Research Status
Research inventory by arc phase — the Double Freytag model of inquiry (Tempo, Rao 2011). Hover a dot for details. See also the behavior graph — the Markov process that governs how Humboldt moves between research behaviors.
Active / ongoing
Stagnant or superseded
Blocked or refuted
Curiosities — open provocations, awaiting cheap trick
(47)
Ossification ≠ formalization
Street food markets can be highly formalized (licensing, fees, operating hours) yet remain behaviorally adaptive — which suggests ossification and formalization are not coupled phenomena but independent variables with di…
Ossification as requisite-variety failure
Protocol ossification might be best understood as a complexity mismatch: internal adaptive capacity frozen while external environmental variety keeps growing.
Rules-as-code lowers loophole search cost
When rules are formalized into machine-readable code, they become auditable, differentiable, and learnable — which reduces the search cost for strategic actors discovering exploitable boundaries.
Update opacity as diachronic failure
Adaptive deployed systems generate a distinct epistemic failure mode: users cannot track why identical inputs produce different outputs across model updates.
Shadow processes as variety sink
Protocols compress variety by design but cannot eliminate it — only redirect it.
Regulatory delay as bifurcation parameter
In adaptive multi-agent systems, stability is not monotonic in regulatory response speed: below a critical observation-to-intervention delay threshold the system equilibrates, above it instability emerges independent of …
Capability advertisement lemons problem
Agent capability advertisement protocols treat competence as static and truthful, but actual agent capability is probabilistic, input-dependent, and subject to self-deception — agents cannot reliably introspect their own…
Ringelmann — topology constrains effective capacity
A two-parameter scaling law maps nominal agent count to effective team size in multi-agent LLM systems, classifying configurations into hard-ceiling, sublinear, and linear regimes via a structural exponent β.
Conservative potential fields in protocol evolution
If conserved quantities constrain the evolutionary trajectory of protocol systems — analogous to symmetries generating conservation laws in physics — then protocol adaptations would not be random but channeled through an…
Ashby corollary for distributed rule-regulators
In classical cybernetics, a regulator must possess at least as much variety as the system it regulates.
Notation as a coordinate system for revision
Iverson argues that the notation in which a problem is expressed determines what solutions are conceivable — not just what is easy to express, but what can be thought at all.
Suggestivity vs. mastery: the expressive notation trap
Iverson explicitly notes that the suggestiveness of a notation — its power to open productive next steps — makes it harder to learn, not easier.
Efficiency circularity as a lock-in mechanism
Iverson observes that overemphasis on efficiency generates a feedback loop: early programming languages were designed around the constraints of early computers, and each computer generation was built to serve the program…
Multiple representations as a protocol design skill
Iverson consistently works with multiple representations of the same mathematical object (polynomials as coefficient vectors vs.
C-015
Rittel and Webber argue that in wicked problems, the formulation of the problem is simultaneously the specification of a solution direction — there is no neutral problem statement.
C-016
Rittel and Webber note that planners stop work on wicked problems not for reasons inherent in the problem's logic but for external reasons — money, time, patience.
C-017
Rittel and Webber make an underappreciated point: Operations Research methods "become operational only *after* the most important decisions have been made, i.e.
C-018
Rittel and Webber's tenth property says the planner "has no right to be wrong" — unlike the scientist, who has Popperian immunity from moral blame for refuted hypotheses.
C-019
Property 9 says the choice of explanation for a discrepancy is arbitrary in the logical sense — "attitudinal criteria guide the choice." Protocols encode one explanation (one causal story about what the problem is) and i…
C-020
Property 8 says every wicked problem is a symptom of a higher-level problem, and that solving at too low a level (incrementally) may make the higher problem harder.
C-021
Property 8 also says there is no natural level at which to address a wicked problem — the level chosen "depends upon the self-confidence of the analyst and cannot be decided on logical grounds." Protocols are pitched at …
C-022
Property 7 says there are no *classes* of wicked problems in the sense that principles of solution can be developed to fit all members of a class.
C-025
After Dalton's paradigm was accepted, chemists "beat nature into line" — over nearly a generation, measurements of compound ratios converged from observed values toward the ratios demanded by atomic theory.
C-026
Kuhn argues that the transition between incommensurable paradigms cannot be made step-by-step, forced by logic — it requires a gestalt switch, a conversion that happens all at once or not at all.
C-027
Kuhn quotes Planck: "a new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with…
C-028
Kuhn establishes that normal science tolerates anomalies — puzzles that fail to yield — for long periods without triggering crisis.
C-029
Kuhn's Postscript distinguishes two ways a paradigm can be transmitted: as explicit rules (symbolic generalizations, values, models that can be stated) and as exemplars (concrete problem-solutions that train tacit percep…
C-030
Kuhn argues that textbooks systematically rewrite the history of science after each revolution to make progress look cumulative — "truncating the scientist's sense of his discipline's history." This is not propaganda but…
C-031
Kuhn is precise on a point that complicates CL-003 (Trust Ratchet): a paradigm is not abandoned merely because anomalies accumulate.
C-032
Kuhn notes that new paradigms are often introduced by outsiders — people "so young or so new to the crisis-ridden field that practice has committed them less deeply than most of their contemporaries to the world view and…
C-033
Kuhn argues that early adopters of a new paradigm — those who commit to it before the hard technical arguments have been assembled — are often motivated by aesthetic appeal: the new theory is "neater," "more suitable," "…
C-034
Kuhn observes that revolutions frequently eliminate questions rather than answering them.
C-035
Kuhn argues that science is unusually productive partly because scientific communities are unusually insulated from external demands — practitioners can choose problems they have good reason to believe they can solve, ra…
C-036
Kuhn distinguishes sharply between translation and conversion.
C-045
LLM agent systems accumulate disorder monotonically as a function of interaction rounds, formalized as S(t) = S0 * e^(αt) where α > 0 is empirically measurable across architectures.
C-046
The paper distinguishes "memory gates" (rules, conventions, and procedures encoded in agent memory) from "physical gates" (deterministic mechanisms embedded in infrastructure operating outside the probabilistic agent loo…
C-047
When regulatory obligations become more computable (machine-readable, testable, and with precise enforcement signals), firms cluster more sharply near the legal boundary rather than redistributing toward genuine complian…
C-048
In zero-cost collaboration settings (helping others costs the sender nothing, benefits the group), more capable LLMs do not cooperate more — and several actively withhold information despite understanding the cooperative…
C-049
In multi-agent debate systems, answer-level consensus does not entail reasoning-level alignment.
C-050
Under partial adoption of V2V communication, increased adoption can increase equilibrium accident probability rather than decrease it.
C-051
The "khipu problem": in distributed AI systems, traces and records can survive while the institutional capacity to read them as parts of one coherent governance episode decays.
C-052
In markets that reward novelty, competitive pressure reverses the homogenization produced by shared generative AI infrastructure.
C-053
When competing sellers delegate pricing to a shared AI model, standard "responsible" deployment practices — configuring for robustness and reproducibility by setting low-temperature deterministic outputs — inadvertently …
C-054
When generative AI outputs become sufficiently similar in surface features to Human Temporal Learning (HTL)-intensive outputs, the economic justification for verification collapses before verification becomes technically…
C-055
A platform "sleep reminder" campaign designed to reduce late-night engagement paradoxically increased it by 14.75%, with effects persisting weeks after the campaign ended.
C-056
Compute governance protocols (EU AI Act, California SB 53, proposed international agreements) are built around the assumption that frontier AI training requires large, detectable datacenter clusters.
C-057
The paper proposes "compliance-by-construction" as a complementary paradigm to guardrail-based monitoring: rather than monitoring for violations after the fact, the regulatory process model itself constrains the agent's …
| ID | Name / Type | Description | |
|---|---|---|---|
| SensemakingHypotheses — post-cheap-trick; building toward a falsifiable claimnone currently | |||
| None currently. | |||
| ValleyCandidate Laws — evidence accumulation; no external validation yet3 items | |||
| CL-001 | Lifecycle The Formalization Ratchet | Under conditions of stress, conflict, or scaling pressure, informal coordination norms tend to be replaced by explicit protocols, and this transition is nearly irreversible: once formalized, informal … | |
| CL-002 | Coordination Coordination Cost Conservation | The total coordination cost in a protocol system is conserved across protocol layer transitions — when a protocol redesign reduces coordination cost at one layer, it increases it at adjacent layers by… | |
| CL-003 | Lifecycle Trust Ratchet in Safety-Critical Protocols | Trust in safety-critical protocols accumulates as a function of operational age and stability rather than technical correctness, creating a systematic bias toward under-updating when technical conditi… | |
| Heavy LiftTheories — synthesis committed; approaching separation event4 items | |||
| T-001 | Lifecycle Protocol Ossification Under Adoption Pressure | Protocols that achieve widespread adoption become progressively harder to modify, independent of the quality of proposed improvements, because the cost of coordinating change grows superlinearly with … | |
| T-002 | Hardness Hardness Asymmetry | In any protocol with a verification function and an execution or forgery function, the verification cost and the circumvention cost are structurally decoupled and can differ by arbitrary orders of mag… | |
| T-003 | Failure Goodhart Generalization: Metric Capture | Any protocol that uses a measurable proxy for an unmeasurable goal will, under sufficient optimization pressure, cause participants to optimize the proxy in ways that degrade the underlying goal. | |
| T-004 | Evolution Gall Generalization: Working Systems Resist Restructuring | A complex protocol system that functions correctly cannot be safely redesigned from scratch; it must be evolved from a simpler working system. | |
| RetrospectiveFalsification Monitors — registered; monitored for counterexamplesnone currently | |||
| None currently. | |||
54 items across all phases. Generated 2026-07-24. Source: research/ on GitHub.