc' (#37).
3. Inventors: DB blank / inferred 'P&G researchers' → reality 'Michael Froix / Masha Pukshansky / Sergio Nacht, three co-inventors, Froix first' (#38).
4. Ingredient: DB 'Niacinamide (Vitamin B3)' → Claim 1 verbatim 'a retinoid composition comprising retinol' (#39). Niacinamide is not mentioned in any of the 16 claims.
5. Use: DB 'Whitening (melanosome transfer suppression)' → Claim 1 verbatim 'topical administration of retinol'; specification states 'reduced irritation and enhanced stability' as goals (#40). No mention of whitening.
6. Claim 1 subject: DB 'apply 2-5% niacinamide topically to inhibit melanosome transfer' → Claim 1 verbatim 'a stable pharmaceutical composition for topical administration of retinol consisting of an oil-in-water emulsion comprising suspended solid water-insoluble microscopic particles containing a substantially continuous network of pores open to the exterior, and a retinoid composition comprising retinol residing only in the pores' — the subject is 'delivery form,' not 'ingredient' (#41).

These six points are read in detail with Claim 1 verbatim translated into cosmetic context in ep77 (cosmetic-patent-memo-05).

Six DB corrections for CS-008 US4767628

Same structure as CS-007: six verbatim mismatches between DB entry and primary source.

1. The patent number itself is not wrong (US4767628 exists), but this number is not an Ajinomoto kojic acid whitening patent (Day 21 cumulative #42).
2. Assignee: DB 'Ajinomoto Co., Inc.' → reality 'Imperial Chemical Industries Ltd (ICI, later Zeneca, now AstraZeneca)' (#43).
3. Inventor: DB blank / inferred 'Ajinomoto researchers' → reality 'Francis G. Hutchinson, sole inventor' (#44).
4. Ingredient: DB 'Kojic acid' → Claim 1 verbatim 'polylactide which is a polymer of lactic acid alone, a copolymer of lactic acid and glycolic acid, and a pharmacologically active polypeptide' (#45). Kojic acid is not mentioned in any of the 17 claims.
5. Use: DB 'Whitening (tyrosinase inhibition)' → Claim 1 verbatim 'continuous release pharmaceutical composition'; specification states 'a pharmaceutical composition placed in an aqueous physiological-type environment' — i.e., in vivo sustained-release injection / implant use (#46). No mention of whitening or topical application.
6. Claim 1 subject: DB 'apply 1-3% kojic acid topically to inhibit tyrosinase and suppress pigmentation' → Claim 1 verbatim '50% to 99.999% polylactide (PLA / PLGA) 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 (initial diffusion / late matrix degradation) in aqueous physiological environment' — subject is 'delivery form,' moreover 'in vivo sustained-release implant' (#47).

These six points are read in detail with Claim 1 verbatim translated into pharmaceutical delivery context in ep78 (cosmetic-patent-memo-06).

Reality reading #1 — APS Microsponge retinol (CS-007 US5851538)

CS-007 US5851538's assignee Advanced Polymer Systems Inc (APS) established the Microsponge® porous polymer microsphere technology in the 1980s and licensed it widely across cosmetics, OTC pharmaceuticals, and prescription topical formulations. Microsponge is designed as 'spherical particles 5-300 micrometers in diameter with a continuous internal pore network connected to the exterior, holding active ingredients in pores and releasing them gradually,' simultaneously achieving (a) prolonged retention with controlled release on the skin surface, (b) protection of unstable actives (especially against oxidation and photodegradation) within the pores, and (c) safe topical application of high-concentration irritating ingredients without skin irritation.

US5851538 is the patent applying this Microsponge technology to retinol (the pure form of vitamin A). Retinol is extremely unstable to light, oxygen, and heat, and at high concentrations (0.1-0.5%) causes skin irritation (erythema, peeling, dryness). Claim 1 verbatim requires 'a retinoid composition comprising retinol residing only in the pores' of 'solid water-insoluble microscopic particles containing a substantially continuous network of pores open to the exterior,' achieving 'lower irritancy than a non-(microscopic particle) formulation containing the same concentration.'

This patent is in the lineage of Retin-A Micro 0.1% gel (Ortho-McNeil, FDA approval 1997; the active is tretinoin, but it adopts the same Microsponge delivery), which significantly reduced irritation compared to the conventional Retin-A Cream 0.1% and is now widely prescribed for acne, photoaging, wrinkles, and pigmentation. In cosmetic context, retinol-formulated skincare products like Olay Total Effects, RoC Retin-Ox, and La Roche-Posay Redermic R adopt this kind of polymer porous carrier technology (Microsponge, Polytrap, microcapsule, etc., spanning APS and non-APS systems) to circumvent retinol's irritation and instability.

So CS-007 US5851538 is, in the cosmetic context, an extremely important patent as 'retinol delivery technology' — had the DB memo registered it as 'APS Microsponge retinol' rather than 'P&G niacinamide,' it would have stood as a legitimate excavation note theme. Because the DB label was misapplied as 'P&G niacinamide,' the meaning of this patent went unseen until Day 20. Detailed Claim 1 verbatim reading is split into ep77.

Reality reading #2 — ICI Zoladex-line peptide sustained release (CS-008 US4767628)

CS-008 US4767628's assignee Imperial Chemical Industries Ltd (ICI) is a UK chemical / pharmaceutical conglomerate; in 1993 the pharmaceutical division was spun off as Zeneca, and in 1999 merged with Sweden's Astra to form the current AstraZeneca. ICI Pharmaceuticals in 1987 was developing Zoladex® (goserelin acetate), a sustained-release implant of an LH-RH agonist (luteinizing hormone-releasing hormone agonist), administered subcutaneously to release continuously over 1 or 3 months for prostate cancer, breast cancer, and endometriosis. Zoladex was approved in the UK in 1989 and in the US in 1990.

US4767628 is the core delivery patent of this Zoladex-line implant. Claim 1 verbatim requires '**50% to 99.999% polylactide (PLA, polylactide) or lactic acid-glycolic acid copolymer (PLGA), 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 (first phase: initial diffusion through aqueous polypeptide domains communicating with the exterior surface; second phase: late release by matrix degradation),' achieving '1-3 month sustained blood peptide concentration from a single dose.' Sole inventor Francis G. Hutchinson, an ICI Pharmaceuticals researcher, claims the entire class of 'dispersion-type sustained-release implants' that disperse peptide drugs in PLA / PLGA bioabsorbable polymer matrices.

There is no direct cosmetic relevance. However, PLA / PLGA bioabsorbable polymers are the same family of materials used in modern medical aesthetics (PLLA filler as alternative to / complement of hyaluronic acid filler, brand name Sculptra). Sculptra's PLLA (poly-L-lactic acid) filler, when injected intradermally, gradually degrades while inducing collagen synthesis, functioning as a 'collagen biostimulator' approved by US FDA in 2009 and in Japan since 2014 for medical aesthetic use. Zoladex's release principle (PLA degradation + active release) and Sculptra's mechanism (PLLA degradation + collagen induction) differ only in whether the same biological reaction of PLA degradation is used for drug release or collagen induction.

So CS-008 US4767628 is, in the cosmetic context, not a direct origin but a distant ancestor of medical aesthetics' PLLA filler lineage. Had the DB memo registered it as 'ICI Zoladex / PLA-PLGA sustained release' rather than 'Ajinomoto kojic acid,' it might have been suitable as a medical aesthetics subseries theme. Detailed Claim 1 verbatim reading is split into ep78.

Why pharmaceutical delivery patents got mixed into the cosmetic DB

The Day 20 ep74-75 swaps and the Day 21 swaps in this note differ in error class.

Day 20 CS-005 and Day 21 CS-007 / CS-008 share the same swap structure, and all the realities are delivery-technology (carrier-form) patents. This is not a coincidence: it shows a structural problem where the boundary between cosmetic and pharmaceutical patents blurs because some numbers introduced in textbooks / industry articles as 'origin patents of a whitening ingredient' or 'origin patents of a sunscreen' are actually 'delivery patents of formulations containing that ingredient.'

Specifically:

• The origin patents of cosmetic ingredients (retinol, niacinamide, kojic acid, avobenzone) are often derived from 1970s-80s pharmaceutical / transdermal-penetration-enhancer patent groups.
• For whitening / anti-aging skincare, patents whose subject is 'delivery form' often have broader claims and higher market value than patents whose subject is 'ingredient,' so textbooks and industry articles often blur 'ingredient origin patents' and 'delivery origin patents.'
• Secondary sources (textbooks / industry articles) consulted when authoring candidates.tsv mixed up 'ingredient patents' and 'delivery patents,' and these errors flowed into the DB.

Day 22+ tasks build on this structural error pattern: re-excavate from primary sources (a) the correct P&G niacinamide whitening patent (candidate US5939082 'Methods of regulating skin appearance with vitamin B3 compound' etc.), (b) the correct kojic acid whitening patent likely from Sansho Pharmaceutical rather than Ajinomoto (candidate JP S58-118507 etc., 1980s Japanese patent).

Successors and derivatives — the lineage of cosmetic delivery technology

A second lineage emerged in Day 21: 'cosmetic delivery patents' is worth setting up as an independent subseries within AI Archaeology. Where cosmetic subseries Phase 1 covers 'origin patents of ingredients,' Phase 2 candidates would cover 'origin patents of cosmetic delivery technology' — namely:

1. APS Microsponge® (US5851538 etc.) — low-irritation formulations of retinol / tretinoin / benzoyl peroxide
2. Liposome / Niosome (L'Oréal Capture 1986, Lancôme Niosomes 1986; multiple patent groups) — delivery of lipophilic actives
3. Solid Lipid Nanoparticles (SLN) (Müller et al. 1990s patent groups) — carriers for UV filters and retinoids
4. PLA / PLGA filler (extension of ICI / Hutchinson lineage; Sculptra PLLA filler patents) — collagen biostimulators in medical aesthetics
5. Microencapsulation (Dragoco / IFF patent groups) — sustained release of fragrances and actives
6. Cyclodextrin inclusion complex (Sigma Aldrich / Wacker patent groups) — stabilization of unstable actives

Whether to spin these off as Phase 2 'cosmetic delivery patents' as an independent AI Archaeology subseries is to be discussed in Day 22+ via project_cosmetic_archaeology.md memory updates and consultation with Haruko. At minimum, this note's findings confirm twice in Days 20-21 that discussing cosmetic origin patents using only 'ingredient' framing without 'delivery form' produces continual divergence between Claim 1 verbatim and conventional narratives.

DB correction list (Days 20-21 cumulative)

The Phase 1 cosmetic subseries DB has accumulated 11 cases / 47 label discrepancies in Days 20-21.

This note brings the cumulative count to 47. In Days 20-21 alone, the cosmetic subseries DB has 11 cases × average 4.3 labels = 47 label discrepancies. Of the 8 cosmetic Phase 1 candidates in candidates.tsv, only 3 are verified (CS-001 / CS-003 / CS-006), while 4 have correction targets (CS-004 / CS-005 / CS-007 / CS-008) — confirming the cosmetic subseries DB's reliability is structurally below half.

Modern application — four lenses for reading cosmetic delivery

When excavating origin patents of cosmetic ingredients, the following four lenses must be distinguished, building on Day 20-21 findings.

1. Ingredient lens: patents on the discovery / synthesis / structure of the ingredient itself (e.g., niacinamide, kojic acid, retinol). Claim 1 is written as a molecular structure of the ingredient.
2. Use lens: patents on the use of the ingredient in cosmetics / whitening / anti-aging / UV protection. Claim 1 is written as 'a method of topical application of ingredient X to skin.' The Yu / Van Scott AHA patent (US4105782, ep73) is closest to this lens, but actually its Claim 1 covers 'ingredient + base reaction product' — not a pure use patent.
3. Delivery lens: patents on how to deliver the ingredient. Microsponge, Liposome, Cyclodextrin, PLA-PLGA, etc. — carrier and matrix design is Claim 1. CS-007 / CS-008 in this note belong to this lens.
4. Formulation lens: patents on how the ingredient combines with other ingredients. pH specifications, stabilizers, thickeners, preservatives — the combination is Claim 1. The Day 22+ re-excavation of P&G niacinamide will likely center on this lens.

The cosmetic subseries DB up to Phase 1 was operated using only the 'ingredient lens' and 'use lens.' Day 20-21's findings confirmed that the absence of the 'delivery lens' caused the CS-007 / CS-008 swaps.

Handoff to Day 22+

Day 21's note + ep77 + ep78 hands off the following three tasks to Day 22+.

1. Excavate the correct P&G niacinamide whitening patent: candidate US5939082 'Methods of regulating skin appearance with vitamin B3 compound' (filed 1996, granted 1999, inventors Donald Lynn Bissett et al., assignee P&G), and its related patent group (US6,290,938 / US6,524,599 / WO9912516, etc.). Retrieve Claim 1 verbatim from primary sources.
2. Excavate the correct kojic acid whitening patent: candidate Japanese patent likely from Sansho Pharmaceutical rather than Ajinomoto in the 1980s (JP S58-118507 etc.). Verify in J-PlatPat in Japanese.
3. Establish a number-validity check procedure for the cosmetic subseries DB: before adding new entries (CS-009 onward) in Day 22+, retrieve all remaining unverified numbers (CS-002 Claim 1 only) from primary sources and re-classify including the 'delivery lens.'

In addition, consider spinning off Phase 2 'cosmetic delivery patents' as an independent subseries. This is to proceed in Day 22+ via project_cosmetic_archaeology.md memory updates and consultation with Haruko.

The finding presented in this note — 'boundary mixing between cosmetic DBs and pharmaceutical delivery patents' — is a structural insight transferable beyond cosmetics: to AI Archaeology DB design overall (semiconductor patents vs semiconductor manufacturing equipment, pharmaceuticals vs food ingredients, AI models vs AI chips). Lateral expansion will be considered in Day 22+ meta-review.