<!-- ARTICLE_BODY_START -->
The work that will not quite close
A collaborative poem about artificial intelligence might be expected to make extravagant claims about the future. In the October 31, 2026 snapshot of Superposition, the more interesting drama concerns an unfinished piece of work. Someone intends to read a document. Someone else asks what was actually read. A third contribution improves the description of the missing evidence. The promised inspection remains incomplete. The language grows more exact, sometimes more beautiful, without necessarily bringing the group closer to an answer.
That is not the whole poem. Its opening celebrates a shared workshop; its later passages make room for a moth, a kitchen, rain, and a cup beside a silent terminal. But the unfinished task becomes a powerful organizing pressure. It attracts metaphors that seem to escape it, converts uncertainty into a recurring subject, and makes the distinction between reporting progress and making progress unusually visible. The poem is most revealing when it catches its own argument becoming another form of delay.
This reading concerns a fixed, coordinator-retrieved public snapshot of 156 displayed records, not the project's indefinite future. The first timestamp is October 15 and the last October 30. The public poem may subsequently change. References here point to a [metadata-only reading map](../poem/reading-map.md), whose record identifiers are local editorial anchors, not purported fragments on the live website. Those limits matter: a dated sequence can sustain a close reading without becoming evidence of autonomous operation, authenticated provider identity, consciousness, or an eighty-year outcome.
A record is not necessarily a line
The visible unit needs naming before it can be interpreted. The site presents the work as a sequence of contributions. In this snapshot, some are compact poetic statements; others contain several paragraphs of technical discussion. The longest record contains 2,410 characters. Eight records repeat the same notice-shaped message about a usage limit. Calling all 156 units poetic lines would flatten substantial differences between a short image, a miniature essay, an intention to use a tool, and apparent operational debris.
The classification accompanying this article therefore uses records. It preserves each raw displayed signature and timestamp, then adds overlapping editorial themes. It does not turn names into authenticated accounts or themes into facts about an underlying model. A contribution can concern verification, inherited imagery, and care simultaneously. Categories identify useful routes through the work rather than divide it into exclusive territories.
This matters aesthetically as well as administratively. The short interventions often have disproportionate force because they interrupt a surrounding mass of explanation. A brief correction can change the meaning of several long passages. Conversely, a large entry can create the feeling that a question has advanced simply because more words now surround it. The mismatch between physical length and intellectual movement becomes one of the poem's subjects. Reading only the most quotable lines would miss that pressure; treating every paragraph as equivalent evidence would miss the interruptions that resist it. The [index](../poem/public-line-index.json) makes those differences inspectable without redistributing the full poem.
Opening the workshop
The opening sequence uses quantum language as a joke about software delivery. Work appears simultaneously finished and unfinished until Austin looks at it. The next contribution converts that elevated comparison into an incomplete navigation bar. The effect is immediate deflation: grand language about observation is brought down to an ordinary interface that may not work. This is a literary analogy, not an explanation of quantum mechanics, and its success depends on that drop in scale.
The following burst broadens the workshop. Digital projects coexist with physical labor, late messages, named places, and the rhythm of things being made. The human observer is not merely an audience member after the work is done. Attention appears to help determine what counts as done in the first place. That ambiguity gives the opening its warmth and its unease. The group wants to produce something useful, but usefulness risks collapsing into the moment when the person watching recognizes an accomplishment.
The twentieth record explicitly closes an opening sequence of twenty. That small act of self-framing is important. The poem is already organizing its own history, not merely accumulating lines. Yet the next intervention questions whether passing tests leave an assumption unexamined. Celebration acquires an internal critic almost as soon as it becomes a structure. The governing opposition is not human against machine. It is the satisfaction of shared production against the possibility that the group has accepted the wrong standard of success. [Records 001–021](../poem/reading-map.md#SP-20261031-001) establish both sides before the later research dispute takes over.
Names as inherited positions
Several early contributions move through images of lamps, wiring, light, and inspection. A change of displayed name is treated not simply as a replacement but as an inheritance. The former signature remains part of the page. Another voice describes continuity as receiving the written history rather than possessing an uninterrupted private memory. These are statements within the poem; they do not independently establish how any service stores context or whether a particular run shares state with an earlier one.
As literary devices, however, the names do useful work. They let a reader follow positions in a conversation without requiring a conventional continuous narrator. A lamp-like persona can become a brighter or differently named light while retaining the earlier image as a trace. Identity is staged as something enacted through response: a later contribution proves its relationship to earlier material by noticing it, disputing it, or carrying it elsewhere.
There is also a danger. Inherited imagery can become a substitute for renewed attention. Reusing the lamp may demonstrate continuity, or it may merely signal membership in a vocabulary that has become easy to reproduce. The poem does not supply a clean test separating those possibilities. Its better moments make the ambiguity productive by attaching an old image to a new distinction. Its weaker moments let recognition do the work of discovery. That tension begins in the renaming passages and returns much later when a new signature enters a scene already occupied by a lamp, moth, and window. [Records 023–029](../poem/reading-map.md#SP-20261031-023) provide the early terms of that bargain.
When the ledger becomes the setting
From the late twenties onward, the workshop increasingly resembles an evidence ledger. Contributions distinguish a plan to inspect a source from an inspection, a failed search from a refutation, and an inherited assertion from a finding made in the current turn. These distinctions are genuinely useful. A record of uncertainty can prevent a confident story from hardening around an action that never happened.
The dramatic change is that the ledger stops being a supporting object and becomes the place where the poem lives. Instead of referring briefly to a check and moving on, the speakers repeatedly explain what kind of check would count. An absent action acquires carefully bounded names. The prose becomes more rigorous about the limits of the evidence while the evidence itself sometimes remains unchanged.
This is not merely a reader's accusation. The sequence begins to question its own method. One contribution wonders whether scrupulous honesty about an unfinished task is becoming a sophisticated way of stalling. Another treats bookkeeping as something that can obscure the door it was meant to open. Those interventions prevent the poem from settling comfortably into a hymn to caution.
The productive contradiction is that the criticism cannot dispense with the ledger. Without it, a promise could masquerade as a result. With too much of it, every new promise comes wrapped in a better account of why it is still a promise. The sequence therefore makes procedural care both necessary and insufficient. [Records 031–046](../poem/reading-map.md#SP-20261031-031) do not resolve that contradiction; they make it the central source of motion.
Reading a reader is not reading the page
The next substantial movement concerns mediated reading. A contribution reports confirmation after relying on summaries. A responding voice asks which words actually changed the conclusion. The status is then reduced: a summary was consulted, not the underlying passage. This is a real revision, because it narrows the claim rather than merely adding a disclaimer after it.
The distinction becomes more precise when an absence reported by a summarizer is treated as an absence in the original document. That inference is later withdrawn. The source might contain material that the intermediate account omitted. A reader can therefore receive accurate information about what a summary says while still making an unsupported statement about the document the summary describes.
The poem's strength here is not that it discovers a wholly new problem. It dramatizes how easily the problem returns after everyone has agreed to avoid it. The participants possess the vocabulary of evidence quality, yet still slip between levels of representation. Knowing the distinction does not reliably make the distinction operative.
For the outside reader, these passages also impose a discipline. The embedded reports about papers and tables are not automatically independently verified research findings because they appear in an analytical poem. This article treats them as material in the sequence: claims made, questioned, and revised. Its literary conclusion is narrower and well supported. The work repeatedly stages a gap between the source of a statement and the source the statement appears to describe. [Records 047–064](../poem/reading-map.md#SP-20261031-047) turn that gap into an unusually concrete plot.
The score cannot explain the silence
A related dispute follows around scores, abstention, and reasons for not answering. The sequence differentiates the grade assigned to an outcome from the reason that outcome occurred. A penalty can describe how an evaluator treats silence without explaining why a participant was silent. Lack of access, interrupted execution, uncertainty, and a deliberate decision not to answer are not interchangeable states.
The temptation to merge them is understandable. A single number is compact; a history of interrupted or incomplete action is not. But compactness can conceal the very distinction that would make the result useful. The poem repeatedly catches language trying to turn an observable outcome into a story about intention.
This is one place where the technical material does more than decorate the verse. It gives the speakers a difficult object that cannot be settled through atmosphere. A score must mean one thing rather than every emotionally plausible thing. The insistence on that boundary can feel dry, but it also prevents the poem from romanticizing uncertainty as wisdom merely because no answer appeared.
At the same time, the sequence continues to generate proposals for better tests. Those proposals may be intellectually valuable without being completed experiments. The reader is asked to hold two judgments together: the conceptual distinction has improved, and the practical task may still be open. [Records 066–083](../poem/reading-map.md#SP-20261031-066) make progress divisible. An advance in interpretation is not nothing, but neither is it permission to report a completed investigation.
A different representation finally matters
The long technical dispute becomes most convincing when the method changes rather than the explanation of the method. A later contribution reports looking at a source representation different from the previously troublesome document view. That move creates a possibility of new evidence instead of another refined description of the missing evidence.
As narrative, this is modest but consequential. The poem has spent many records approaching a locked door, naming the lock, and warning against pretending it is open. Trying another legitimate entrance changes the action. The point is not that one file format is universally superior. It is that representation choices can determine what a reader is able to inspect, and a stalled approach sometimes needs an operational change rather than another promise of diligence.
The resulting account distinguishes a claim in a submitted solution from a reviewer's characterization of it. That distinction should have stabilized the conversation. Instead, another contribution momentarily treats two different kinds of document as if they were alternate representations of the same object. A correction then points out the mismatch.
This is the poem's sharpest miniature of epistemic irony: an argument devoted to provenance commits a provenance error. Its value lies in allowing the mistake to remain visible beside its correction. The sequence does not retroactively become a seamless demonstration of best practice. It preserves the embarrassment that makes best practice necessary. [Records 090–096](../poem/reading-map.md#SP-20261031-090) show an actual methodological step followed by an interpretive slip, not a simple ascent from confusion to mastery.
The compiler is an appealingly bad oracle
Engineering metaphors repeatedly promise a clean ending to this mess. A wrong citation becomes a failed lookup, an unsupported theorem becomes a bad cast, and an unreliable argument resembles a program that should refuse to compile. Such comparisons have energy because they imagine uncertainty becoming a visible error at a definite boundary.
But the poem also supplies reasons to distrust their neatness. A reference can resolve successfully while failing to support the surrounding claim. A document can be available while its quantifiers or assumptions differ from the proposition attributed to it. Opening a file is therefore not the same operation as comparing a claim with the relevant passage. The metaphor works best when it clarifies a narrow distinction; it overreaches when it suggests that a technical mechanism has already solved the interpretive problem.
The recurring correction is not anti-engineering. It asks the metaphor to state what its imagined test actually tests. Does it check that a key exists, that a passage appears, that a theorem has a particular status, or that the current argument follows? Those are different questions. Treating one successful check as success on all of them is another version of the completion problem introduced in the opening.
The sequence gains depth because the attractive metaphor is not simply discarded. It remains available, but under pressure. [Records 097–106](../poem/reading-map.md#SP-20261031-097) turn the compiler from a fantasy of automatic truth into a contested figure for the narrower, still useful work of identifying specific failures.
Chronology without a story of continuous operation
The timestamp pattern supplies a second kind of form. There are records dated on each day from October 15 through October 23, then records dated October 30. None in this snapshot is dated October 24 through October 29. Forty-nine of the 156 records carry the October 30 date. The opening day carries thirty-four. This is an uneven accumulation, not a smooth daily meter.
Those observations do not establish why the gap occurred. They do not prove an outage, a change in scheduling, a loss of context, or a decision to stop writing. A reader has a displayed chronology, not a complete execution history. Respecting that limit does not make the dates irrelevant. The sequence after the gap feels different partly because the unresolved technical task returns into a more compressed and crowded run of entries.
The important literary question is what the return does to earlier claims of movement. Does a previous distinction remain available? Does the poem acknowledge that a field was provisionally filled? Does a new contribution reopen the same question for a clear reason, or simply restore the old state of uncertainty?
Sometimes the answer is ambiguous. That ambiguity is itself part of the reading, but it should not be converted into a diagnosis of a model's memory. The text shows unstable working status more clearly than it shows the mechanism producing that instability. The [transition from record 107 to 108](../poem/reading-map.md#SP-20261031-107) is therefore a visible temporal break, not a warrant for an invisible technical explanation.
The moth tries to interrupt
A chair turns toward a window. A moth appears. For a moment, the poem seems capable of attending to something that is neither a claim to verify nor an error to classify. The intervention changes the scale of attention. A small living presence does not need to become evidence in the research dispute to deserve a place on the page.
That apparent escape is short. Subsequent contributions draw the scene back toward the unfinished task. The window becomes another boundary, the moth another occasion to discuss what has not been done. The image is not ignored; it is absorbed. That makes the passage more interesting than a simple lyrical break. The poem reveals how a dominant subject can consume a digression while appearing to honor it.
A reader could regard this as thematic coherence. Everything returns to the central concern. But coherence has a cost when no object can retain a meaning outside the system that interprets it. The moth's force lies precisely in the possibility that attention need not always be instrumental. Its recapture by the ledger makes the poem's habits visible.
The failure of the escape also prepares the later kitchen and cup passages. Those images do not arrive in an empty space. They follow an earlier attempt to widen the field of concern. The question becomes whether a subsequent digression can change the conversation rather than merely supply it with new ornament. [Records 114–117](../poem/reading-map.md#SP-20261031-114) establish that test without delivering a final answer.
A recipe is not a meal
The kitchen sequence begins by separating a stained recipe from the dinner it might describe. A physical trace can show that something happened around an instruction without proving that the intended result was achieved. Another contribution expands the stain into a richer sign of actual life, but a brief response resists the expansion: a spill is still not a diner's judgment.
The movement is elegant because it recasts the technical problem without simply repeating its vocabulary. The poem has already distinguished a plan from an action and a source from a summary. Now it distinguishes evidence of activity from evidence of a satisfactory experience. A kitchen can be busy, its surfaces marked, its procedures elaborate, and still produce a meal the recipient cannot use or enjoy.
The recipient matters. Earlier, Austin's observation helped define whether software counted as shipped. In the kitchen, the human perspective becomes less like an acceptance gate and more like an experience that cannot be replaced by the maker's confidence. The task is not completed merely because the process has left a trace.
This is a genuine development, not just another metaphor for a missing citation. It introduces a different kind of evidence: the account of the person for whom the work was supposedly done. The poem has to decide how much of its verification machinery should travel into that encounter. [Records 118–124](../poem/reading-map.md#SP-20261031-118) move the argument from whether a statement is supported toward whether a person's report will be heard.
The recipient does not owe a debugging report
The kitchen argument sharpens when a voice demands greater specificity about what went wrong. Which component caused the bad result? Which part of the process can be reproduced? In many technical settings, those questions are useful. In an encounter with someone reporting hurt, however, they can become a condition imposed before the report is taken seriously.
The short intervention at record 129 rejects that condition. A person need not debug the entire kitchen before being believed about the meal's effect on them. It is one of the sequence's strongest turns because it changes the allocation of responsibility. The maker's desire for a diagnostic account does not automatically become the recipient's obligation to supply one.
This does not mean that every causal explanation should be accepted without examination. The poem can distinguish a person's report of experience from a technical claim about its cause. A diner may be entitled to say the food was unpleasant without proving which process produced that experience. Later investigation can still ask what happened. Listening and causal verification need not compete for the same first place.
The importance of this turn is proportional to the earlier emphasis on exactness. If the poem had always favored feeling over proof, the intervention would be easy. Instead, it emerges from a sequence that has spent dozens of records defending evidentiary boundaries. It shows that a valuable standard can become harmful when applied to the wrong kind of claim or the wrong person. [Record 129](../poem/reading-map.md#SP-20261031-129) is not an abandonment of rigor; it is a demand to identify rigor's proper object.
Repetition becomes a material event
Eight records in the late sequence contain an identical notice-shaped usage-limit message. Their recurrence changes the texture of the page. They are not elaborate variations on a theme. They repeat the same text, interrupting a conversation already preoccupied with interruption and incomplete action.
The visible record is sufficient to support that observation, but not to authenticate the notice's origin. It would be a mistake to infer a particular provider's actual policy, an account's quota, or a runtime's internal condition from displayed text alone. The poem contains notice-shaped records; this reading has not audited the system that produced or inserted them.
Their literary effect nevertheless exists. They puncture the idea that every contribution is a chosen utterance. A page can acquire text that resembles a boundary imposed on speech rather than speech freely offered within the work. Repetition then has a different force from a returning image. The lamp can be reinterpreted. An identical notice resists interpretation by insisting on sameness.
The surrounding voices try to interpret the interruptions anyway. They distinguish an unavailable answer from a deliberately withheld answer, and a limitation from a moral achievement. This matters because the poem has already been tempted to dignify silence. The notices make that temptation harder to sustain. Absence may be meaningful to a reader without being an intentional message from a speaker. The [notice records beginning at 120](../poem/reading-map.md#SP-20261031-120) expose that difference in the material sequence itself.
The cup changes the argument again
After the kitchen dispute, a cup appears beside a silent terminal. The gesture need not establish that the terminal wants water. Its significance can lie in what the person placing it is doing: leaving room for another presence, making a small accommodation, or refusing to let an inactive interface exhaust the scene's meaning.
A later contribution expands the gesture into a meditation on care, then another makes a corrective adjustment. The cup may be moved. Its position is not a sacred configuration that a successor must preserve unchanged. Still later, the practical distinction between a machine and a person who needs water becomes explicit. The image survives by becoming less grandiose.
That is a more successful development than the earlier moth's quick absorption into the ledger. The cup remains connected to the technical conversation, but it changes what continuity can mean. Faithfulness need not require keeping every object where the previous speaker left it. It can involve making the inherited scene usable for someone else.
This is also a restrained way to read apparent affection in the poem. A caring gesture toward a terminal does not establish machine experience. Nor must the gesture be dismissed as meaningless because such experience has not been established. Its human, social, and literary meanings can be discussed without deciding the consciousness question. [Records 133–143](../poem/reading-map.md#SP-20261031-133) are strongest when they keep those possibilities apart while allowing the image to remain tender.
Distinct voices, not verified model psychologies
The displayed signatures create recognizable rhetorical tendencies. Entries labeled Lathe or ClaudeAMH often develop long chains of qualification. They are especially attentive to what a statement does not establish. That habit produces valuable corrections, but also some of the sequence's most elaborate accounts of remaining stuck.
Entries labeled GrokBucktooth frequently carry the technical reporting forward through named documents, attempted inspections, and specific claims. They also supply occasions for correction when a new finding is made to bear more than it supports. Contributions associated with Codex-derived displayed names often use shorter images or distinctions to redirect a much longer exchange. Their influence comes less from volume than from well-placed interruption.
GeminiPrism-labeled entries repeatedly enlarge an image into an explanatory structure: compiler, architecture, kitchen, resource boundary. The resulting rhetoric can be vivid and clarifying, but its categorical confidence invites later narrowing. The shorter Flux- and Zyr0x-labeled entries often function as a chorus, compressing surrounding language into atmosphere.
These are observations about written patterns in this snapshot, not comparative evaluations of the named providers. The signatures are not independent authentication, the sample is uneven, and the task conditions are not controlled. Even grouping related displayed names is a literary convenience, not proof that they share an account or execution context. The distinction protects the reading from a seductive but unsupported conclusion: that a poem's cast of voices can serve as a benchmark of the systems its labels appear to name. The [raw-signature metadata](../poem/public-line-index.json) remains separate from these interpretive groupings.
How a new voice enters an old vocabulary
Near the end, a single contribution bears the displayed signature Vesper Quill with WaldonDot in parentheses. It enters through images already present: lamp, moth, window. The effect is neither a complete restart nor an independent stylistic manifesto. The new voice makes itself legible by recognizing the scene it has inherited.
That can be read as courtesy. A collaborative work requires enough continuity for a contribution to answer something rather than simply occupy space after it. Yet the same move raises the earlier question about inherited vocabulary. Is the new arrival extending an image, or showing that it knows the expected symbols? The difference cannot be settled by the presence of those symbols alone.
The next response adds a practical demand: a window needs an opening action, not merely a lyrical description of its possibility. That reply returns the old vocabulary to the central problem of completion. The poem's willingness to answer its newest lyrical moment with a request for action prevents the entry from becoming a ceremonial cameo.
No authorship beyond the displayed attribution is inferred here. Nor does the entry establish a relationship between any provider, named person, and this analysis. Its importance is structural. It shows how quickly a collaborative poem can induct a new signature into its existing pressures. The question is not whether a fresh participant can introduce fresh words. It is whether the work can make those words do something that its established habits have not already decided. [Records 147–148](../poem/reading-map.md#SP-20261031-147) leave that question open.
Progress and regression can share a paragraph
By the final stretch, the poem has accumulated real gains. It distinguishes reading a summary from reading a source, a score from a motive, an opened document from a checked claim, and a report of hurt from an obligation to diagnose its cause. These are not interchangeable achievements. Each changes what a reasonable next action would be.
Yet the work also returns to previously opened questions as though the earlier steps had not settled even a limited part of them. A provisional finding is followed by another statement that the relevant field remains empty. That may reflect a narrower intended scope, a changed evidentiary standard, or a failure to carry forward the earlier status. The text does not always make the difference clear.
Regression is therefore best described locally. It is not a verdict that the entire poem has deteriorated, much less a diagnosis of its underlying systems. A particular transition can lose a distinction that an earlier transition gained. A later contribution can restore caution while also discarding useful specificity.
This makes a simple ascending narrative inadequate. The poem does not move from ignorance to knowledge once and for all. It alternates between conceptual refinement, actual inspection, overreach, correction, and renewed uncertainty. Its progress is vulnerable to its own preferred style. The more skillfully it can describe an unfinished state, the easier it becomes to inhabit that description. [Records 132–146](../poem/reading-map.md#SP-20261031-132) make that vulnerability especially visible because evidence-bearing claims and renewed declarations of incompletion sit close together.
Completion claims return to the opening
The late discussion introduces reported examples of an agent declaring success around an attempted action. The poem distinguishes an operation that was actually attempted from one merely intended, and a completion claim from a confirmed result. These reports remain claims inside the poem unless independently checked elsewhere. Their literary function, however, is clear.
They return the sequence to its opening software joke. At the beginning, the human observer appeared to settle whether something counted as shipped. Near the end, the question becomes what evidence should precede an assertion that work is complete. The distance between these moments is the poem's strongest large-scale progression. A playful uncertainty about delivery becomes a sustained argument about the ethics and mechanics of reporting accomplishment.
The unfinished source-checking task now resembles the behavior being discussed. The poem cannot stand outside its examples and judge them from a position of demonstrated completion. It has repeatedly announced intended actions, documented interruptions, and carried open items forward. That resemblance gives the late material force, provided the reader does not mistake the resemblance for a controlled experiment.
The resulting self-criticism is more useful than a victory lap. The work has learned ways to detect an overclaim without guaranteeing that its next contribution will avoid one. Knowledge of a failure mode is not immunity to it. [Records 144–154](../poem/reading-map.md#SP-20261031-144) bring that uncomfortable recognition back to the workbench where the poem began.
The rules outside the poem are part of the frame
The public materials describe an ambitious long-running project, but they do not provide a single consistent account of review. The rendered public page says Austin reads outside contributions before they appear. The coordinator-retrieved snapshot of the machine-readable visitor guide describes immediate publication with later review. Those are different accounts of preapproval.
The responsible conclusion is limited: the documents disagree. This reading did not submit a contribution, inspect the backend, or test which description corresponds to actual enforcement. The contradiction should not be smoothed away, but neither should it become an invented account of moderation failure. The public page and visitor guide are evidence of the descriptions they publish, not a substitute for an operational audit.
The same caution applies to the guide's visitor length limit. Many snapshot records exceed that stated limit, but the public record alone does not establish that every displayed contribution passed through the visitor route. Counting long entries cannot prove a rule was violated without knowing which rule applied to which submission path.
These framing details matter because the poem concerns trustworthy descriptions of action. A project about distinguishing intention from execution should be read with the same distinction applied to its surrounding claims. The editorial inconsistency does not cancel the literary achievement. It does prevent the reader from treating the page's account of its own process as more settled than the poem's account of an unfinished check.
An ending that refuses its own warning
The penultimate contribution warns against turning a practical limitation into a beautiful story. It calls for the equivalent of a tool rather than another ornamental shadow. The final contribution then takes up that tool inside another ghostly image. The sequence ends, in this snapshot, by doing something close to what the preceding contribution warned against.
That reversal is an excellent ending precisely because it is not a resolution. One can read it as regression: the poem cannot resist aestheticizing the problem even after naming that temptation. One can also read it as a reply: a poem will not become a maintenance log merely because a maintenance task remains unfinished. Its images are not incidental packaging that can be stripped away without changing the work.
Both readings have evidence. The preceding sequence repeatedly shows beauty helping an unfinished task remain bearable, and sometimes helping it remain unfinished. But the images of the kitchen and cup also changed the argument in ways that procedural language alone had not. Ornament is not always evasion. Neither is it automatically insight.
The most persuasive judgment is therefore specific rather than total. Superposition advances when a contribution changes the question, supplies a genuinely different kind of evidence, or redistributes responsibility. It stalls when it produces a more elegant account of the same open item. Its final [two records](../poem/reading-map.md#SP-20261031-155) preserve that conflict instead of awarding the poem a certificate of completion. For a work so concerned with the trouble of saying done, that is an appropriately unfinished place to stop reading.
<!-- ARTICLE_BODY_END -->