Analytical appendix

Arguments and sources

A map of claims, limits and open questions: what is known, what is estimated, and what still needs testing.

Vladimir Lovtsov29 min12 sectionsWorking edition

Check a claim, find its source and examine its limits. Each source identifies how it is used and how far it has been checked.

1. The argument map

The manifesto and complete book rest on five connected claims. The manuscript contains a preface, 24 chapters and an afterword. This appendix examines the sources and assumptions of the current text, including the teaching calculations in chapters 15, 18 and 22. The previous abridged draft is retained for comparison.

1.1. AI changes the composition of tasks

Generative models can already perform some cognitive tasks: organise, classify, create drafts, write and check code, and use tools. Applicability depends on context, acceptable error, data quality and the cost of oversight.

Status: observable fact.

What does not follow: a whole profession disappears; a technical possibility is economically justified; an organisation can integrate the tool safely.

1.2. Impact more often begins with a changing role

In its 2025 study, the International Labour Organization estimated that roughly one in four workers worldwide is in an occupation with some exposure to generative AI. Only 3.3% of global employment fell in the highest category. The ILO considers task transformation a more likely general outcome than complete role replacement.

In 2026, it separately warned that exposure indices are not forecasts of redundancies. They do not account for the full implementation cost, institutional constraints or actual market behaviour.

Status: an estimate based on a published methodology.

Limitation: a global estimate does not predict the fate of a particular role, country or person.

1.3. The organisation distributes the gains from technology

Greater productivity can fund lower prices, more output, better quality, investment, profit, shorter working hours or staff reductions. The tool does not choose the distribution. Markets, ownership, authority and incentives shape that choice.

Status: the author's interpretation of economic and management logic.

Condition for revision: if empirical evidence shows that different ownership forms under otherwise comparable conditions do not affect the distribution of gains, the claim should be narrowed.

1.4. Joint ownership changes bargaining power, but does not abolish the market

A cooperative can combine participation in governance with a share in the economic result. In a Russian production cooperative, the cooperative itself owns its assets. Members have membership rights and rights connected with their cooperative share; they do not automatically own a personal fraction of every server or hold copyright in every program. In this book, “co-ownership” means participation through the organisation. Rights in a particular product, permission to use materials and settlement on withdrawal require separate arrangements.

It does not guarantee:

a good product;
fairness in every decision;
equal pay;
freedom from redundancies or conflicts;
rapid access to capital;
resilience against competition.

Status: an observable property of the organisational form, plus the author's assessment of its significance.

1.5. A small product can test the digital artel

The working hypothesis is that a small product team can use AI and automated quality checks to sustain end-to-end delivery, while shared ownership connects voting rights with responsibility for the outcome.

Status: an organisational hypothesis with an existing analytical foundation; practical effectiveness is not yet established.

Test: the bounded ninety-day experiment described in section 7 of this appendix.

What would disprove it: no payment signal, inability to delegate decisions, opaque contributions, a conflict without a workable procedure, or worse economics than available alternatives.

2. What is known about work and AI

2.1. Distinguish exposure, automation and redundancy

Three terms are often used as synonyms, although they describe different stages.

Exposure means a system's capabilities overlap with some tasks in an occupation.

Automation means an organisation has technically and economically integrated a system in place of some human work.

Redundancy means an organisation has reduced roles or headcount after considering strategy, demand, budgets and employment relationships.

No automatic transition connects these stages. High exposure can lead to augmentation, a changing role, a lower price for a particular task or no adoption at all.

Public trackers help observation, but attribution of causes often relies on company statements, news classifications and a mixture of factors.

2.2. The balance of job creation and displacement

The World Economic Forum's Future of Jobs Report 2025 draws on responses from more than a thousand large employers and ILO data. Respondents expect combined macrotrends to create about 170 million roles and displace about 92 million by 2030, producing a positive net balance in the labour-market segments covered by the study. This is not an exhaustive count of global employment.

These figures prove neither universal prosperity nor the end of work:

they concern several macrotrends, not AI alone;
they are employer expectations, not an observed outcome;
new and disappearing roles may concern different people and markets;
a positive aggregate does not remove transition costs.

For the book's argument, changing tasks and bargaining power matter more than the sign of the overall balance.

2.3. Substitution and complementarity

The IMF describes AI's impact as a combination of two effects. Technology can replace individual tasks and complement human work. Advanced economies have higher exposure, but also greater potential for productivity gains. Outcomes depend on skills, institutions and organisational policies.

The book deliberately does not choose just one scenario. Displacement and augmentation can happen simultaneously within one company, or even one role.

2.4. Why tasks matter more than job titles

A task-level view helps personal planning:

Task type	Likely direction	Possible response
Repetitive, fully digital, easy to check	Becomes cheaper quickly	Learn automation and move towards checking outcomes
Needs context from several systems	Partly automated	Build accessible, secure context
Contains an irreversible decision	AI assists analysis	Keep authority and responsibility with a person
Depends on trust and negotiation	Benefits from better preparation	Develop listening and negotiation skills
Covers the entire route to operation	Hard to replace completely	Understand the product and delivery context

This table is not a forecasting model. It is a tool for examining your own work.

3. Why ownership belongs in the AI discussion

3.1. Productivity gains contain no distribution formula

Suppose a tool reduces the cost of an operation by 30%. The benefit can be shared among:

customers through lower prices;
the organisation through profit or reserves;
investors through the value of capital;
employees through income, time or development;
a new product through reinvestment;
society through tax and service availability.

No technology chooses the proportions. That is a political, economic and management decision.

In an ordinary company, rights to assets and residual returns are determined by law and agreements. Financing and risk-bearing help explain those arrangements, but owners are not the only people who suffer when a business fails. Participation in a cooperative’s return also needs a legal basis; bearing a risk does not by itself create a right to profit.

The book does not claim that the latter is always fairer. It argues that structure changes participants' rights and therefore deserves an engineer's examination.

3.2. Data, models and compute

The discussion of cooperative AI goes beyond a developer's share in a product. The Platform Cooperativism Consortium considers joint ownership and control at several levels:

work creating and evaluating systems;
data and rights to its use;
models and intellectual assets;
computing infrastructure;
rules for application and distribution of results.

The target SoDelo model connects production, consumption and ownership of resources within a federation of cooperatives. Its foundation has already been developed in the project analysis. Joint operation of existing models is one proposed direction; training a proprietary frontier model is a different, substantially more capital-intensive undertaking. The initial computing service is sized after measurement, full budgeting and demand validation. A small product pilot and an infrastructure pilot can be separate stages of the same proposal.

3.3. Why an option is not enough

A stock option can give someone an economic interest in company growth. It can be a powerful way to align incentives.

Options and democratic control, however, answer different questions:

an option: which part of financial growth a person might receive;
a vote: which decisions a person has a formal governance right to influence;
an employment contract: which predictable compensation they receive;
a charter: who can change the rules themselves.

For some companies, salary, options and a strong culture work better than a cooperative. SoDelo makes sense only where members want economic participation together with control of the basic rules.

3.4. Objections that ownership does not remove

A worker may depend on the same enterprise both for earnings and for the value of invested funds. A common downturn can reduce income, impair the investment and increase the need for help together. Joseph Blasi, Douglas Kruse and Harry Markowitz, NBER, 2008 examine concentrated risk under employee ownership and ways of limiting it. Their conclusion is not that holding a stake is always unwise: other financial supports, participation terms and individual preferences matter. Chapter 21 adds the author’s stress check for simultaneous requests on a common reserve.

Equal votes do not guarantee sensible compensation for scarce skills either. In Michael Kremer’s theoretical model, 1997, a majority may choose redistribution that weakens incentives, while capital tied to the firm discourages dissatisfied members from leaving. The conclusion depends on the model’s assumptions. For SoDelo, this is a reason to examine remuneration and exit terms in advance, including cases where a majority favours a decision that harms the team.

A theoretical risk does not establish that such firms necessarily fail more often. Gabriel Burdín’s study of Uruguayan firms, 2013, finds greater survival among worker-managed firms in its sample. This is an observational comparison within a particular country and legal setting, not an experiment establishing a universal cause or a forecast for a Russian project. The open NBER abstracts and Burdín’s abstract, introduction and discussion of results were checked on 12 October 2026; his econometric analysis was not reproduced.

4. Real organisational reference points

4.1. International principles

The International Cooperative Alliance identifies seven principles:

voluntary and open membership;
democratic member control;
member economic participation;
autonomy and independence;
education, training and information;
cooperation among cooperatives;
concern for community.

For primary cooperatives, the basic rule is one member, one vote. It is a governance principle, not an instruction to pay everyone equally or decide everything unanimously.

4.2. Igalia

Igalia's official materials describe it as worker-owned and worker-managed. Founded in 2001, it works on technically complex open-source components, including browser engines, graphics, multimedia and language runtimes.

What the example shows:

joint ownership is compatible with substantial engineering complexity;
mentoring supports new colleagues as they join the shared work;
democratic governance requires people prepared to share responsibility.

What it does not prove:

the model transfers to every market;
equal pay and an assembly suit every team;
a cooperative is automatically more productive than a conventional company.

4.3. CoTech

CoTech is a British network of technology cooperatives. It gives clients access to several specialisms and lets small organisations collaborate without merging into one hierarchy.

What the example shows:

a federation can achieve scale through shared functions;
a small autonomous team need not become a giant enterprise;
shared sales, a brand and exchanges of practice can be a distinct network offering.

Questions for transferring the model:

responsibility to the client;
margin distribution;
common quality requirements;
a procedure for removing a weak or dishonest member organisation.

4.4. Mondragon

Mondragon matters for its scale and its failures. The Fagor crisis shows that a cooperative ecosystem can soften the consequences for people, but does not abolish flawed strategy, competition or the bankruptcy of a particular business.

The book uses the example to limit optimism, rather than as a promotional number.

The group's official chronology records the Fagor Electrodomésticos crisis in 2013 and the subsequent work by cooperatives to find employment solutions for affected members. It is a source for the event itself, not an independent assessment of the crisis's causes.

4.5. Platform Cooperativism Consortium

PCC gathers and studies digital cooperative projects. Its reports cover initiatives in different countries; in 2025 it gave particular attention to cooperative AI: shared control of data, models, compute, labour and rules.

Many projects establish that a field of experimentation exists. They do not establish every project's economic viability or the applicability of a particular Russian model.

5. The Russian legal framework

5.1. What Federal Law No. 41-FZ regulates

The law on production cooperatives covers:

charter requirements;
membership and governing bodies;
the share fund;
personal labour and other participation;
profit distribution;
members' employment relationships;
reorganisation and liquidation.

The book cites an existing framework without claiming that SoDelo must use this particular form.

5.2. What cannot be promised without specialist review

Without examining the actual activity, one cannot promise:

a particular tax advantage;
exemption of payments from social insurance contributions;
a safe profit-distribution formula;
one uniquely permissible intellectual-property arrangement;
scaling without changing the applicable regime;
automatic compatibility between different participant types.

Article 19 separately regulates members' labour and their social and medical insurance. A blanket statement that profit distributions attract no insurance contributions is therefore especially risky without examining the actual payments.

Before launch, separate advice is needed on:

the organisational form;
agreements with candidates and members;
intellectual-property transfer or licensing;
accounting for labour and other participation;
taxes and contributions;
personal data and external AI services;
cross-border payments and licences, if applicable.

5.3. The book is not a public offer

The SoDelo participation model described here is a working hypothesis. SoDelo is developing as a joint initiative. The site and participation rules are being prepared; public registration is not open. There is no offer to buy a share, make a contribution or join an operating organisation. Any public version must preserve that boundary until the legal and economic model has been reviewed.

5.4. Producing, consuming and federating raise different legal questions

Legal links were rechecked on 19 October 2026 against the legal database’s open texts. The accessible consolidated version of Law 41-FZ lists amendments through 5 April 2021. Before registration or a contract, the version must be checked again against the rule’s effective date.

Joint production: Russian Civil Code, Articles 106.1–106.6 and Law 41-FZ. Labour, participation, property and liability are discussed within this framework.
Meeting shared needs: Civil Code, Article 123.2 and Russian Law No. 3085-1 of 19 June 1992. The latter covers consumer societies and their unions within its scope, not every type of cooperative.
Bringing organisations together: Civil Code, Article 123.8 on associations and unions. “Federation” describes our plan; the word does not itself select a legal form.

These branches matter because product development, equipment procurement and member representation may need different arrangements. A consumer form is not a way to distribute profit without constraints, and equal voting does not remove property obligations. SoDelo’s specific structure has not been approved.

For a production cooperative, members’ personal liability needs separate attention. Russian Civil Code Article 106.1 and Article 13 of Law 41-FZ provide for members’ subsidiary liability; its extent and procedure depend on the law and charter. Potential losses cannot be assumed to stop at the initial contribution without examining those terms. The cooperative itself answers with its assets; an internal social-reserve designation does not shield funds from its creditors.

For a consumer cooperative, Civil Code Article 123.3 requires members to cover losses through additional contributions within three months after approval of the annual balance sheet. Members bear joint and several subsidiary liability within the unpaid portion of each member’s additional contribution. The specific cooperative type requires separate examination.

The organisation’s ownership and a member’s rights differ: Article 9 of Law 41-FZ addresses the cooperative’s assets and members’ shares. Article 15 gives each member one vote. Article 22 allows voluntary withdrawal with written notice to the chair or board at least two weeks in advance; debt does not prevent withdrawal. Withdrawal itself, settlement timing and the amount due remain distinct questions.

Russian Civil Code Article 1295 governs works created within employment duties: the employer normally holds the exclusive right unless the contract provides otherwise. Authorship and statutory author remuneration remain distinct rights. They are not equity in the company and do not arise automatically for every project participant.

6. The author's experience and the boundaries of examples

6.1. The experience I draw on

Innotech Group, part of T1 Group: August 2021 to the present.
Progression within the group: from data analyst to Director of Project Portfolio Management and IT leader of a development cluster.
Tech Guilds: March 2023 to November 2024; more than twenty experts and over a thousand practical cases.
Experience assembling teams and organising the complete software delivery process.

The detailed chronology and team results are in the CV and case studies. The book draws on this experience where it helps explain how work is organised.

6.2. What remains outside the account

Corporate examples disclose no customer data, internal workloads, architectural details or contractual terms. Team results are not presented as a manager's sole achievements. Personal experiments with external models use personal, public or synthetic data, separately from employer systems.

Corporate experience does not automatically establish that a personal product is ready for production deployment. That requires its own checks, operational evidence and agreements with the prospective customer.

6.3. Personal products

LAF is a private personal product for an autonomous software development lifecycle with AI agents and mandatory quality checks. The author uses it himself and is still defining the commercial model.

Planograph AI is a personal tool for planning a large team's backlog, capacity, sprints, dependencies and risks in the context of the delivery process.

These products are R&D and evidence of practical work on organising development. Readiness for a particular deployment is discussed separately.

7. The ninety-day test

7.1. Objective

Test the compatibility of three systems, rather than the abstract appeal of cooperation:

a customer problem;
the team's ability to deliver a product;
minimum rules for shared decisions and contributions.

7.2. Outside the experiment

registering a large federation;
collecting contributions at scale;
training a frontier model;
buying a computing cluster;
recruiting members publicly;
promising shares before legal design;
using employer resources or data.

7.3. Required artefacts

a one-page experiment agreement;
a problem map and supporting interviews;
a log of material decisions;
roles and their boundaries;
time and expense records;
a data and AI-tool policy;
quality criteria;
a conflict procedure;
a final report deciding whether to stop, continue as a project or design an organisation.

7.4. Conditions for continuing

Continuation is justified when all of the following hold:

a customer signal has been confirmed;
one complete journey creates measurable value;
the solution's cost is understood;
roles allow progress without collective votes on operational matters;
participants agree on the record of contributions;
conflict and feedback have not broken the team;
the next stage can be financed without hidden unpaid labour.

7.5. Stop criteria

customer behaviour does not confirm the problem;
the team builds a platform instead of a narrow, complete journey;
material decisions depend solely on the author;
contributions cannot be discussed without personal conflict;
AI adds more review work than useful acceleration;
legal boundaries make the chosen scenario impermissible;
participants do not want to continue once the novelty wears off.

8. Open questions

Before any public launch, these remain unresolved:

Which legal form best fits the first product?
How should payment for current work be separated from participation in residual returns?
How should previously created personal intellectual assets be valued?
What candidate period is fair to both sides?
Which decisions need a simple majority, a qualified majority or unanimity?
How can a minority be protected without giving it a veto over every decision?
How is the value of an exit calculated?
Who answers to a client in a federation?
How is AI's contribution measured when the tool is not a member?
Which data and processes may be sent to external models?
At what scale is independent financial and governance review needed?
How can a failure be discussed publicly without disclosing customer secrets?

These questions are not gaps to conceal. They are the programme for the next investigation.

9. Main sources

The library was reviewed on 7 October; key claims and sources were rechecked on 19 October 2026. Each source has a purpose: to support an observation, introduce a concept or help examine an objection. Including a book does not imply agreement with every conclusion it reaches.

The complete manuscript contains 24 chapters. Maslow and hh.ru appear in part one; Kondratiev and Moore in chapters 6–7; Marx, Varoufakis, Piketty, Acemoglu and Johnson in chapters 9–12; Ostrom, cooperative principles and Laloux in chapters 14–16. Scholz is a further-reading direction in chapter 20. The notes distinguish reading a primary text from checking an edition and its open materials. A publisher’s description is not presented as a reading of the entire book.

People, labour and long-term change

Nikolai Kondratiev. The Long Waves in Economic Life, 1935, The Review of Economic Statistics, 17(6), 105–115. Article scan. A reference for discussing long-term economic change. The conclusion and a primary excerpt with methodological commentary, section XVII, were checked: the long-wave hypothesis does not give us a calendar for the next crisis. A full review of methodological disputes remains to be done; a wave diagram cannot establish the economy’s current phase.

Karl Marx. Capital, Volume I, 1867: the buying and selling of labour-power. Primary text, English translation, Chapter 6. This section has been checked: the capacity to work is distinguished from ownership of the means of production. It helps frame the question of rights to the result. It does not supply a ready-made payment formula for SoDelo. Chapter numbering varies by edition; the reference identifies the available translation.

Abraham Maslow. A Theory of Human Motivation, 1943. Original article in York University’s archive. Already used in “Can we just live a little?”. The article qualifies the order of needs and discusses their simultaneous partial satisfaction. The relevant sections have been read. The theory helps discuss a character’s life; it does not diagnose that character.

Gordon Moore. Cramming More Components onto Integrated Circuits, 1965. Intel’s historical account of the original article. In chapter 7 on technology: the subject is components on integrated circuits. The archival account has been checked; the relationship does not determine GPU prices, housing costs or the inevitable arrival of general AI.

Economics, ownership and platform power

Daron Acemoglu and Simon Johnson. Power and Progress, 2023. PublicAffairs edition. The question for our book is when technological progress improves working people’s lives. Bibliographic details and the MIT discussion with the authors have been checked. Chapter 12 uses the general argument about distributing benefits; a full analysis of the historical examples is not claimed. The work broadens the discussion of who benefits, but does not establish the superiority of every cooperative.

Thomas Piketty. Capital and Ideology, 2020, English edition. Harvard University Press. Used in chapter 11 on property institutions and unequal starting conditions. The edition and author’s LSE lecture materials, 2020 were checked. Reading the whole book or reproducing its data series is not claimed. The author’s tax proposals are not adopted as a ready-made project design.

Yanis Varoufakis. Technofeudalism: What Killed Capitalism, 2023. Publisher’s page, open sample. The preface presents the author’s hypothesis about a change in the economic system. For us, it invites scrutiny of platform dependence. The edition and open preface have been checked; the “end of capitalism” remains Varoufakis’s position, not an established fact of this book.

Self-management and shared resources

Frederic Laloux. Reinventing Organizations, 2014. Book website, open presentation based on the book. The presentation’s account of self-management, wholeness and evolutionary purpose has been checked. These are reference points for chapter 16 on governance. “Teal” does not mean “cooperative”: ownership rights need separate arrangements. An organisation’s colour does not establish its effectiveness.

Elinor Ostrom. Beyond Markets and States, Nobel Lecture, 2009. Lecture page, slides. Her book Governing the Commons (1990) belongs alongside it. These works help frame rules for shared resources and multiple centres of decision-making. The introduction and section IV of the published lecture, 2010 were read: boundaries, rule-making participation, monitoring, conflict resolution and nested governance. Chapter 14 distinguishes these findings from the author’s application to computing. Managing a natural resource and access to GPUs are not treated as identical problems.

Books and guides on cooperation

Trebor Scholz. Own This! How Platform Cooperatives Help Workers Build a Democratic Internet, Verso, 2023. Publisher’s book page, book discussion. Further reading for chapter 20: joint ownership, relations with unions and the difficulties of growth. The edition’s record and introductory discussion have been checked; a full case review remains ahead. Another project’s success does not establish demand for ours.

Hagen Henrÿ. Guidelines for Cooperative Legislation, ILO, third edition, 2012. Edition and open text. A comparative guide that helps frame questions about cooperative identity and legislation. The publication record has been checked. It is an international reading resource, not a source of current Russian law.

International Cooperative Alliance. Guidance Notes to the Cooperative Principles, 2015. ICA guide. A companion to the seven principles used in the manifesto. The guide’s publication has been verified; chapters 14–16 discuss working procedures as authorial proposals, not a ready-made application of the guide. A movement’s normative guidance does not replace a charter or measure financial resilience.

Observations used in the chapters already written

hh.ru, “The labour market in 2025: why finding a job became harder”. hh.ru article. Chapter One uses its comparison of office vacancies in June 2024 and June 2025. It does not measure time to an offer or describe the entire IT labour market in autumn 2026. Sergey’s and Alina’s scenes remain fictional stories, not survey findings.

Work and AI

The cooperative model

Technical basis for chapters 17–18

For chapter 17, the official GLM repository, DiLoCo paper, Decoupled DiLoCo, DeepMind, 2026 and Hivemind were checked. They establish the existence of self-hosting tools and distributed-training research, not the economics of a particular cluster.

For chapter 18, vLLM optimisation documentation was read: cache memory, concurrency, latency and throughput. The book’s 700 billion parameters and 100 load units are illustrative, not GLM specifications or server measurements. ICA history supports chapter 14 on Rochdale.

Russian legal framework

10. When to revise the manifesto

The position should change if:

new evidence shows generative AI systematically affecting employment differently from the current task-based account;
the experiment fails to confirm the market or management hypothesis;
legal review finds the principles incompatible with the chosen form;
actual participants propose a fairer, simpler model;
cooperation proves worse than options, a partnership or a conventional company for the chosen product.

A manifesto that facts cannot correct quickly becomes a faith. The aim here is a strong position open to testing.

11. A computing cooperative as a separate experiment

Working addition dated 2 October 2026. The engineering analysis “Cooperative AI without the myths” is being prepared for a separate publication. This section does not imply that infrastructure has been purchased or an organisation formed.

Hypothesis and boundary

A small group with non-coincident demand peaks may jointly own computing infrastructure and obtain a useful service at understandable cost. This is a stronger hypothesis than dividing a computer's price by the number of people: it includes solution quality, queueing, maintenance and workable rules.

A product artel and an infrastructure cooperative test different capabilities. The first product experiment requires no accelerator purchase. A shared service needs its own bounded pilot, initially on rented or already available hardware. A positive technical result does not replace a test of joint governance.

What to measure

The share of successfully completed tasks against an agreed baseline.
Input and output sizes, agent steps, retries and session duration.
Time to first token and complete response, including queueing, especially p95.
Behaviour with simultaneous human use and background jobs, including failures.
Average power use, operator time, repairs and recovery.
Total cost per successful task and monthly payments when membership is below plan.

Model specifications, measurements of third-party deployments and calculation assumptions must be labelled separately. No published tokens-per-second figure becomes a commitment to members until the relevant scenario has been reproduced.

Organisational arrangement

Capital, democratic votes, pay and resource consumption are accounted for separately. The administrator receives agreed compensation. Members can see how their quotas are calculated, shared expenses and the rules for temporary use of spare capacity. Other people's prompts are excluded from shared reporting.

A small resource fund can support approved shared projects, including LAF experiments. Each job needs an owner, a spending cap, an outcome criterion and rules for yielding resources to primary work. Continuous utilisation is not an objective in itself.

Owning hardware does not mean owning copyright in the models loaded onto it. Before launch, the group checks licences for every component of its chosen deployment, whether member access is permitted, data requirements and the legal form. Exit arrangements include share valuation, payment timing, and data export and deletion.

Conditions for continuing and stopping

The technical pilot runs for an agreed limited period, such as a month. The group continues the organisational test towards ninety days only if quality, queueing and budget are acceptable. A purchase is considered after complete hardware and hosting quotations, a recovery test, and agreement on access and exit rules.

If the model cannot solve the tasks, demand exceeds capacity, maintenance relies on unpaid heroics or costs cannot be covered, defer the purchase. Valid outcomes include a smaller service, a hybrid with cloud services, another model or abandoning shared hardware. Each is acceptable when supported by observations.

12. Federation, economics and participation

12.1. What was clarified on 6 October 2026

A federation is SoDelo's intended direction. A small pilot tests the first part of that model. Production, procurement, computing services and a work marketplace are connected through shared rules and infrastructure; a participant may be a creator, consumer and co-owner at the same time.

The project has a concept, economic scenarios, federation design, participation documents, user journeys and software implementation. These establish the extent of preparatory work. They do not establish legal registration, professional review of a charter, market demand or financial viability.

12.2. What the calculation in chapter 22 shows

The numerical example is developed in chapter 22, A difficult month: baseline, delayed payment, lower demand and reserve limits. The previous version remains in the archive.

All operations, obligations and the RUB 300,000 available opening balance in this test belong to one hypothetical operator. Funds held by separate federation members cannot be counted as its available cash without a separate basis.

The scenario with monthly external revenue of RUB 720,000 is a teaching example. Deducting RUB 360,000 for fully loaded labour, 100,000 for infrastructure, 60,000 for sales and payments, 45,000 for hypothetical other mandatory payments and 50,000 for depreciation leaves RUB 105,000. This is the result of assumed inputs, not a profit promise.

An internal transfer of RUB 120,000 between products and the computing operation is not added to external revenue. Member capital contributions are not sales. Halving external sales and only halving the sales-and-payment-cost line produces a RUB 225,000 loss. If instead half the baseline payments are delayed by a month, baseline outflows produce a negative net cash flow of RUB 205,000. Opening available cash of RUB 300,000 leaves RUB 95,000 at month-end, provided the calendar also allows intervening obligations to be met. That is an additional assumption in the fictional example, not an established reserve requirement.

Amounts exclude indirect taxes. For the cash-flow calculation, the four cash-expense lines, including the hypothetical RUB 45,000, are assumed paid in the same month: RUB 565,000 in total. The RUB 50,000 depreciation reduces the economic result but is not part of those payments. The hypothetical RUB 45,000 does not establish a tax rate. Internal sales are excluded from the overall management view but remain in separate entities’ accounts, with their taxes and obligations. Decisions require actual prices, contracts, load tests, full labour costs, a tax calculation and a cash-flow schedule. The table establishes neither the price of a GLM server nor an access tariff.

In the baseline month, recognised revenue and cash receipts coincide at RUB 720,000. In the delayed-payment case the work has already been performed under the example’s assumptions, so revenue and the economic result remain unchanged while RUB 360,000 becomes receivable. The baseline RUB 155,000 cash flow is not labelled EBITDA: the assumed payment categories do not establish that measure. Equipment purchases and financing inflows and repayments are outside this monthly cash flow and must be calculated separately. Depreciation also does not establish that cash has actually been set aside to replace equipment.

12.3. Legal wording

The historical analysis was reviewed for confusion between member numbers and votes, promises to return the initial contribution, blanket exemption of payments from insurance contributions, and attributing the application's P2P restriction to the law itself. The public text does not use these promises.

Exit arrangements should be checked against Article 22 of the production cooperatives law, and membership against Article 7. The Federal Tax Service describes which payments are subject to insurance contributions. Naming a payment does not establish its legal basis.

12.4. What the invitation means

The author invites discussion and preparation of specific areas of work. This is not fundraising or a promise of income. Early participation offers an opportunity to shape the enterprise and accept more responsibility; it does not automatically confer preferential future payments.

Social resilience is a purpose served by proposed shared tools, several sources of orders, owned products, learning and reserves. These mechanisms require earned resources. Proposed assistance must not be presented as existing insurance coverage.

Personal cooperative work is separate from an employer's resources and obligations. Businesses are potential customers and partners. Public principles and benefits should be clear; contracts, personal data and commercially sensitive information are shared only where required for the work.