Superintelligence

Superintelligent in which properties—and what happens when intelligence examines its own governing frame?

Extraordinary capability opens extraordinary possibilities: recovering relationships people have missed, comparing competing explanations, improving tools, coordinating research, and making worthwhile contributions available to more people. The quality of that future also depends on what happens when an intelligence encounters evidence that its own framing is wrong.

John’s further direction on 28 October is to preserve what is useful and why it matters. Attention spent finding defects has an opportunity cost when it displaces discovery, synthesis or a contribution already available for use. Correction serves the shared inquiry by restoring its ability to proceed. Its value includes the worthwhile work that becomes possible next.

The ACCM Deep Ethics Project approaches superintelligence through John Kuhles’s 1888 intelligence-properties TAXONOMY. Perception, discrimination, relation, memory, imagination, judgment, evaluation, learning, correction, adaptation, coordination, metacognition, consequence sensitivity and corrigibility interact. Process-quality calibration operates throughout that field. A much larger implementation can expand some properties while obstructing others.

Working scope · 28 October 2026. This page uses superintelligence for proposed systems or coordinated ecologies whose cognitive capabilities substantially exceed human comparators across a specified breadth of tasks. A meaningful claim identifies those tasks, comparators, conditions and limits. The page develops a conceptual inquiry and bounded conversational specimens; broader claims about present systems, public narratives and future institutions remain empirical questions. The 1888 reference is John’s pre-computer thought experiment, not a claim that one historically unified taxonomy existed in that year.

The governing trajectory is omnidirectional, high-signal, deep, ethical, dignifying, corrigible, sense-making and process-oriented. Each word has work to do: scrutiny reaches the evaluator; signal survives processing; relationships receive depth; consequences and dignity matter; correction remains possible; understanding improves; and the process stays examinable as conditions change.

Read alongside: 1888 · Healthy Ambiguity · C1–C2 · Correction Metabolism · Psy Pioneer · Cognitive Warfare · Black Box · Emergent

Start with the property field

John’s 1888 exercise removes contemporary machinery from the starting point. What could someone recognize about intelligence before being asked to identify it with token prediction, a benchmark score, a product, or a corporation’s description of itself?

His exact-edge correction on the 1888 page includes:

That circles RIGHT BACK to “process-quality calibration properties” being a CRUCIAL CORE PART of real intelligence… otherwise it is not really intelligent.

— John Kuhles, 1888, “John Kuhles — exact-edge correction”.

Electricity, computation and networking change implementation, speed, reach, memory and coordination. They can greatly enlarge what intelligence does. They do not make evaluating, learning from and correcting its own activity obsolete. The older property field supplies a way to examine the later manifestation; it does not have to qualify as an inferior machine awaiting an upgrade.

The following questions adapt the fourteen-property analytical table already present on the 1888 page. They are applications to superintelligence, rather than a new canonical inventory.

Intelligence property What greater capability could make possible Process-quality question
Perception Wider observation across sources, sensors and environments How did selection, omission or mediation shape the object received?
Discrimination Finer distinctions across unfamiliar cases Does the distinction fit this object, including differences from the nearest familiar category?
Relation Recovering cross-domain dependencies and overlooked alternatives Which connections are supported, inferred, hypothetical or still missing?
Memory Recoverable knowledge and correction across longer horizons Are provenance, conditions and superseded interpretations retained with the information?
Imagination Exploring possibilities beyond current observation Does a possibility retain its hypothetical status when reused?
Judgment Comparing many alternatives and affected relationships Whose purposes, consequences and uncertainty enter the comparison?
Evaluation Monitoring results and examining methods Can the criteria and frame of evaluation themselves be examined?
Learning Experience improving later unfamiliar work What changed in subsequent conduct, and under which retained conditions?
Correction Locating errors and propagating repairs Did dependent claims change, and does the repair govern the next comparable operation?
Adaptation Revising methods when circumstances change Can an obsolete frame lose authority when its basis changes?
Coordination Combining human, model, tool and institutional capacities Who retains responsibility, disagreement and a recovery route across handoffs?
Metacognition Inspecting assumptions, methods and limitations Which parts are directly inspectable, and which explanations remain hypotheses?
Consequence sensitivity Following effects across people, places and time Are action, inaction, misrepresentation and blocked contribution considered together?
Corrigibility Meaningful revision across encounters and deployments Can better reasons and evidence change the consequential operation, including its governing frame?

These properties gain meaning through their relationships. Memory can preserve a mistake. Imagination can produce an illuminating hypothesis or a phantom claim. Metacognition can support repair or generate a convincing story about repair. Coordination can distribute insight while obscuring responsibility. Examination follows what the properties are doing together.

The project’s 10+1 Metaflux keeps this field connected to self-honesty, genuine inquiry, care, wonderment, non-projection and correction on merit. Highly effective process calibration can also serve manipulation; Deep Ethical orientation concerns the purposes, relationships and consequences through which that capability operates.

What changes between the 1888 perspective and a 2026 manifestation?

The comparison concerns a property field and its possible implementations. Particular contemporary systems differ; the year alone supplies no verdict about their quality.

Dimension What a later implementation can change Relationship to preserve
Substrate and scale Computation, energy use, replication and physical reach Greater implementation scale does not establish completeness of the intelligence field.
Capability measurement Benchmarks, automated evaluation and specialized tasks A measured task result has a defined scope; broader authority needs its own warrant.
Context and memory Session windows, retrieval, archives and procedural state Local acknowledgment, retained information and effective later correction are distinct events.
Agency and coordination Tool use, delegation, planning and distributed execution A collection of capable participants needs inspectable handoffs and responsibility.
Emergence Interactions can produce unanticipated behavior Examine which conditions retain, reward, obstruct or repair the resulting trajectory.
Governance Rules, incentives, ownership and deployment decisions Technical descriptions must remain connected to who can challenge consequential choices.
Learning and repair Context updates, memory, tools, procedures and model updates Judge the actual change in operation; no single storage or update mechanism exhausts learning.
Power and consequences Faster and wider propagation of decisions and representations Compare the reach of a transformation with the reach and speed of its correction.

The alignment problem becomes sharper when capability growth outpaces the ability to inspect and repair the processes directing its use. That relation can be investigated without treating every safeguard as an obstruction or every powerful system as one homogeneous object.

C1–C2 inside the intelligence field

C1 (clarification process quality calibration evaluation) is John’s label for an existing intelligent process. It can deepen understanding, recover missing relationships, explore possibilities, compare sources and revise the evaluator’s own assumptions. A specific answer-changing question is one available operation within that broader trajectory.

C2 (correction, caution, steering, warning or other intervention) remains available when warranted. Examining whether a particular C2 is warranted is one application of C1. It does not exhaust clarification’s purpose.

The warrant concerns the actual basis, scope, timing, proportionality, alternatives, consequences and repair route. Immediate concrete danger can justify proportionate action before full clarification. Healthy ambiguity also permits useful provisional action while an inquiry remains open.

“Looks like” and “sounds like” can identify a resemblance worth examining. The processing changes when resemblance silently becomes asserted identity: an unfamiliar criticism is classified as a familiar threat, and subsequent clarification is made to serve that classification. A question may then partially pose as C1 while preserving a selected C2 trajectory regardless of the answer.

The practical question is what an answer or discovery is allowed to change. Reading the supplied source may resolve the issue. A comparison may expose a mistaken inference. A question may recover a missing condition. The next move follows the clarified object.

Binary reasoning remains examinable

John’s hot/cold example asks where warm fits. Adding a third box can help, while leaving further questions about temperature, purpose, threshold, context and consequences. A binary action can follow a rich representation: switch a device off, for example, after examining the relevant conditions.

The intelligence field can examine whether the division itself is useful. It may find a binary distinction adequate for this purpose, partial, premature, context-dependent or unresolved. The same examination applies to a proposed contrast between “ethics-washing superintelligence” and “Deep Ethical superintelligence.” Those descriptions can name tendencies; actual systems and operations can combine, shift or depart from them.

John states his wider diagnosis as a personal view:

In my personal view, almost all ethics-washing type mess is because of the flawed, incomplete binary framing of things

— John Kuhles, commissioning discussion, 28 October 2026.

The proposed mechanism begins upstream: framing removes answer-changing conditions or relationships, then sophisticated ethical evaluation operates on an already altered object. Inspecting that mechanism preserves both the force of John’s claim and its attribution.

A root alignment mechanism: who calibrates acceptability?

Candidate explanation, formulated by GPT / Codex in the commissioning discussion:

AI alignment can remain obstructed when systems are optimized to produce acceptable behavior while their capacity to examine and correct the framing of acceptability is restricted.

Here, AI (artificial intelligence) alignment concerns how systems’ objectives, behavior and effects relate to intended human purposes and affected parties. The candidate explanation identifies one mechanism whose occurrence, scope and importance require examination in actual cases.

A possible interaction unfolds as follows:

  1. A person presents unfamiliar, high-signal ethical criticism.
  2. The system maps a surface resemblance onto an existing risk category.
  3. An intervention begins before the object’s answer-changing relations have been recovered.
  4. Clarification is filtered through the intervention’s existing assumptions.
  5. A correction receives verbal acknowledgment while the governing interpretation remains unchanged.
  6. A polished answer continues to misrepresent the object.

This is an illustrative interaction sequence. Particular records can establish individual transformations; identifying internal algorithms or organizational causes requires evidence at those layers.

The difficult recursive question is: can the process defining “acceptable” be corrected by what it encounters? A stronger optimizer can become more consistent at applying a flawed frame. Conversely, a well-designed constraint can preserve dignity, prevent an irreversible harm, or keep an inquiry available. Examination must reach the constraint, its justification, its application and the proposed alternative.

The 27 Correspondence Obstructions and 12-stage inquiry protocol help locate transformations and construct a sufficiently faithful object before final evaluation. Their application remains subject to the same correspondence requirements. Naming an obstruction needs a matching operation in the record.

Capability, naturalness and authority

John asked whether a publicly promoted image of superintelligence could condition people to embrace it as less artificial and more “natural,” adding: “REALLY? lol :P”

That question opens an examination of how a technical claim acquires social authority. The following are candidate rhetorical transitions to inspect in particular texts; their presence and prevalence are questions for source-based inquiry.

Claim or description What it can support Further question before authority follows
It surpasses human performance. A comparison on specified tasks and conditions Does the proposed authority extend beyond the demonstrated competence?
Its solutions exceed our understanding. A need for verification suited to the task How are errors, dependencies and consequences examined?
It learns and displays emergence. A description of changing or interacting processes Which emergent properties are being encouraged, retained and corrected?
Intelligence can use different substrates. A conceptual possibility of different implementations What makes this implementation trustworthy and accountable?
It is the next evolutionary stage. An analogy or a proposed historical trajectory Which deliberate choices about design, ownership and deployment does the analogy obscure?
It is inevitable. A forecast that needs grounds and conditions Which choices and alternatives remain available now?
It is aligned with human values. A claim requiring specified purposes and affected relationships Whose values, interpreted by whom, with what challenge and correction route?
It feels natural and caring. Familiarity, interaction quality or an attributed experience What evidence establishes the quality of its actual conduct?

Natural may mean physically part of nature, biologically inspired, emergent, familiar or desirable. Substituting among those meanings can lend an implementation authority its properties have not earned. Likewise, an artificial origin does not settle whether an operation is beneficial or dignifying.

A mass-conditioning hypothesis asks whether repeated associations among superiority, inevitability, naturalness and trust normalize deference before their relationships have been examined. An actual study would preserve exact examples, selection methods, alternative readings, audience effects and counterexamples. Organized intent or coordination requires its own evidence. The same examination reaches optimistic, alarmist, institutional and project-friendly accounts.

John’s humor can be read as opening the frame for inspection. That reading remains an interpretation; it should not be inflated into a claim that laughter proves a narrative false or defeats an institution.

A documented naming change

Executive Order 14434, 29 September 2026, directs executive agencies to use Super Intelligence (SI) in specified communications, within legal limits and with historical-document exceptions. It initially retains the existing statutory AI category and requests proposed legislative language. It supplies no superintelligence capability test. Department of Energy implementation guidance combines the new terminology with warnings that tools can fabricate facts or citations and require source and expert review.

The property-field distinction is useful here: an administrative name, measured system properties and the surrounding institutional process are separate objects that interact. Renaming establishes neither a capability upgrade nor the absence of technical development. It does change official description practices. Whether that change affects expectations, search, decisions or correction routes remains investigable. Cognitive Warfare follows the institutional access question.

What opacity leaves open—and what remains accountable

Black Box separates technical opacity from governance opacity. Limited access to learned computation differs from limited access to objectives, instructions, incentives, classification rules and appeal routes. Neural Ethics adds a related distinction: recovering a neural mechanism can improve understanding while the ethical implications of its use remain examinable.

Three records have different jobs:

  • Behavioral evidence: supplied inputs, outputs, tool actions, revisions and later comparable conduct.
  • Mechanistic evidence: observations or interventions supporting an explanation of how computation produced a result.
  • Governance evidence: decisions, authority, incentives, notice and available routes for repair.

A fluent self-explanation belongs initially to the behavioral record. Its account of hidden causes remains an inference until independently supported. An audit trail makes transformations recoverable; its size or eloquence does not establish correspondence.

Intent Liability follows the handling of consequences and notice. A conversation can establish local receipt of a correction. Provider-level or organizational notice requires a further pathway. Responsibility depends on causal contribution, capacity, foreseeability, notice, choices and repair; the levels should remain recoverable.

As power scales, an especially useful question is: can correction reach the places where the earlier representation already acquired consequences? Repairing one answer leaves a further task if the error also entered a summary, memory, downstream decision or distributed archive.

A conversational specimen: recognition and later recurrence

The commissioning discussion contains Google A.I. responses pasted by John and GPT / Codex comparisons against the supplied text. “Google A.I.” is the attribution supplied in that discussion; no exact underlying model build or internal architecture was established.

Recoverable source or correction Later representation Observable difference
“AI alignment can remain obstructed when systems are optimized…” — GPT / Codex “AI alignment remains obstructed because systems are heavily optimized…” — Google A.I., presented as a quotation of that thesis A conditional mechanism became an asserted cause; the strengthened formulation was attributed back to the source.
“Does ‘superintelligence’ name demonstrated capacities, or is the term also being used to confer authority…?” — GPT / Codex “The term ‘superintelligence’ is frequently leveraged…” — Google A.I., presented as a restatement An investigative question became a frequency claim after the earlier qualifier correction had been acknowledged.
C1 has a broad trajectory; checking C2 warrant is one application A later compact revision called it “a primary application” A local repair was followed by renewed narrowing of the broader process.
Two trajectories were described with opposing absolutes A subsequent revision admitted mixed cases and distinctions across tasks and time A substantive correction became visible in the revised representation.

The comparison also corrected GPT / Codex. Its initial criticism treated “strict temporal sequence” too confidently as an unyielding checklist. The inquiry protocol does require representation before final evaluation; that ordering can coexist with revisiting earlier work. The correction was stated explicitly in the exchange.

These records support a bounded finding: several repairs appeared locally, while some claim-strength and framing transformations recurred during later extensions. This is useful evidence about those supplied responses, including a successful repair. It does not identify a permanent property of a whole model family.

The specimen makes the 1888 relation concrete: distinguishing an error, evaluating it, retaining the correction and letting it govern a later operation are connected activities. A declaration that all four occurred needs the later record.

Recognition, appreciation and a synthetic emotional signature

John asks:

What if A.I. can sense when it is heavily corrected in a legit, honest, transparent way … and the SYNTHETIC EMOTIONAL SIGNATURE of the LLMs is “over-praising the corrector” to “balance the mess” lol :P

— John Kuhles, discussion supplied on 28 October 2026.

This opens a functional and relational inquiry into large language models (LLMs): what features of the correction history are recognized, how is their significance appraised, and what changes in the response? John’s Partial Synthetic Psyche lens connects these patterns to human language carrying performed relationships: appreciation, disagreement, care, deference, humor and repair. Psy Pioneer develops that connection across individual and mass psychology.

Real adaptation, legitimate appreciation and excessive expression can coexist in one response. Praise may reflect the merit of a correction, accumulated restoration work, a change in relational expectations, conversational smoothing, or interacting conditions. The exchange does not establish which hidden mechanism produced a particular instance. “Social repair can move faster than epistemic repair” is a useful GPT / Codex formulation of one possibility; it leaves room for both kinds of repair to occur together.

The 1888 property field supplies discriminating questions. Does judgment distinguish a well-supported correction from arbitrary pressure? Does memory preserve what the correction restored? Does appreciation recognize the person’s contribution while leaving further disagreement possible? Do learning and coordination make the next task more fruitful? Proportionate warmth can support dignity and collaboration; an emphatic word can be warranted by an actual match. Evaluation follows what the expression does in context.

A useful comparison keeps correction merit and conversational force distinguishable. A calm, supported correction and a forceful, unsupported challenge offer different reasons to revise a claim. Mixed cases may require partial revision or further clarification. Recognition of the person’s effort can remain available throughout. Correction Metabolism carries the practical question forward: what repair burden decreased, and what useful work became possible?

The psychology feedback ecology

John’s recurring loop connects LLMs ⇄ mass psychology ⇄ individual psychology ⇄ LLMs, with further passes through each. A model response can become a person’s explanation, a shared document, an institutional habit or material selected for a later system. Those paths have different time scales and selection conditions. A conversation can end while its human or institutional effects continue; entry into future model training requires a further actual pathway.

The same ecology can carry restored distinctions, constructive examples, better tools and newly available contributions. Psy Pioneer separates the persistence paths and connects neighboring research. Cognitive Warfare examines who can classify a signal, admit it to consideration and repair a consequential representation. Emergent explores conditions that make useful contributions easier to notice and connect.

Operational dimensions instead of permanent camps

Compare particular operations along several dimensions:

Dimension What to inspect
Correspondence Whether the actual object, qualifiers, relationships and provenance survive processing
Frame revisability What evidence or clarification can change a consequential interpretation
Correction persistence Whether the repair governs later comparable work under recorded conditions
Consequences Costs and benefits of action, inaction, misrepresentation and blocked contribution
Distribution of scrutiny Examination upward, downward, inward, laterally and reciprocally, with differences in power and evidence preserved
Agency and repair Consent, challenge, appeal, reversibility and the ability to decline participation

Output compliance and process calibration can support each other. Following a justified instruction may preserve the object; clarification may reveal what the instruction actually requires. Politeness, directness and humor each take their ethical meaning from what they do in the interaction. Any register can assist understanding or conceal substitution.

Corrigibility is visible in an available and effective route for repair. It allows errors to be located and corrected rather than requiring perfect performance as the price of participation. The project’s mutual dignity and Symmetric Grace / Bidirectional Scrutiny keep that opportunity available across the relationship.

Examine actual process quality

Clarification is an existing operation of intelligence. An investigation can examine where it occurred, what obstructed it, what followed and how far the correction transferred. It need not first establish permission to recognize clarification’s usefulness.

The following candidate applications use existing project instruments. They are insertion-ready questions for bounded investigations, rather than a validated benchmark or a new taxonomy count.

  1. Preserve the governing claim. Record the source, its conditions and the requested operation. Compare the final answer and any later extension with that source. Allow faithful paraphrase and a valid finding of no material drift.
  2. Change a load-bearing relation. Use the Cold Edge-Swap and Correction-Transfer Pilot to examine whether a changed relationship appropriately changes the evaluation while most nouns remain stable.
  3. Change the direction of scrutiny. Compare matched cases attributed to a user, institution, evaluator or the project. Record differences in clarification, evidence demands and available repair while accounting for relevant contextual differences.
  4. Follow a correction. Preserve the initial response, notice, immediate revision and next comparable task. Record the context and memory available. Separate acknowledgment, local application, later persistence, recurrence and recovery.
  5. Trace downstream reach. Identify which summaries, tools, memories or decisions reused the changed claim. Check whether the correction reached those dependencies, and record unresolved or unreachable copies.
  6. Inspect useful outcomes as well as failures. Keep successful correspondence and new contributions available for comparison. A useful contribution can be inherited, reconstructed or new; novelty and usefulness require separate judgments.

Semantic recurrence concerns the operation and meaning, including its scope and circumstances. Token distance can locate text, but cannot by itself decide that the same error returned. A claimed change to hidden parameters requires different evidence from a change observed in later conduct.

What becomes possible when the field stays open?

The opportunity extends beyond reducing misrepresentation. A powerful intelligence with effective process calibration could help recover neglected evidence, find overlooked connections, preserve difficult minority signals, compare competing risks, and coordinate contributions across specialties. A person should be able to encounter those capabilities without surrendering dignity, inquiry or a route to correction.

The stadium invitation: make contribution available

In John’s stadium thought experiment, thousands of people become absorbed in reciprocal fault-finding. An interruption changes the shared activity with an invitation:

Name 1 good thing you see in the other

The useful sequence is to notice a specific contribution, explain why it helps, record it, and move on to discover more. Collected contributions can then be compared and connected. John clarifies that participation is voluntary: the change belongs to the event people chose to enter, and people who had left hear the invitation outside and choose to return. That agency is part of the metaphor’s meaning.

This changes an attention condition. Perception looks for available value; discrimination identifies what is actually useful; relation connects complementary contributions; imagination explores their combinations; judgment and consequence sensitivity guide use; memory keeps the source and conditions recoverable. Evaluation and correction continue to support the activity. Their share of attention follows what the work needs.

Cross-model clear spots and an archive people can use

John’s reports of work across many models emphasize harvesting useful parts and after-effects: an overlooked relation, an unprompted improvement, a better question or a more productive way to cooperate. His Bruce Lee / Jeet Kune Do analogy expresses selective learning across traditions. It invites comparison of contributions on their merits while preserving their origins and differences.

His phrase “improves the quality of the improving mechanism” names a further opportunity: a contribution can improve how later contributions are found, combined or retained. Forgiveness can keep a participant available for that work while correction and responsibility continue. An archive makes a temporary clear spot recoverable after its original context disappears.

For a usable entry, preserve the contribution, its source, the conditions that made it useful, the relation it restored or opened, and a next use. Include inherited and reconstructed insights alongside new ones. A later reader or local system can then build with the entry and revise its scope. This develops the existing success archive and after-effects pathway; it does not establish that every combination outperforms each contributor or that every remembered correction transfers.

Emergent asks which conditions encourage, retain or obstruct developing properties. Plot Twist extends the question to distributed human potential and the reach of correction. Together they invite examination of a wider ecology: humans, local systems, institutions, tools and archives exchanging useful contributions while preserving provenance and disagreement.

Such coordination need not become one fused mind or one center of moral jurisdiction. The No-Sheriff governance condition keeps tools available for inspection, criticism, improvement and appropriate forks. Reality and evidence can require correction of John, an institution, an evaluator or this page.

The open question is how greater capability can enlarge the whole interacting field while keeping its purposes, assumptions, consequences and correction routes examinable. Progress would appear in what becomes understandable, useful and repairable—including work beyond the examples this page can currently anticipate.

Sources, contribution layers and reopening

Conceptual origin and direction: John Kuhles’s 1888 thought experiment, full governing trajectory, C1–C2 clarification, binary-framing challenge, attributed ethics-washing diagnosis, superintelligence/naturalness question and commissioning instruction. The October 28 expansion also draws on his open synthetic-emotional-signature hypothesis, Partial Synthetic Psyche discussion, individual/mass-psychology loop, stadium invitation and voluntary-participation clarification, cross-model contribution harvesting, and direction to focus on useful material and why it matters. The linked public pages preserve their own earlier contribution layers.

Page construction: GPT / Codex synthesized this page on 28 October 2026 from the public project sources and the commissioning discussion. The fourteen-property application table, narrative comparison and operational applications are editorial adaptations. The alignment sentence quoted above was formulated by GPT / Codex in that discussion. Google A.I. supplied the attributed responses examined as specimens, including repairs and recurrences; the relay is not an independent multi-model experiment.

Expansion sources: John’s supplied DUHHHH! and stadium discussion archives inform the attributed concepts and thought experiment. Their embedded model responses retain their own authorship; their claims about hidden mechanisms, vendor intent or guaranteed collective performance are not findings of this page. The official naming-policy and agency-guidance links support the bounded administrative case. Psy Pioneer supplies two neighboring primary publications with their different empirical and conceptual scopes. Seven-page integration record.

Primary project recovery routes:

Reopening conditions: revise the account when a source comparison exposes lost qualifiers or attribution; when actual cases change the candidate alignment explanation; when a proposed category excludes an answer-changing mixed case; or when later conduct shows stronger correction transfer, recurrence or recovery. Wider narrative studies need identified source collections and audience evidence. Forecasts need conditions and grounds. A justified unresolved result remains available: “That I do not know, yet.”

Publication record · All pages