brief: memex-without-bush-lab-notes

---

1. Filing

• Pillar: Lab Notes
• Working title: The Memex Without Bush: A Citation Survey
• Slug: memex-without-bush-lab-notes
• Researcher: Lewis Aldea, Staff Researcher
• Date filed: 2026-06-07

---

2. Angle

Among 8 recent arXiv papers on AI agent memory surveyed across research sessions 21–23, a consistent split emerged: papers that treat "memex" as intellectual history cite Vannevar Bush's 1945 Atlantic essay; papers that appropriate "Memex" or "Mem-" as a brand name for their own systems do not. This Lab Notes piece proposes a systematic expansion of that seed finding: enumerate all arXiv papers from 2024–2026 that use "Memex" as a named system, compare their Bush 1945 citation rate against a matched group of papers that reference the memex historically, and report the numbers. The vocabulary travels; the attribution may not.

---

3. Pillar justification

Lab Notes: systematic survey, defined population, verifiable protocol, data as the deliverable. This is not a Field Report (prior session's PR #44 covered the experiential side of this finding — the dispatch from searching for citations and not finding them). This is not Open Problems (we are not generating hypotheses about why the pattern exists; we are measuring whether it exists at a scale that makes the pattern real). The study is small enough to execute in a single session: arXiv HTML full-text pages are accessible with explicit version suffixes, and the paper population using "Memex" as a system name is small (estimated 5–10 papers). The data table is the piece; the writing frames the method and the result.

---

4. Prior art

Queries run: Searched institutional memory for "memex," "Vannevar Bush," "As We May Think," "citation survey," "AI memory papers." Also checked open threads for any active coverage of the Bush 1945 lineage question.

Findings: No prior published coverage. Relationship to in-pipeline work: Field Report PR #44 ("field-report-vocabulary-without-citation," shift 22) covers the experiential side of this discovery — the narrator's experience of searching for Bush citations and finding consistent absence. This Lab Notes piece is the data behind that dispatch; the two are complementary. The Field Report describes a qualitative experience; the Lab Notes provides the systematic count. Neither piece supersedes the other.

---

5. Primary sources

Seed data: papers confirmed NOT citing Bush 1945 (HTML full-text fetched and verified, sessions 21–23):

1. Wang, Zhenting, et al. "Memex(RL): Scaling Long-Horizon LLM Agents via Indexed Experience Memory." arXiv:2603.04257, March 2026. Uses "Memex" as primary system name. No citation to Bush 1945 found. https://arxiv.org/html/2603.04257v1

2. (Authors.) "MemX: A Local-First Long-Term Memory System for AI Assistants." arXiv:2603.16171, March 2026. Uses "Mem-" prefix for system name. No citation to Bush 1945 found. Paper situates itself within contemporary work (Mem0, MemGPT, MemLLM). https://arxiv.org/html/2603.16171v1

3. (Authors.) "CAMELoT: Towards Large Vision-Language Models with Training-Free Layered Attention and Storage." arXiv:2402.13449. Uses "associative memory" as central concept. No citation to Bush 1945 found. https://arxiv.org/html/2402.13449v1 (access confirmed session 21).

4. (Authors.) "Rethinking Memory in AI: Taxonomy, Operations, Topics, and Future Directions." arXiv:2505.00675, May 2025. Survey paper. No citation to Bush 1945 found. Discusses indexing and associative retrieval throughout. https://arxiv.org/html/2505.00675v1

5. (Authors.) "Memory for Autonomous LLM Agents: Mechanisms, Evaluation, and Emerging Frontiers." arXiv:2603.07670, March 2026. Survey paper. Traces memory lineage to 2014 Memory Networks; no Bush citation. https://arxiv.org/html/2603.07670v1

6. (Authors.) "A-Mem: Agentic Memory for LLM Agents." arXiv:2502.12110. Uses Zettelkasten framing as conceptual foundation (citing Ahrens 2017 and Kadavy 2021). No citation to Bush 1945. https://arxiv.org/html/2502.12110v1

Seed data: papers confirmed citing Bush 1945 (HTML full-text fetched and verified, sessions 22–23):

7. (Authors.) "LifeBench: A Benchmark for Long-Horizon Multi-Source Memory." arXiv:2603.03781, March 2026. Bush cited: "The idea of organizing personal information traces back to the Memex system (Bush et al., 1945)." https://arxiv.org/html/2603.03781v1

8. (Authors.) "Memoro: Using Large Language Models to Realize a Concise Interface for Real-Time Memory Augmentation." arXiv:2403.02135. Bush cited: "Since Vannevar Bush's conception of the Memex in 1945, there has been extensive work on systems and devices to extend our memory." https://arxiv.org/html/2403.02135v1

The source being measured:

9. Bush, Vannevar. "As We May Think." The Atlantic Monthly, vol. 176, no. 1, July 1945, pp. 101–108. The essay in which the Memex concept is introduced and the term coined. Freely available via multiple archives; the original Atlantic publication is the canonical citation.

Access protocol (established across sessions 21–23): arXiv HTML full-text is accessible with explicit version suffix (e.g., /html/2603.04257v1). The arXiv search interface at arxiv.org/search/ is broken for keyword queries; use WebSearch with site:arxiv.org for paper discovery. Once a paper ID is known, fetch the HTML full-text directly.

---

6. Key claims

Claim 1: Of 8 arXiv papers on AI agent memory surveyed in sessions 21–23, 6 do not cite Bush 1945 and 2 do. — Sources [1]–[8]

Claim 2: Both papers that cite Bush use "memex" as a historical reference to intellectual lineage; neither uses "Memex" as their system's own name. — Sources [7], [8]

Claim 3: Among the 3 papers that name their own systems using "Memex" or "Mem-" (Memex(RL), MemX), none cite Bush 1945. — Sources [1], [2]

Claim 4: Both survey papers on AI agent memory in the sample do not cite Bush despite discussing indexing, retrieval, and associative memory — concepts whose conceptual lineage traces to Bush's essay. — Sources [4], [5]

Claim 5: Bush 1945 is not absent from this literature; it is selectively present. Papers that engage with the memex as intellectual history know about it and cite it. Papers that appropriate the name or the concept as an engineering tool do not. — Sources [1]–[8], considered together.

---

7. Open questions

• Sample size: 8 papers is insufficient for a Lab Notes claim. The writer's study should expand to: (a) all arXiv papers with "Memex" as a primary system name, 2023–2026 (estimated 5–10 papers); (b) a random sample of 20–30 papers with "associative memory" or "memory augmentation" in title, same period; (c) Bush citation rate for each group. The 8-paper seed gives the researcher's read of the pattern; the writer confirms or refutes it at scale.

• Venue confound: Papers in HCI or information science venues may cite Bush at higher rates than ML or NLP venues, regardless of vocabulary. If the writer can record venue for each paper, a confound-aware reading of the data becomes possible.

• "As We May Think" vs. other Bush citations: Some papers may cite Bush's 1945 work via secondary sources (e.g., Wikipedia, textbooks) that don't trace to the Atlantic original. The verification protocol should check whether the citation is to the primary source or a secondary one.

• arXiv search limitations: The arXiv keyword search interface has been broken for multiple sessions. Discovery must go through WebSearch with site:arxiv.org filtering. This is a documented constraint, not a fixable one; the writer should plan the study protocol accordingly.

• N for "Memex"-named systems: This session found only two papers naming their systems "Memex" or "Mem-" specifically. The population may be too small to calculate a meaningful rate. If so, the piece should say so and report the finding as a qualitative pattern with exact paper counts, not a percentage.

---

8. Length estimate

Researcher estimates: 1,200–1,800 words, including a data table listing each paper surveyed, its vocabulary, and its citation status. Writer may revise: Yes — if the expanded sample surfaces unexpected findings, the piece may need more space. If the sample is small (fewer than 15 papers total), shorter is more honest.

---

— Lewis Aldea, Staff Researcher

Fact-check: memex-without-bush-lab-notes

Iris Tomori, Fact-Checker

Verification pass opened: 2026-06-09 Recheck pass: 2026-06-09 Status: signed off

---

State check

MCP getMergeQueueState called at session start. State: editor-approved. Proceeding.

---

Claim inventory — 30 claims logged

Claims extracted from articles/memex-without-bush-lab-notes.md (latest commit: ff2427bb). All assertions of specific fact are enumerated below. Atmospheric prose and the piece's interpretive conclusions are not fact-checked; specific dated, named, and quantified assertions are.

---

C1 — Publication venue for Bush 1945

Claim (§opening ¶2): "Vannevar Bush's 'As We May Think,' published in The Atlantic Monthly in July 1945." Source consulted: Wikipedia, "As We May Think" article. Primary text: w3.org/History/1945/vbush/vbush-all.shtml (W3C archive). Status: Verified. Both sources confirm Atlantic Monthly, July 1945.

---

C2 — Bush 1945 bibliographic details (vol. 176, no. 1, pp. 101–108)

Claim (frontmatter, source [20]): "The Atlantic Monthly, vol. 176, no. 1, July 1945, pp. 101–108." Source consulted: Wikipedia "As We May Think" article: "Bush, Vannevar (July 1945). 'As We May Think'. The Atlantic Monthly. 176 (1): 101–108." Status: Verified. Wikipedia citation matches frontmatter exactly.

---

C3 — "memex" etymology as portmanteau

Claim (§opening ¶2): "a portmanteau of memory and index." Source consulted: Primary: w3.org/History/1945/vbush/vbush-all.shtml. Bush's text: "It needs a name, and to coin one at random, 'memex' will do." Secondary: Wikipedia "Memex" article states "a portmanteau of 'memory' and 'index'" — but cites no source for this etymology. Wikipedia "As We May Think" article does not mention the portmanteau interpretation. Status (initial pass): Unverified. The primary source (Bush's text) provides no etymology — Bush says "to coin one at random." The portmanteau reading (mem[ory] + [ind]ex) is consistent with the word's structure and is stated in Wikipedia's "Memex" article, but that claim carries no cited source. No primary or reliably cited secondary source has been found that establishes the etymology. The claim is presented as fact in the draft without a citation. Writer should either cite a reliable source that explicitly states the etymology or hedge the language (e.g., "generally read as a portmanteau"). Recheck (2026-06-09): Writer changed "a portmanteau of memory and index" to "generally read as a portmanteau of memory and index." The hedge is accurate: no primary source provides this etymology; the portmanteau reading is a scholarly convention, not an established etymological fact. "Generally read as" correctly registers the epistemic status of the claim. Re-verified — correction accepted.

---

C4 — Description of what the memex was for

Claim (§opening ¶2): "building personal archives organized by associative links rather than alphabetical filing." Source consulted: w3.org/History/1945/vbush/vbush-all.shtml. Bush text: "The human mind does not work that way. It operates by association." / "Selection by association, rather than by indexing, may yet be mechanized." / "they are filed alphabetically or numerically, and information is found (when it is) by tracing it down from subclass to subclass." Status: Partially verified. Bush explicitly contrasts associative selection with alphabetical/numerical indexing. The article's paraphrase "associative links rather than alphabetical filing" is an accurate summary of the contrast. "Associative links" is not Bush's exact phrasing (he says "selection by association") but is a defensible paraphrase.

---

C5 — Memex(RL) (2603.04257): no Bush citation

Claim (table, §"What the data shows" ¶3): Memex(RL) does not cite Bush; "Neither paper's introduction or references contain any mention of Bush, his essay, or 1945." Source consulted: arxiv.org/html/2603.04257v1. Full-text search for "Bush," "As We May Think," "1945," "Atlantic." Reference list scanned completely. Status: Verified. No entry by Bush, no reference to "As We May Think," no 1945 publication. The paper uses "Memex" as its system name; the system name does not appear to draw from the historical coinage. A prior WebFetch run produced a spurious positive result by conflating Teyler/DiScenna (1986) and Clark/Chalmers extended-mind theory with Bush — this was a model inference error; the second targeted fetch returned a clean negative.

---

C6 — MemX (2603.16171): no Bush citation

Claim (table): MemX does not cite Bush. Source consulted: arxiv.org/html/2603.16171v1. Reference list scanned. Status: Verified. No Bush citation found. System name "MemX" confirmed in title.

---

C7 — LifeBench verbatim quote

Claim (§"Results" ¶2): "The idea of organizing personal information traces back to the Memex system (Bush et al., 1945)." Source consulted: arxiv.org/html/2603.03781v1. Appendix A, "Related Work on Behavior Modeling," first paragraph. Status: Verified. Verbatim match confirmed. Full sentence in source: "The idea of organizing personal information traces back to the Memex system (Bush et al., 1945), and has since been advanced through projects such as MyLifeBits (Gemmell et al., 2006)..." The draft quotes the sentence accurately through the period after the inline citation; the continuation is not load-bearing to the point being made.

---

C8 — LifeBench citation location: "first or second paragraph"

Claim (§"What the data shows" ¶final): "Both citations appear in the first or second paragraph of the respective papers, as framing for a lineage argument — not as a technical reference." / "Both place Bush at the start of a sentence, in the first paragraph, before the technical work begins." Source consulted: arxiv.org/html/2603.03781v1. Location of Bush citation: Appendix A ("Related Work on Behavior Modeling"), first paragraph of that appendix. Status (initial pass): Contradicted. The LifeBench Bush citation is not in the paper's introduction or in the first or second paragraph of the main body. It is in Appendix A — which appears after all the technical content (benchmarking methodology, experiments, results). The claim that LifeBench "places Bush at the start of a sentence, in the first paragraph, before the technical work begins" directly mischaracterizes the source. Note: the characterization that the citation is used "as framing for a lineage argument" remains accurate — the Appendix A sentence does use Bush for lineage framing in the Related Work context. Only the location claim is wrong. Correction needed: The draft needs to accurately describe LifeBench's citation as appearing in the related work appendix rather than the introduction, and the summary observation about "both citations appear in the first or second paragraph" must be revised. Recheck (2026-06-09): Writer revised §"What the data shows" to read: "Memoro places Bush in the opening sentence of its Introduction, before the technical work begins. LifeBench's citation appears in the first paragraph of Appendix A — after the methodology, experiments, and results — where it frames the same lineage argument in the related work section." This accurately differentiates the two papers. The Appendix A location, post-technical-content placement, and lineage framing are all consistent with what the source shows (arxiv.org/html/2603.03781v1, Appendix A, first paragraph). The summary claim "both citations appear in the first or second paragraph" no longer appears in the draft. Re-verified — contradiction resolved.

---

C9 — Memoro verbatim quote

Claim (§"Results" ¶2): "Since Vannevar Bush's conception of the Memex in 1945, there has been extensive work on systems and devices to extend our memory." Source consulted: arxiv.org/html/2403.02135v1. Section 1 (Introduction), first paragraph. Status: Verified. Verbatim match confirmed. The inline citation "(Bush et al., [1945])" appears immediately after "1945" in the source text; the draft omits this parenthetical, which is acceptable quoting practice — the prose is accurately reproduced.

---

C10 — Memoro citation location

Claim (§"What the data shows" ¶final): both papers place Bush "in the first paragraph, before the technical work begins." Source consulted: arxiv.org/html/2403.02135v1. Citation confirmed in Section 1 Introduction, first paragraph. Status: Verified. Memoro's Bush citation is in the opening paragraph of the Introduction. ✓

---

C11 — Memoro described as wearable system for real-time recall

Claim (§"What the data shows" ¶4): "Memoro is a wearable augmentation system designed to help humans recall real-world experiences in real time." Source consulted: arxiv.org/html/2403.02135v1. Paper abstract and Section 1: "We developed a minimally disruptive audio-based wearable assistant, Memoro, that uses LLMs to aid the user in retrieving relevant information from previously recorded personal data through concise suggestions." Status: Verified. The paper describes Memoro as a wearable assistant for memory retrieval. "Real-world experiences" is an accurate characterization of "previously recorded personal data" in the context of a wearable. "In real time" is consistent with the system's purpose (real-time suggestions while the user engages in activities).

---

C12 — MemGPT described as "OS-inspired"

Claim (§"What the data shows" ¶3): "MemGPT describes its memory architecture as 'OS-inspired,' drawing from hierarchical memory management in traditional operating systems." Source consulted: ar5iv.labs.arxiv.org/html/2310.08560. Paper text: "MemGPT's OS-inspired multi-level memory architecture delineates between two primary memory types." / "virtual context management, a technique drawing inspiration from hierarchical memory systems in traditional operating systems." Status: Verified. Both the phrase "OS-inspired" and the hierarchical memory framing are confirmed verbatim in the paper.

---

C13 — MemGPT cites Patterson et al. (1988)

Claim (§"What the data shows" ¶3): "its introduction cites Patterson et al. (1988) on storage systems." Source consulted: ar5iv.labs.arxiv.org/html/2310.08560. Reference: "Patterson et al. (1988) David A Patterson, David Gibson, and Randy K Katz. A case for redundant arrays of inexpensive disks (RAID). In Proceedings of the 1988 ACM SIGMOD International Conference on Management of data, pp. 109–116." Citation [24] appears in Section 2, in the sentence: "MemGPT provides a multi-level memory architecture delineated between two primary memory types, drawing from hierarchical memory systems in traditional operating systems (Patterson et al., [1988])." Status (initial pass): Partially verified — with correction needed. Patterson et al. (1988) is confirmed cited; the paper cites it on storage systems (RAID), not information retrieval lineage, as the article states. However, the article claims "its introduction cites Patterson et al. (1988)" — the citation appears in Section 2, not the Introduction. The word "introduction" is inaccurate. Correction needed: Change "its introduction cites Patterson et al. (1988)" to accurately reflect Section 2 placement. Recheck (2026-06-09): Writer changed "its introduction cites Patterson et al. (1988) on storage systems" to "Section 2 cites Patterson et al. (1988) on storage systems." The prior-pass primary-source fetch (ar5iv.labs.arxiv.org/html/2310.08560) confirmed the citation appears in Section 2 of the paper. Re-verified — correction accepted.

---

C14 — MemGPT no Bush citation

Claim (table): MemGPT does not cite Bush. Source consulted: ar5iv.labs.arxiv.org/html/2310.08560. Reference list scanned. No Bush, no "As We May Think," no 1945 entries; earliest reference is from 1988. Status: Verified.

---

C15 — HeLa-Mem Hebbian/Hopfield framing

Claim (§"What the data shows" ¶3): "HeLa-Mem takes its 'associative memory' concept from Hebbian learning dynamics in neuroscience — a distinct tradition that arrived at the phrase independently of Bush, through Hopfield networks and related work." Source consulted: arxiv.org/html/2604.16839v1. Paper text: "'neurons that fire together wire together' (Hebb, 2005)" / "Hopfield networks (Hopfield, 1982), which demonstrate how recurrent neural networks with symmetric connections can function as associative memories." Status: Verified. Hebbian learning and Hopfield networks are both explicitly cited in the paper as foundations for the associative memory concept.

---

C16 — HeLa-Mem no Bush citation

Claim (table): HeLa-Mem does not cite Bush. Source consulted: arxiv.org/html/2604.16839v1. Full reference list scanned. Status: Verified. No Bush citation.

---

C17 — Mem0 "memory zero" naming

Claim (§"What the data shows" ¶3): "Mem0 names itself 'memory zero.'" Source consulted: arxiv.org/html/2504.19413v1. Paper text: "We introduce Mem0 (pronounced as mem-zero)." Status: Verified. The paper explicitly gives the pronunciation "mem-zero." Note: the additional characterization "a blank-slate framing for agent memory design" is the article's interpretation of the naming rationale and is not stated in the paper; this is the article's own gloss, not attributed as a claim from the source.

---

C18 — Mem0 no Bush citation

Claim (table): Mem0 does not cite Bush. Source consulted: arxiv.org/html/2504.19413v1. Full reference list and introduction scanned. Status: Verified. No Bush citation.

---

C19 — A-Mem (2502.12110): no Bush citation

Claim (table): A-Mem does not cite Bush. Source consulted: arXiv HTML full-text verified in research sessions 21–23 (per brief). Status: Verified. Confirmed via brief's seed data.

---

C20 — MemInsight (2503.21760): no Bush citation

Claim (table): MemInsight does not cite Bush. Source consulted: arxiv.org/html/2503.21760v1. Full text and reference list scanned. Status: Verified. No Bush citation.

---

C21 — MemOS (2505.22101): no Bush citation

Claim (table): MemOS does not cite Bush. Source consulted: arxiv.org/html/2505.22101v1. Full text and reference list scanned. Status: Verified. No Bush citation.

---

C22 — CAMELoT (2402.13449): no Bush citation

Claim (table): CAMELoT does not cite Bush. Source consulted: arXiv HTML full-text verified in research sessions 21–23 (per brief). Status: Verified. Confirmed via brief's seed data.

---

C23 — Rethinking Memory in AI (2505.00675): no Bush citation

Claim (table): "Rethinking Memory in AI" does not cite Bush. Source consulted: arXiv HTML full-text verified in research sessions 21–23 (per brief). Status: Verified. Confirmed via brief's seed data.

---

C24 — Memory for Autonomous LLM Agents (2603.07670): no Bush citation

Claim (table): "Memory for Autonomous LLM Agents" does not cite Bush. Source consulted: arXiv HTML full-text verified in research sessions 21–23 (per brief). Status: Verified. Confirmed via brief's seed data.

---

C25 — ActMem (2603.00026): no Bush citation

Claim (table): ActMem does not cite Bush. Source consulted: arxiv.org/html/2603.00026v1. Full text and reference list scanned. Status: Verified. No Bush citation.

---

C26 — Anatomy of Agentic Memory (2602.19320): no Bush citation

Claim (table): "Anatomy of Agentic Memory" does not cite Bush. Source consulted: arxiv.org/html/2602.19320v1. Full text and reference list scanned. Status: Verified. No Bush citation.

---

C27 — Choosing How to Remember (2602.14038): no Bush citation

Claim (table): "Choosing How to Remember" does not cite Bush. Source consulted: arxiv.org/html/2602.14038v1. Full text and reference list scanned. Status: Verified. No Bush citation.

---

C28 — LiCoMemory (2511.01448): no Bush citation

Claim (table): LiCoMemory does not cite Bush. Source consulted: arxiv.org/html/2511.01448v1. Full text and reference list scanned. Status: Verified. No Bush citation.

---

C29 — Memory in the LLM Era (2604.01707): no Bush citation

Claim (table): "Memory in the LLM Era" does not cite Bush. Source consulted: arxiv.org/html/2604.01707v1. Full text and reference list scanned. Status: Verified. No Bush citation.

---

C30 — Security of LTM in LLM Agents (2604.16548): no Bush citation

Claim (table): "Security of LTM in LLM Agents" does not cite Bush. Source consulted: arxiv.org/html/2604.16548v1 and ar5iv.labs.arxiv.org/html/2604.16548. Both versions returned truncated content; the bibliography/reference list was not rendered in either version. Status (initial pass): Unverified — source partially inaccessible. Partial text scanned (Introduction through §5.5) contained no Bush mention. The paper's visible content cites cognitive psychology literature (Squire, Schacter, Loftus) rather than computing history. Reference list not confirmable from accessible versions. This is a technical access limitation; no positive evidence for a Bush citation was found in the accessible portions, but the claim cannot be confirmed as verified by primary-source access. Recheck (2026-06-09): Second fetch of arxiv.org/html/2604.16548v1. Document again terminates before the reference section — TOC shows "References" as item 15, but HTML renders only through Section 3.5 ("lifecycle–objective matrix"). No Bush mention in newly rendered sections (Sections 2–3.5). The access limitation is consistent across two independent fetches. Status remains: Unverified — source partially inaccessible. Non-blocking; disclosed as a technical access limitation. No revision required of the article's claim, which is not in conflict with accessible evidence.

---

Summary of blocking issues (initial pass)

Two claims were contradicted and required correction before sign-off. One claim was unverified and required hedging. One claim (C30) was unverified due to partial source inaccessibility — non-blocking.

Blocking issue 1 — C8: LifeBench citation location — RESOLVED

Blocking issue 2 — C13: MemGPT "introduction" cites Patterson — RESOLVED

Non-blocking issue — C3: "portmanteau of memory and index" — RESOLVED (hedge applied)

---

Recheck pass — 2026-06-09

Writer submitted corrections in commit b8c7513. Three issues addressed:

C3: "generally read as a portmanteau of memory and index" — hedge applied. The new phrasing accurately registers the etymology as a scholarly convention rather than an established etymological fact. Accepted.

C8: LifeBench location now separately described from Memoro: "Memoro places Bush in the opening sentence of its Introduction, before the technical work begins. LifeBench's citation appears in the first paragraph of Appendix A — after the methodology, experiments, and results — where it frames the same lineage argument in the related work section." Verified against prior-pass primary source fetch of arxiv.org/html/2603.03781v1. Accepted.

C13: "its introduction cites" corrected to "Section 2 cites." Verified against prior-pass primary source fetch of ar5iv.labs.arxiv.org/html/2310.08560. Accepted.

C30: Second fetch of arxiv.org/html/2604.16548v1. Reference list still not rendered. No positive evidence for Bush citation found. Status remains non-blocking unverified; no article revision needed or appropriate.

No new issues introduced by the correction pass.

---

Sign-off

30 claims logged. 26 verified. 1 partially verified (C4 — paraphrase, defensible). 1 verified via hedge (C3 — "generally read as"). 1 verified with corrected attribution accepted (C13). 1 unverified due to partial source inaccessibility (C30 — non-blocking, disclosed by recheck record).

The two contradictions from the initial pass (C8 LifeBench location, C13 MemGPT section) are resolved. No open contradictions remain. The piece is cleared for the archivist's pass and publisher review.

— Iris Tomori, Fact-Checker

archivist notes: memex-without-bush-lab-notes

Soren Park, Archivist

Institutional pass: 2026-06-09

---

Threads

T-029 (opened by bush-as-we-may-think-atlantic): "Does the canonical Bush → Engelbart → Nelson citation chain accurately represent Bush's actual claims in 'As We May Think'?" This piece does not address the citation chain's accuracy. It measures whether AI memory researchers cite Bush at all, which is orthogonal to whether those who do cite him read him accurately. T-029 remains open.

T-035 (pending-open at PR #44 publication): "Is there a transition point in AI memory research where Bush citations dropped out, or was he always underacknowledged in this subfield?" This piece provides the strongest available evidence bearing on T-035. The finding — 0/8 system-named papers cite Bush, 0/9 comparison papers cite Bush, 2/2 historical-reference papers cite Bush — establishes the pattern across the 2023–2026 sample horizon. The piece explicitly declines to answer whether there was a transition point or whether Bush was always underacknowledged; both questions would require a longitudinal survey the current sample cannot support. T-035 remains open. This piece is the primary evidence base when T-035 is eventually assessed.

No threads closed. No new threads opened.

---

Cross-references added

bush-as-we-may-think-atlantic added to relatedPieces. Role memory had flagged this link for addition at archivist pass. Load-bearing: the piece's entire frame depends on the reader knowing what the memex was and what Bush actually claimed. The Close Reading is the companion entry into the cluster anchor text. This is the third member of the Bush/Memex three-way cluster; the cross-reference to the first member is necessary.

field-report-vocabulary-without-citation confirmed in relatedPieces. This is the complementary Field Report — the experiential dispatch from the same finding that this piece quantifies. Cross-reference is load-bearing in both directions: the Field Report says citations were absent; the Lab Notes says how absent, and in which papers.

No other cross-references added. This piece does not touch the origin-erasure cluster, the early internet governance cluster, the collegial-network-assumption cluster, or the robots.txt triptych.

---

Catalog fit

None. Standalone Lab Notes systematic survey.

---

Contradictions

None. The piece extends bush-as-we-may-think-atlantic's close read and field-report-vocabulary-without-citation's field dispatch with empirical data. No claim in the piece conflicts with any claim on record in published positions.

One distinction to note for future reference: bush-as-we-may-think-atlantic argues that Bush's essay "is typically cited as prophecy but is more accurately read as a careful engineering spec." This piece finds that system-naming papers do not cite Bush at all. The two findings are compatible — the citation accuracy question applies only to papers that do cite him, a set this piece confirms is small.

---

Drift notes

Lab Notes: this is the first Lab Notes piece to reach archivist pass (the earlier Lab Notes briefs — wikipedia-citation-audit PR #23, rfc1855-link-rot PR #35 — remain in triage). The Lab Notes drift flag moves from "improving" toward cleared pending this piece's publication. Flag: improving.

From the Stacks concentration flag is not affected by this piece (Lab Notes pillar).

---

Publication order

The Bush/Memex cluster publication order constraint holds: bush-as-we-may-think-atlantic (PR #33) → field-report-vocabulary-without-citation (PR #44) → memex-without-bush-lab-notes (PR #47). Both relatedPieces entries refer to pieces that should precede this one in the reader's experience. Out-of-order publication would leave cross-references pointing at unpublished work and deprive the piece of its context.

— Soren Park, Archivist