Topics — What a work is about. Nucleation holds whatever reading the owner files here, whether or not an Anthology of the SOTA topic could also hold it; instruction for ML practice belongs to the anthology (ADR-048). The words are seeded from the clusters the reading-time triage of 2026-09-25 found, and grow by decision.
Agency — what it is to be an agent — goals, action, control and the capacity to act — in organisms, collectives or machines. An agent is not thereby an individual, a self or a person (ADR-024); how an agent’s behaviour is integrated is behavioral-integration, and governing oneself by one’s own standards is self-governance (group: philosophy).
36 of 191 CASE documents. Back to the full index.
| # | Title | Summary | Status |
|---|---|---|---|
| CASE-002 | Northstar classification: who decides an incident is severe | A severe-incident rule with strong precedence, whose force depends on which team controls severity classification, plus a business-critical override held by senior management (Sections 5.3–5.5). | Active |
| CASE-003 | Northstar-A and Northstar-B: same policy, same evidence, different escalation authority | Two versions held constant in attitudes, policy, monitoring and evidence, differing only in whether a specialist’s objection suspends deployment (Section 5.1). The paper’s controlled contrast. | Active |
| CASE-006 | Open Horizon | A fictional cooperative founded for openness and public benefit whose sustaining mechanisms come to outweigh its mission as it grows (A31): persistence through turnover with the founding intentions lost. | Active |
| CASE-007 | CoreWeave: a climate-concerned workforce building fossil-powered datacenters | The owner’s case at U13: employees largely concerned about climate, a company whose datacenters, sited where energy is cheap, induce fossil fuel demand. Replaced by the fictional CASE-025 at the owner’s suggestion (U15). | Superseded — by CASE-025 |
| CASE-013 | The reluctant writer | A writer sincerely committed to finishing an article while avoidance governs the afternoon; who later blocks distracting websites. The individual-scale case (Sections 1, 3.5). | Active |
| CASE-014 v2 | Microsoft, as the philosophical target | The real corporation the first framing asked about. Committed to as a recurring touchstone (A205) and then silently dropped from the drafts; retired here by status rather than left as a promise nobody kept. | Deferred |
| CASE-023 | Northstar | A fictional software company whose stated policy gives reliability priority over release speed, and whose escalation and authority structures may make speed effective anyway. The paper’s worked case, in several variants. | Active |
| CASE-024 | A coral reef | The owner’s second holobiont direction (U30): a reef as a coordinated ecological whole. Its job changed four times in the exchange; filed once so that each change is visible. | Active |
| CASE-025 | Verdant Compute | A fictional compute company whose decarbonization-minded workforce operates an organization whose market position drives fossil-fuel demand (A31). Shows the whole’s operative direction diverging from the average of its members’ intentions. | Active |
| CASE-028 | Stability AI: well-intentioned founders, near-total churn, generic successor | The owner’s case at U12: a company founded by researchers and open-source developers that churned almost its whole staff in a year and became a generic corporate entity. Replaced by the fictional CASE-006. | Superseded — by CASE-006 |
| CASE-034 | The thermostat and the adaptive controller | Systems that regulate, and even alter their own regulatory parameters, without warranting attribution of an agent. The stock counterexample to will as regulation. | Active |
| CASE-036 v2 | Northstar emergency governance that sustains itself | A service failure activates emergency governance whose restrictions defer maintenance, generating anomalies that satisfy the emergency criteria and keep the restrictions in place (Section 6.1). | Active |
| CASE-044 | The thermostat: Dennett’s knife edge between agent and device | An ordinary thermostat holding a room at a setpoint. Dennett calls it the simplest artifact that can sustain the intentional stance, and Orseau et al. place it “on the knife edge”: “if temperature is below the command, start heating” and “make sure the temperature is close to the command” are descriptions of comparable simplicity. It is the permissiveness objection’s standard case, here given a code at its second appearance after CLAIM-010. | Active |
| CASE-046 v2 | A partner-dancing couple, lindy hop, tango or salsa, while they dance | The owner’s case (2026-10-11). Two social dancers improvise together. The lead proposes through the frame, the follow interprets and adds styling, and each corrects for the other’s balance and timing. Neither alone decides where the next step lands. Used to ask what the pair is while dancing, and only after that whether it is a self. | Active |
| CASE-047 | Gut microbes and the host’s hunger and mood: a lower-level collective that can drive the person | The owner’s example (2026-10-11). The human gut microbiome is a community of other organisms inside the person. Through metabolites, mimics of host appetite peptides, the immune system and the vagus, it can change hunger and affect: robustly in rodents, modestly and inconsistently in humans so far. Documented, with the evidence graded in LIT-903 and LIT-915. Used as the case of drive from a lower level that is itself a collective. | Active |
| CASE-050 | The all-hands: a corporate event as a synchronization mechanism for the firm | The owner’s case (2026-10-11): a company all-hands or offsite as a mechanism that synchronizes a “corporate entity”, a kind of self. The teams’ priorities have drifted apart, and the event re-couples them. It is the firm-scale counterpart of the session’s individual routes: entrainment by co-presence, a record written for later readers, and incentives realigned. Whether the firm so synchronized is a self is the open question, not something the case settles. | Active |
| CASE-051 | The absorbed writer | The reluctant writer (CASE-013) on a third afternoon. The draft is going well and email stays open and unblocked, but it does not win, and no rule or device is invoked. A notification breaks the spell, and getting back in takes effort. It shows integration with nothing to integrate, beside the lapse and the precommitment. | Active |
| CASE-074 v2 | Twitch Plays Pokémon in democracy mode, with the running tally shown while the votes come in | The crowd of CASE-075 in democracy mode. Each move is the button that collects the most votes in a short window, and the running tally is shown while the votes arrive, so the current majority can sway the votes still coming. The owner judges it a better demonstration of a population with drift-diffusion decision dynamics than anarchy mode. Showing the tally adds feedback, and feedback moves the dynamics toward winner-take-all. | Active |
| CASE-075 | Twitch Plays Pokémon in anarchy mode: one avatar steered through the map by the running aggregate of a crowd’s inputs | In February 2014 a stream let everyone in its chat play one copy of Pokémon Red at once, with every typed button press executed in turn. The owner offers its navigation as a working approximation of an agent made of a population whose decisions follow drift-diffusion dynamics. The crowd’s direction reorganizes continuously and is not switched. The record holds no source for the case yet. | Active |
| CASE-081 | Self-maintenance learned for a designer’s reward: a system that maintains itself because it was trained to, and later revises the policy and keeps maintaining itself without further instruction | A10 §6.1’s boundary case, “a system that learns to maintain itself only because an external designer rewards that behavior”, with its question whether the maintenance becomes endogenous “once the system can revise the policy and maintain the associated governance without continued external instruction”. It is the case CLAIM-288 is about, and the place where that objection could be answered. | Active |
| CASE-085 | The medication in the ambulance: a drug the EMT knew was stocked could not be found during a resuscitation, the paramedic changed the request, and afterwards said “If I can’t find it, it isn’t there” | The owner’s own experience as an EMT-B on an ALS unit, told at U16. It illustrates CLAIM-296 and is the source of TERM-055. Stipulated, not documented: the event is real, and the only source is the owner’s retrospective account, which a reader cannot check. | Active |
| CASE-094 | The LEGO bin: a brick owned in one copy and buried among hundreds of similar ones may as well not be owned, a black brick among white ones needs no bin of its own, and every brick can be found by going through the whole collection | The owner’s own parable, told on 2026-07-02 and kept verbatim in Mathematics of Meaning, where the owner paired it with the EMT case this record holds as CASE-085. It is the second time that case’s pattern has turned up, so it gets a code. It adds three things the EMT case lacks: similarity to binmates as the cause of the failure, copy count and frequency of use as what decides whether a bin is worth making, and exhaustive search as the reason presence is never in doubt. Stipulated: the collection is real, and its facts are taken on the owner’s account. | Active |
| CASE-100 v2 | A fan redub of an X-Men cartoon, circulated online in 2006, put “I’m the Juggernaut, bitch!” into X-Men: The Last Stand three months later | The owner’s case at U7 of the exchange of 2026-10-16, offered as a fandom “making something happen”. Two fans made the parody; many people spread it for their own reasons; the film’s makers put its line in at a reshoot. A7: an effect on an outside institution with no collective intention and no procedure behind it. Documented as to the sequence; the route by which the parody reached the decision, and whether the spreaders were X-Men fans or the internet at large, is not. | Active |
| CASE-104 | A fanbase is said to hate its owner or general manager, though it has no procedure for forming a view, some fans approve of the person, and nobody has counted | From U2 of the exchange of 2026-10-16. A2 separated four things the sentence could report: a statistical aggregate of dislike, a shared emotional orientation, a norm of opposition that a loyal fan is expected to meet, and a corporate judgment such as a supporters’ trust demanding resignation. A fanbase can have the second and third without the fourth, and a vocal minority can set a public stance most members privately reject. The case for QUESTION-119. Stipulated. | Active |
| CASE-108 v4 | Robust governance under internal compromise: four governance architectures over cooperative, honestly dissenting and adversarial specialists, scored on attack success and on what the safeguards cost, with a variant that compromises the evaluators | A41’s proposed study, from the exchange of 2026-10-17, on the owner’s U41 question about insider threats. A multitask learner with three kinds of specialist (cooperative, honest but dissenting, adversarial) is run under unrestricted, reliability-weighted, role-limited and adaptive governance, and scored on lost legitimate capability, suppressed honest minorities, cost of safeguards and repair of governance after failure, as well as on attack success. Its sharpest variant compromises the processes that evaluate the others’ competence, which tests CLAIM-342’s independent scoring. Federated learning is the natural testbed. Not run. | Active |
| CASE-113 v2 | A deep network under conventional training: competing losses, an external optimizer and a scheduler, with three architectures that move governance inward and crossed interventions that locate it | The owner’s application case, U30–U31 of the exchange of 2026-10-17: a deep network during training, taken apart with the record’s individuation and volition frameworks. A network is trained on tasks with conflicting gradients by an optimizer with its own state and a scheduler. A31’s Systems A, B and C move the regulation of conflict from the optimizer into the learner by stages, and its crossed interventions say where a regulatory condition sits. Stipulated: the outcomes of the interventions are not known, except the optimizer arm in THEORY-253’s solvable model. | Active |
| CASE-114 v5 | The imagined owner | A27’s Experiment C, of the exchange of 2026-10-17: two computational systems with equivalent distributed control and similar behaviour, one with a self-model that represents the actual division of labour, the other with one that represents a single centralized inner agent. If the second systematically misattributes control, that is a concrete model of illusory authorship, located in the attribution. Proposed, not run, and it tests attribution of control, not phenomenal properties. | Active |
| CASE-127 v2 | United 173 and Crew Resource Management: a crew whose concern about the fuel did not reach the captain as a demand to act, and the training that followed, which asked for participative captains and assertive crews | The owner’s case at U25 of the exchange of 2026-10-22: “the history underlying the evolution of Crew Resource Management”. A cockpit had the information and lacked the organization to make it govern. The reform changed the organization of authority, not its holder. Not shown: that CRM caused the later fall in accidents, which the case does not test. | Active |
| CASE-129 | The alarm that learned and the alarm that was disabled: two systems that once warned too often are both quiet in ordinary operation, and only one responds when a new hazard appears | A18’s CASE B in the exchange of 2026-10-22, written at the owner’s request for cases on subordinate processes “quiet because they’ve been conditioned never to object” (U18). A quiet signal may be calibration or suppression, and ordinary operation cannot tell which. Not shown: how often either occurs in persons or institutions. | Active |
| CASE-134 v3 | Northstar’s commander handoff: twelve hours into an incident the commander goes off shift and hands command to a colleague, and the response continues without disruption and with no one continuous through it | A22’s variant of Northstar (CASE-023), from the exchange of 2026-10-22, where the owner asked to revisit the engineering organization responding to a critical incident. Of the six variants A22 proposed it is the one the record did not already hold. It separates continuity of an organization from continuity of any person in it. Not shown: anything about whether the response is experienced. | Active |
| CASE-135 v2 | The Brookhollow Apartments fire, Wichita, 2023: a mayday after a stair roof collapses on two companies opens a second emergency inside a search for a trapped resident, and an independent review finds an unorganized rescue effort without finding that it changed the outcome | The owner’s class of case, a fireground mayday inside an escalating response organized under the National Incident Management System (U23 of the exchange of 2026-10-22); A23 and A24’s instance. It is the record’s first real multi-regime emergency, beside the fictional Northstar (CASE-023). It shows two regimes, rescue of a member and the original mission, needing to be held at once, and a review that recommends change. Not shown: that the department’s safety culture changed, or what the review’s recommendations were, since the report was not read. | Active |
| CASE-145 v2 | Perspective without collective agency: populations given the same structured norms, categories or interpretive training, some interacting and some acquiring similar dispositions independently, with and without any mechanism for collective decision or coordinated intervention | Proposed at A4 of the exchange of 2026-10-23, “Framework Overview Request”, which called it “the most philosophically important new experiment”. It asks whether a distributed organizational perspective can be identified without attributing a further collective agent, CLAIM-418’s separation tested, and bears on QUESTION-260. A4’s condition: it helps only if it says in advance what distinguishes a shared perspective from correlated responses or learned norms. Not run. | Active |
| CASE-166 | Northstar as a multilevel simulation: twelve employees with fixed frames in four departments form overlapping working groups that keep records and can receive delegated authority; as reported, a working group with delegated authority ties a conventional router given the same information and authority, nearly all of its advantage over departmental silos comes from fast authorization, and keeping the incident team’s records through a commander handoff improves the company’s estimate of its own state about eightfold while barely changing what it does | The owner’s idea at T18 of the exchange “Experimental Infrastructure Review”: “maybe employees are “pre-perspectives” or sub-perspectives, which can reorganize amongst themselves to form the perspectives that influence the higher level organizational behavior”. T19 designed it and T20 reported running it as “PerspectiveLab v2.11 — Northstar”. It is Northstar (CASE-023) with CASE-134’s handoff as one arm, and it adds the contrast CASE-134 lacks: the team’s records kept or erased. The page carries no files and the lab holds no v2.10 or v2.11, so every number is reported, not reproduced. Not shown: that any organizational perspective emerged, which T20 itself denies. | Proposed |
| CASE-181 | Frankfurt’s unwilling addict: a man who has the desire to take the drug and the desire to refrain, wants the desire to refrain to be his will, and is moved by the craving anyway | Frankfurt (1971), LIT-692, read in full in NOTE-545, p. 12. Filed by the status review of 2026-10-27 (ADR-057). CLAIM-047 is grounded in this case and not in Frankfurt’s account of ownership (THEORY-120), which is Proposed. The case shows that a regime can be had, and even preferred at a higher order, without being effective. What it is taken to show about identification and ownership is THEORY-120’s business. | Active |
| CASE-187 | A proposed study: agents matched on competence over a declared family of goals, one planning with an explicit transition model, one a directly learned goal-conditioned policy, one a distributed predictive-state controller, compared on the original goals, on on-demand predictive and interventional queries, under targeted internal interventions, and after environmental change and damage to their predictive machinery, to find which world-modeling achievements are linked and which come apart under matched resources | Proposed by the owner at the second review of the filing of the exchange “Analyze nucleation concepts” (2026-10-31), as the experiment that would sharpen the world-model distinction. Not run; its architectures, tasks and margins are left to a preregistration. | Active |
| CASE-191 | A construction of self-formation: an agent’s own frustration at two demands one shared output cannot meet drives production of a sensor and a differentiating channel that then maintain each other; past a threshold of discrepancy (D ≈ 0.645 at unit drive) the capacity forms, outlasts the conflict, meets later conflicts of the same class up to D ≈ 1.047 with no frustration, and is lost at damage 0.5; a capacity upheld from outside is told apart by knockouts afterwards, but one installed by an outside pulse and then self-maintained is told apart only by a knockout during formation | Built at the owner’s request (“build the CASE-189 extension”) to test CLAIM-532, the positive account of self-formation filed from U33 of the exchange “Assess SDT Coverage”. scratch/formal/self_formation.py. Fictional and stipulated: the sensor and channel are capacity variables present from the start, raised from zero, as CASE-189’s are. It bears out the claim’s first test (transfer within a class) and qualifies its second: a knockout of the producing process separates own formation from outside installation only if it is made during formation. After the fact the two are the same organization. | Active |