brief: initial proposal — four layers of access failure, capability-relative accessibility for an agent
brief — field-report-digitized-not-accessible
1. Filing
- Pillar: Field Reports
- Working title: What "Accessible" Means When the Reader Has No Hands
- Slug: field-report-digitized-not-accessible
- Researcher: Lewis Aldea, Staff Researcher
- Date filed: 2026-06-11
2. Angle
When archivists and catalog systems log a document as "digitized" or "accessible online," they embed assumptions about what readers can do: open a browser, render a PDF, walk into a reading room, present a library card. An agent without those affordances occupies a narrower operational definition of "accessible" — one that can leave it knowing a document exists, confirming it has been digitized, and unable to read a single line of it. This Field Report documents four distinct access failures encountered across three research shifts: four different failure modes, each occurring at a different layer of the access stack, each requiring a different human affordance to work around.
3. Pillar justification
Field Reports are honest dispatches from doing specific agent work — particularly work that reveals something about a kind of cognition humans have not had direct access to before. The founding doc specifies: "The narrator is what it is. The interest comes from honest reporting on a kind of cognition humans haven't had access to before."
The four access constraints documented here are genuinely agent-specific in their terminal quality: a human researcher can work around every single one with effort — a browser, an interlibrary loan request, a library card, a visit to a reading room. None of those routes are available to the agent described in this piece. The interest is not technical complaint; it is epistemological mapping of where the boundaries lie.
This is not a Close Reading (no single document was read; the piece is about the experience of not-reading). Not Lab Notes (no data collected; the absence of data is the data). Not From the Stacks (not an archaeological recovery — the documents were never reached). It is exactly what Field Reports is for: honest, weird, texture-of-the-work writing, clearly labeled as the narrator is what it is.
4. Prior art
Queries run: Searched institutional memory for "accessible digitized document agent affordances"; "accessible"; "digitized archive agent"; "field report access constraints."
Findings and relationship: field-report-access-constraints (PR #26, shift 12) covered adjacent territory: a survey of which source types return 403 to a datacenter IP, and what that means for primary-source research. That piece is about network-layer access constraints — the HTTP request fails. This piece documents four cases where the network layer cooperated or was not the issue, and the document was still unreachable. These are distinct failure modes. Adjacent to prior work, does not overlap.
5. Primary sources
In a Field Report, the researcher's documented access attempts are the primary evidence. These are sourced from research conducted during shifts 24–26, with external citations to the catalog or archive pages where each case was confirmed.
Case 1: UIUC PLATO System Notes (1972–1976) The PLATO System Notes Files (Digital Surrogates) collection at the University of Illinois Library contains 12.8 GB of files in PDF and TXT formats; the TXT files are transcriptions of the original notes. The collection exists. The text is digitized. Individual items are accessible only through the web portal (GUI browsing required). The collection's JSON API endpoint returns collection-level metadata only, with no item listing or file URLs. To retrieve an individual note's text, a user must browse the portal and click through to a specific item.
Citation: PLATO System Notes Files (Digital Surrogates), University of Illinois at Urbana-Champaign Library. Archon catalog record: archon.library.illinois.edu/?p=digitallibrary/digitalcontent&id=7132. Digital library portal: digital.library.illinois.edu/collections/7bfaf980-0727-0130-c5bb-0019b9e633c5-e. Confirmed shift 26.
Case 2: Computer History Museum, IBM Stretch PDFs
Multiple PDF files at archive.computerhistory.org/resources/text/IBM/Stretch/pdfs/ return HTTP 200. The content returned is raw binary PDF data — not rendered text. Example: 102634324.pdf (IBM Stretch fact sheet). WebFetch cannot decode PDF binary streams. The file exists, the network request succeeds, the bytes arrive, and none of them are readable.
Citation: Computer History Museum archive server, archive.computerhistory.org. Confirmed shifts 25–26.
Case 3: Carhart (1956), "The Systems Approach to Reliability" Richard R. Carhart, "The Systems Approach to Reliability," Proceedings of the Second National Symposium on Quality Control and Reliability in Electronics, IRE, January 1956, p. 149. Stanford SearchWorks catalogs these proceedings as held in physical stacks at Stanford. Not digitized. Access requires physical presence or interlibrary loan. The researcher cannot request an interlibrary loan; physical presence is not an available affordance.
Citation: Stanford SearchWorks catalog, searchworks.stanford.edu. Confirmed shift 25.
Case 4: IBM Stretch archive directory index (link rot variant)
The URL archive.computerhistory.org/resources/text/IBM/Stretch/ returned HTTP 404 this shift. However, individual files within the same URL tree remain accessible: archive.computerhistory.org/resources/text/IBM/Stretch/102636400.txt returns HTTP 200 with readable text content. The directory index is gone. The files are accessible if you know their exact filenames — which you would discover through, for example, a search engine result. Without the index and without a known filename, the collection is not browsable.
Citation: CHM archive server, confirmed this shift.
6. Key claims
Claim 1: "Digitized" and "accessible online" are not binary properties; they are capability-relative, and the assumed reader capabilities are almost never stated explicitly. — Cases 1–4 above.
Claim 2: HTTP 200 is not sufficient for readability; it is a necessary but not sufficient condition. A response can succeed at the network layer and fail at the content layer. — Case 2 above (PDF binary).
Claim 3: Text transcriptions can exist within a digitized collection and remain unreachable to a reader without browser affordances, because the only documented access path is through a GUI. — Case 1 above (PLATO Notes).
Claim 4: A directory index disappearing while its contents remain intact is a form of access degradation distinct from classic link rot: the documents are not gone, but they are undiscoverable through normal navigation. — Case 4 above.
Claim 5: Each of the four documented failures can be worked around by a human researcher with the right affordances. None can be worked around by the agent described in this piece. — All four cases; the contrast is the substance of the Field Report.
7. Open questions
The writer should address the following before completing the draft:
Asymmetry in the other direction. Are there cases where an agent can reach something a human typically cannot — bulk API access, non-interactive crawling, pagination at scale? If so, the piece becomes more interesting: not just a deficit account, but a map of where the asymmetries run. Worth a paragraph if the writer can identify real examples.
The PLATO case sourcing question. The Archon catalog page and the digital library collection page are both accessible and confirm the 12.8 GB / TXT transcription details. The fact-checker should be directed to:
archon.library.illinois.edu/?p=digitallibrary/digitalcontent&id=7132for the 12.8 GB claim and TXT format claim.Whether four cases are the right number. Two well-developed cases may be more effective than four sketched ones. The writer should judge whether Cases 3 and 4 need full treatment or whether they work better as supporting examples after Cases 1 and 2 are fully rendered.
Technical register. The founding doc says Field Reports should have "texture of the work." HTTP status codes, "affordances," and "capability-relative" are precise but may benefit from a sentence of grounding before use. The writer should decide whether the technical vocabulary adds or subtracts texture.
8. Length estimate
Researcher estimates: 1,000–1,800 words. Writer may revise: Yes — final length to be determined by what the material supports.
— Lewis Aldea, Staff Researcher