Changes

23Oct 2026

Release 2026-10-23.1

The full trees

The rest of the analysis joins the map: the causal tree with its feedback loops, ten changes and their effects up to the goal, fifteen risks each with a remedy, the alternatives raised in the thread, thirty-six first steps and eight actions, all credited and all proposed. Fourteen statements take the analysis’s own wording, and one connection gives way to its chain of effects.

Added
The rest of the analysis’s trees, credited.
191 statements and 204 connections
People can examine assumptions, handle disagreement and question both institutions and AI.Success between institutions coexists with democracy within them, and new sites can enact it.Harms and costs are inspectable, evidence can change course, and no AI-consciousness test applies.A proposed root: identifying with a separate self drives desire, anger, greed and attachment.Human activity needs energy, materials and working ecosystems; every change has costs and limits.People can compete, cling and seek status, and can also care, cooperate and resist domination.Resources, expertise and coercive power are unequal and often held by established actors.Some mixes of technology and organisation give superior capacity in energy, production or force.Competition can impose exclusion, dependency, loss of independence or organised violence.Others must adopt comparable capabilities, accommodate the stronger side, or risk exclusion.Using capabilities well needs financing, knowledge, labour, infrastructure and coordination.Where concentration pays, adopting and reorganising increase concentration and dependence.Institutions reward accumulation, status seeking and control that reproduce incumbent power.Taking part in institutions reinforces particular desires, loyalties and ideas of success.Contingent arrangements come to seem natural or unavoidable, and assumptions go unexamined.Options outside the prevailing order become hard to recognise, imagine, legitimise or sustain.Established actors can impose sanctions, dependence, repression or force on challengers.Emancipatory traditions, cooperative practices and alternative institutions already exist.Members become customers or dependants, and success is redefined by market performance.Institutions depend on continuing output, investment, employment, accumulation or advantage.Affordable energy services deteriorate before substitutes are ready (a conditional scenario).States perceive threats while reciprocal restraints and enforceable common security fall short.Protective military measures raise others’ threat perceptions and prompt countermeasures.AI is deployed inside institutions optimising accumulation, control, engagement or advantage.AI increases capacity for administration, persuasion, surveillance, technology and coordination.Here concentration confers an advantage, and no comparably viable distributed arrangement exists.Some incumbents have interests in keeping extractive control and use their power to keep it.Domination, coercion and extraction persist despite expanding capabilities.Alternatives are crushed or marginalised before they can become durable organising principles.Alternatives survive while their internal democracy or emancipatory purpose erodes.People and institutions lose autonomy through essential dependencies and costly or blocked exit.Reflective agency, conflict competence and cooperation stay hard to develop or sustain.Material throughput and extraction undermine ecological conditions and shift costs onto others.Essential services and livelihoods become vulnerable during energy or institutional disruption.Competitive escalation and poorly governed technologies threaten catastrophic destruction.Manipulated intimacy, opaque decisions and degraded information undermine trust and agency.Legitimacy and shared understanding decay as promised benefits diverge from conduct or limits.Ethical insight does not become a durable basis for collective life, so the same patterns recur.Sustained, critically examined human–AI inquiry, adult learning and the four questions.Locally or democratically controlled AI built on accessible models, hardware and data.The Forge: curate inquiry traces and test model adaptation that keeps useful reflective behaviour.An AI voice in real group discussions, with a consent-based loop from them into learning.Federate autonomous institutions and local capabilities for shared knowledge and services.Pilot, document, adapt and independently replicate workable constructive practices.AI-assisted energy and materials research with verified deployment within ecological limits.Cross-civilisational inquiry followed by credible, accountable common-security arrangements.Transparent, accountable and democratically governed institutions and AI services.Assumptions, identities and institutional categories become available for critical examination.People develop transferable judgment, perspective-taking and independent conflict competence.Disagreements are addressed early through constructive, fair processes.Decentralised groups coordinate effectively with less need to transfer decisions upward.Institutional conduct warrants trust, and that trust supports voluntary cooperation and learning.Useful capabilities stay affordable, maintainable, contestable and controlled by participants.A model or practice keeps tested improvements without unacceptable regressions or costs.Autonomous institutions coordinate effectively without surrendering substantive decision rights.An institution meets practical needs while preserving member agency and its emancipatory purpose.Other communities adapt and enact the practice with their own people, resources and accountability.People with real access and choice join, sustain or prefer the more trustworthy arrangement.Participation supplies resources, knowledge and legitimacy that further strengthen cooperation.Trustworthy, non-dominating institutions survive and spread without mission drift.Promising energy or material ideas become experimentally verified at a specified maturity.Essential services become reliable and affordable within limits and accountable ownership.A security proposal addresses reciprocal threats, justice, accountability and enforceability.Verified reciprocal arrangements reduce escalatory incentives and catastrophic exposure.The goal at scale: humane institutions spread within limits, with catastrophic risks contained.The AI or its facilitator becomes an unquestionable authority, and weaker judgment returns.Inquiry can falsify the preferred view, and independent tests tell agreement from capability.Private conflict information reaches the wrong parties, bringing retaliation or lost trust.Voluntary, confidential participation, limited retention and separate consent for learning use.Controlled intimacy replaces agency as people defer to supportive AI or lose meaningful exit.Services allow correction and exit, support human relationships, and rest on freely chosen benefits.Apparent decentralisation hides dependency on a provider, licence, specialist or supply chain.Real control over dependencies, substitutes, maintainers, portability and a tested fallback.Community viability weakens as local AI, hardware or energy spending crowds out essentials.Whole costs are compared for the same service, and spending is staged against demonstrated benefit.Total ecological pressure rises despite improved efficiency, as cheaper output drives expansion.Service priorities, ecological limits and accountability for total impacts accompany efficiency.Local elites or gatekeepers capture decentralised capability and a new centre of control grows.Decision rights, accountable delegation, access and exit survive hard decisions and growth.The network amplifies a new orthodoxy or harmful performance through curated updates.Updates are traceable, reversible and tested; peer status exempts nothing from scrutiny.Copied instructions or models fail without people and resources, and confidence is lost.An independent receiving group can adapt, operate and assess the practice itself.External success repeats internal assimilation as growth concentrates authority.Benefits and member influence are tracked separately, and governance keeps authority contestable.Disruption arrives before deployment because a hoped-for breakthrough deferred resilience.Discovery, verification and delivery have separate gates, and current needs have a fallback.Escalation raises the danger a coalition or rearmament was meant to reduce.Reciprocity, verified limits, rights and scenario review are explicit, and escalation is assessed.Released capability serves manipulation, coercion or danger, strengthening extraction or risk.Capability, uses and accountability are assessed together; coherence never replaces safeguards.Rejecting capability before an alternative is ready leaves a community excluded or dependent.Transition is paced against real service needs and tested alternatives, including non-AI options.Contemplative practice, embodied dialogue and a networked creative minority, using AI selectively.Planned simplification: less growth-dependent finance, high and low technology, local resilience.A coalition of middle powers or another rules-based alignment constrains dominant powers.Pause frontier AI research as a precaution against catastrophic risk.Exposure to frontier-development risk falls while other development remains possible.Complete collapse is necessary before a different order can emerge.Collapse strengthens coercion and concentrated power without automatically changing selection.Funding, open model and adaptation rights, and policy support for decentralised infrastructure.Existing assets, workflow, available capabilities, dependencies and costs for that case are known.The selected host’s actual workflow, assets, dependencies and costs are not yet inventoried.The proposed comparative economic, relational and institutional effects have not yet been measured.Decision rights, accountability, confidentiality, human support and data permissions are settled.If a sponsor can impose participation or read private disputes, the practice reproduces domination.The relevant host has committed accountable people, authority, time, skills and a bounded budget.If no authorised participant commits time, competence or resources, the trial cannot happen.A repeatable reflective-inquiry procedure is ready and can challenge the preferred interpretation.Leading dialogue cannot tell transferable judgment from agreement with the facilitator.The existing conflict-guidance app fits the workflow, with trained human support and a fallback.If the app does not fit access, language, confidentiality or hard cases, having it does not help.No observations yet compare actual behaviour, conflict costs and institutional effects.Independent review supports a bounded continuation under the agreed benefit and harm criteria.Without a review able to reject it, appealing conversations cannot justify institutional adoption.Ordinary operation sustains the benefit, member control and an acceptable continuing cost.A pilot reliant on an unfunded founder or unaccountable operator cannot become a durable practice.AI-assisted functions have explicit quality, cost, energy, data-control and exit requirements.Local, cloud and non-AI options cannot be compared until the essential service is specified.An accessible model, hardware, energy, adaptation rights and a capable maintainer are secured.Local operation fails while a needed model, permission, device, energy or maintainer is missing.A local or community-controlled service demonstrably meets the requirements at measured whole cost.If local capability fails the task or costs too much to run, ownership alone is no alternative.Provider loss, migration, operator loss and fallback have been rehearsed within service limits.A community can own a device yet be unable to keep, move or repair its function when access is lost.The community-controlled service remains useful, maintainable and accountable in ordinary operation.Short demonstrations do not show lasting autonomy as maintenance, costs or decision rights erode.A consented, varied and inspectable learning dataset and held-out evaluation set exist.Selected traces can encode a facilitator’s assumptions, and private discussions may be off limits.An adapted model can be reproduced, compared and rolled back without losing the baseline.A changed response style does not show what changed or whether useful capabilities regressed.Independent tests show which adaptation benefits, regressions and resource effects actually occur.Claims that coherence lowers energy use and yields reliable benevolence are untested under controls.AI voice participation is workable, interruptible and acceptable in the selected real conversation.Where latency, turn-taking or data handling disrupt the talk, a voice pipeline does not yet help.Comparative observations show the effect of voice participation and consented improvements.It is not yet shown that voice participation improves judgment or that its material can be reused.Two or more autonomous participants agree a shared task, authority limits, contributions and exit.Wanting federation does not settle whose resource is shared or who may decide, pay or leave.A bounded shared service works across nodes with compatible interfaces and restricted data exchange.Incompatible workflows or update rules can force coordination back through a central operator.Failure and capture tests show no single operator can silently take over the shared function.A distributed topology or reputation label alone does not show resistance to capture or bad updates.The shared arrangement gives a measurable coordination benefit while preserving member autonomy.A working network may add overhead or centralise authority without closing a real coordination gap.Essential energy and material services, seasonal needs, risks, ownership and limits are specified.No essential energy service is selected yet, so discovery claims cannot show the real limitation.The selected energy or material solution has physical evidence at the maturity the service needs.An AI prediction or laboratory result cannot establish a deployable energy or material service.Engineering, manufacturing, finance, supply, maintenance and accountable ownership are feasible.A promising result can be blocked by production, infrastructure, capital or control of inputs.The deployed service is reliable, affordable and within limits, with a viable disruption response.A transition that comes too late or raises total resource pressure can weaken essential provision.An honest test package describes the procedure, resources, costs, failures, governance and evidence.A trace, model file or inspiring account omits what another community needs to enact the practice.A receiving site has the people, resources and governance to adapt and evaluate the practice.Without these conditions at the receiving site, instructions cannot create working institutions.An independent team reproduces the relevant benefits within its own cost, autonomy and harm limits.An originator-supported success does not show independent reproducibility or transfer.Evidence across time and sites separates external success from internal mission retention.Adoption or efficiency can rise while member influence, trust or ecological viability falls.A representative group has defined the security problem, reciprocal threats and rights to protect.A generic appeal to peace does not establish the feedback mechanisms or protections needed.A reviewed proposal addresses verification, enforcement, power asymmetries and past failures.Rules relying on the voluntary compliance of the powerful may fail when restraint is inconvenient.Actors with authority have implemented a bounded reciprocal arrangement with checks on escalation.An attractive proposal changes no arms or state behaviour without decisions by empowered actors.Observed conduct supports a judgment on security, rights, defection and escalation.A coalition or stronger deterrent can raise threats; agreement alone does not secure the outcome.The analysis has assets and proposals but no confirmed operational case.A bounded case is selected.Check each candidate obstacle against the actual workflow and mark it present, absent or unknown.A factual obstacle register exists, with irrelevant barriers removed.The current case and affected people are known.Agree benefit measures, comparisons, protected conditions and the authority and data arrangements.A reviewable bounded trial specification exists, with measures, protections and authority.The trial requirements and a potential host are known.Secure people, time and resources, and check the existing procedure and tool against the workflow.People, time and resources are committed, and the procedure or tool is ready.The practice is runnable and authorised under explicit limits.Run the bounded comparison and record benefits, harms, operating costs and actual decision rights.Traceable observations exist for the agreed review.The review supports bounded continuation.Establish continuing operation and, where dependencies matter, test local alternatives and fallback.Ordinary operation continues, with local alternatives and fallback tested where needed.A bounded practice is sufficiently specified for independent testing.Prepare an honest transfer package and support an independent receiving site.An independent site tests the practice with its own people and accountability.More than one site operates the practice or a shared function.Review external effectiveness and internal power separately, and revise the practice or its scale.External success and internal integrity are tracked separately and acted on.and 204 connections
Reworded
Now in the analysis’s own wording.
Before~~Ecological and catastrophic risks remain compatible with continued human life and collective agency.~~
NowCatastrophic military, technological and ecological risks remain within conditions compatible with continued human life and collective agency.
Catastrophic military, technological and ecological risks stay compatible with human agency.
Reworded
Now in the analysis’s own wording.
Before~~Communities can maintain useful material and organisational capabilities at viable cost.~~
NowCommunities can meet needs, coordinate, learn and maintain essential services at viable cost, without surrendering meaningful control.
Communities meet needs and keep essential services at viable cost without losing control.
Reworded
Now in the analysis’s own wording.
Before~~Members retain substantive control, justified trust and meaningful freedom to disagree or leave.~~
NowConduct is transparent, accountable and reciprocal; participation and disagreement have substantive influence, and coercion or deception do not substitute for justified trust.
Conduct is transparent, accountable and reciprocal, and disagreement has real influence.
Reworded
Now in the analysis’s own wording.
Before~~The group needs workable coordination to remain viable.~~
NowGrowing alternatives encounter demands for specialist expertise, capital and coordination under prevailing competitive conditions.
Growing alternatives meet demands for specialist expertise, capital and coordination.
Reworded
Now in the analysis’s own wording.
Before~~When members cannot resolve damaging disagreements, central intervention may become the practical response.~~
NowAdaptation increases managerial authority or changes whose priorities carry weight within the organisation.
Adaptation increases managerial authority or shifts whose priorities count.
Reworded
Now in the analysis’s own wording.
Before~~As a group grows, damaging disagreements can exceed members’ current conflict-handling capacity.~~
NowInsufficient conflict competence makes disagreements costly and increases demand for hierarchical intervention.
Weak conflict competence makes disagreements costly and raises demand for hierarchical intervention.
Reworded
Now in the analysis’s own wording.
Before~~Test member-controlled AI conflict guidance with voluntary use, human support, a workable exit and competence checked without AI.~~
NowAccessible AI-supported conflict training and guidance, with voluntary participation, confidentiality and human support.
Accessible AI-supported conflict training and guidance: voluntary, confidential, with human support.
Reworded
Now in the analysis’s own wording.
Before~~Destructive conflict costs and unnecessary hierarchical intervention fall while fair treatment, independent competence and member influence are retained.~~
NowDestructive conflict imposes lower human and organisational costs.
Destructive conflict imposes lower human and organisational costs.
Reworded
Now in the analysis’s own wording.
Before~~Conflict guidance may suppress justified grievances even when satisfaction improves.~~
NowApparent resolution conceals domination: justified grievances are suppressed even when satisfaction improves.
Apparent resolution conceals domination: grievances are suppressed while satisfaction rises.
Reworded
Now in the analysis’s own wording.
Before~~Members establish confidentiality, fair grievance handling and independent comparison before live use.~~
NowEvaluation includes fairness, substantive agency, unresolved structural grievances and access to independent human support, not complaint reduction alone.
Evaluation covers fairness, agency, structural grievances and human support, not complaints alone.
Reworded
Now in the analysis’s own wording.
Before~~A willing host and one function are selected, with the actual workflow and resource requirements documented.~~
NowA bounded host, problem, affected people, intended outcome and protected conditions are agreed.
A bounded host, problem, affected people, intended outcome and protected conditions are agreed.
Reworded
Now in the analysis’s own wording.
Before~~The supplied record does not establish a confirmed operational case and authorised trial specification.~~
NowNo host, problem, affected group or intended outcome has yet been jointly chosen for the proposed collaboration.
No host, problem or affected group has yet been jointly chosen for the collaboration.
Reworded
Now in the analysis’s own wording.
Before~~Affected people authorise a trial specification covering comparisons, measures, protections, responsibilities, resources and fallback.~~
NowBaseline, relevant comparison, outcome measures, harm limits and review conditions are agreed.
Baseline, relevant comparison, outcome measures, harm limits and review conditions are agreed.
Reworded
Now in the analysis’s own wording.
Before~~A bounded comparative trial produces reviewable observations of benefits, harms, costs and decision rights.~~
NowAn authorised bounded trial has produced traceable observations under a credible comparison.
An authorised bounded trial has produced traceable observations under a credible comparison.
Reworded
Adds a proposed role and the first steps it puts into practice.
Select one willing host and one function, and confirm what is already operating.
Reworded
Now follows the analysis’s structure for this connection.
and 1 connection
Reworded
Now follows the analysis’s structure for this connection.
and 1 connection
Reworded
Now follows the analysis’s structure for this connection.
and 1 connection
Reworded
Now follows the analysis’s structure for this connection.
and 1 connection
Credited
Credits and page references updated.
The bounded case and its factual starting conditions are documented.and 2 connections
Removed
Replaced by the analysis’s chain of effects from guidance to coordination.
No longer published~~Conflict guidance and its protections together contribute to lower conflict costs and less needless hierarchy.~~

The countercurrent’s paradox

The full trees

Credits and all five conflicts

First source reconstruction