Build the Open Fork

1.Value without central permission. A person should be able to possess and transfer value without asking a central intermediary to authorize the transaction or preserve the account.

2.No unilateral ruler. No developer, miner, pool, company, custodian, or government should be able to redefine ownership or the shared rules for everyone by declaration.

3.Legible and predictable rules. Participants should be able to learn what the system accepts, understand material changes, and adopt them voluntarily.

4.Independent verification and honest exit. Participants should be able to verify what they accept and disagree or leave without being deceived about the history, rules, or asset they hold.

5.Resilience under failure and disagreement. The system should remain usable and recoverable when operators fail, coalitions split, networks partition, or adversaries apply pressure.

6.Voluntary adoption. A change earns participants by making its exact rules, costs, risks, and exits inspectable. Urgency does not substitute for consent.

7.The proposal closed after a split and stalled minority mining. BIP-110 version 1.0.1 was marked Closed on August 9, with the changelog explanation: a chain split with stalled mining. Closure describes this deployment, not every reduced-data proposal.¹

8.The minority branch later advanced. OCEAN's endpoint records four post-split blocks: 961632 and 961633 on August 8, followed by 961634 and 961635 on August 12, all solver-labeled Roughnecks. That is an endpoint record, not independent proof of beneficial ownership.²

9.Policy and consensus are different rule surfaces. Bitcoin Core 30's release notes specify changed default relay and mining policy for data-carrier outputs. BIP-110 specified additional consensus restrictions. Neither supplies a proof-of-work fork specification.⁴

10.Reconstruction is not completion. The evidence record is reconstructed but incomplete. Every known source has a disposition, while missing captures, unprocessed corpora, publication reviews, and all twenty human proposition-edge decisions remain visible gaps.

11.Restored has a strict meaning. A source is restored only when its bytes or a reproducible primary locator, identity, provenance, processing status, publication boundary, and relation to a proposition are retained. A filename in an inventory is accounted for but not restored. A captured source is retained but not therefore admitted. An edge recommendation is not a human decision.

12.The candidate graph stays visibly candidate. The research audit recommends eleven edges as supports, seven as bounds, one as insufficient and retargeted, and one as withdrawn from current claims. The canonical evidence manifest still records all twenty as unreviewed. This edition does not counterfeit the adjudication needed to cross that gate.

13.Grin's scheduled proof-of-work evolution. Grin's scheduled proof-of-work evolution is evidence of a protocol deliberately revising eligibility as hardware conditions change; whether it distributed beneficial ownership remains unestablished. Its historical schedule makes evidence-responsive revision thinkable. Predictable transitions can also favor specialized supply chains and impose unequal upgrade costs.⁵

14.RandomX and commodity processors. RandomX targets general-purpose CPUs through a large memory dataset, a virtual machine, and fast and light modes. That supports a commodity-processor eligibility mechanism, not an ownership outcome. Rentable compute, botnets, operating cost, and eventual specialization remain countercases.⁶

15.Mature commodity ASIC markets. Known equipment, supply history, resale, and established operators may widen day-one work availability. Manufacturing, finance, hosting, firmware, and common beneficial ownership can reconcentrate the same surface. Implemented proof-of-work mechanics do not decide that empirical contest.⁷

16.Multi-work rotation and capped lanes. Multiple eligibility lanes or a changing schedule may prevent one hardware class from becoming the whole network. DigiByte provides an exact multi-algorithm scheme to study, but hopping, weak-lane attacks, independent retarget rules, cross-lane fork choice, and common ownership must be modeled before the idea can travel beyond its source.⁸

17.Template, accounting, payout, and broadcast separation. DATUM documentation specifies a capability in which a gateway obtains templates from a local node, generates miner work locally, and submits solved blocks directly while a pool coordinates rewards without creating the template. Stratum V2 Job Declaration specifies another miner-side template declaration capability. Capability does not establish adoption, and distributing one role does not automatically distribute the others.⁹

18.Independent verification under bounded resources. Consensus limits can bound some validation work, and Bitcoin Core exposes the actual validation path. The hypothesis succeeds only if measured ordinary and low-power operators can validate, recover, and retain the chain under realistic workloads. A rule definition is not that measurement.¹⁰

19.Booted. Valid blocks continue from the declared split point, independent peers can recover from ordinary failures, and no hidden checkpoint or unilateral emergency key supplies the missing agreement.

20.Live. Block intervals, backlog, and reorganization depth remain within declared service bounds across arrival, departure, timestamp, and partition shocks.

21.Secure against a bounded adversary. The claim names attacker capacity, duration, information, cost, and objective. No open proof-of-work system is simply secure against an unbounded attacker.

22.Distributed control. Work ownership, template authorship, payout custody, peer connectivity, validation capacity, and economic-service access are measured separately and by beneficial owner where evidence permits.

23.Open participation. Entry cost includes hardware, memory, bandwidth, storage, power, operational knowledge, variance, market access, and exit liquidity. Being runnable is not the same as being meaningful participation.

24.Economically recognized. Wallets, exchanges, custodians, merchants, payment services, and users commit to an exact chain identity and safety procedure. Simulated prices may be conditions; they are not predictions of legitimacy.

25.What v1 is. The checked-in receipt compares idealized difficulty responses across stipulated hash shocks, then compares pool-only and optional local-template interventions across stipulated cost-latency worlds. It uses common-random-number seeds so paired differences are inspectable.

26.What v1 is not. It does not contain persistent persons or firms, endogenous ownership, capital and service markets, local information, strategic entry and exit, an empirically admitted population, or an adversary that chooses campaigns. Its ranges are not a live-network probability distribution, and external validation has not been performed.

27.Narrow DAA conclusion. DAA responsiveness changes short-run block progress after a stipulated hash departure; it does not determine whether a fork attracts hash.

28.Narrow template conclusion. Optional local construction can distribute template authorship only when miners' stipulated local value clears operating and latency costs.

Custody corollary. Custody and template authorship are distinct but potentially coupled control surfaces: shared plumbing can help, burden, or fail to attract miners.

29.Control is a lived relationship, not a pool label. A useful model must let beneficial owners, facilities, gateways, template authors, share accountants, payout custodians, broadcasters, and validating upstreams separate, recombine, fail, and recover through time.

30.The required object is an agent simulation. Miners, hardware owners, pool operators, template authors, validators, exchanges, custodians, merchants, attackers, and users need persistent identity, local information, resources, objectives, memory, relationships, jurisdiction, and explicit entry or exit. Ownership and service markets must evolve rather than arrive as static concentration constants.

31.Mechanisms must face endogenous response. A scheduled work change can alter investment. A local-template option can alter latency, defaults, and pool bargaining. A resource limit can alter who remains able to verify. An attacker can wait, rent, bribe, partition, or exploit the weakest lane. These are state transitions among actors, not labels on a scenario.

32.The first world is alive. One receipt-bound Agent-v2 pair now contains thirty-three persistent actors. In the constructive world, owners activated capacity, agents chose six distinct control relationships, a coalition paid to make an offer, messages moved through the topology, recipients acknowledged locally, and online work assignment chose the lots that produced blocks. Both paired worlds remained live with ten independent producers. This gives us an executable causal spine, a visible path for participation, and a real engine-to-figure instrument for designing the fork. The full matrix, finite ownership and hardware markets, payout default, verifier exit and recovery, chain reorganization, scheduled multi-work, bounded attack campaigns, and empirical review are the next construction frontier.

33.Editorial canonization is typed. Star Heartsong is the initial editorial canonizer: he may admit material and release an edition, but that role does not make him the source author, represented speaker, proposition adjudicator, model executor, or launch authority.

34.Speaker authority remains living. A represented living speaker remains the authority on what they mean. A quotation can support what was said. An editorial interpretation must remain labeled and directly correctable; engagement cannot stand in for consent or technical authority.

35.Corrections preserve lineage. A correction names the speaker or editor role, the exact proposition or passage, the requested change, the reason, and the affected edition. It creates a reviewable candidate. It never rewrites a frozen ancestor in place.

36.The invitation is presently a seed. All contribution routes remain pending; the current frontier is a seed invitation, not an open intake or an automated mutation path. Public chat may nominate evidence, questions, mechanisms, countercases, benchmarks, or corrections. It cannot submit, admit, merge, execute, or publish them.

1. 1

   Bring primary evidence.

   Identify exact bytes or a reproducible locator, provenance, date, publication boundary, and the proposition the source may support, bound, contradict, or leave unresolved.

2. 2

   Adjudicate the candidate graph.

   A named human reviews each exact proposition edge. The record preserves rationale, uncertainty, time, and later correction.

3. 3

   Specify mechanisms as interventions.

   Name the state transition, intended outcome, strongest countercase, measurement, and abort condition. Brand or community enthusiasm is not a mechanism.

4. 4

   Build and execute Agent-v2.

   Admit actors and ranges from evidence, publish the model and receipt, test reversals, and keep failure a possible result.

5. 5

   Open approved Git and non-Git routes.

   Each route needs a real receiver, identity and consent rules, fulfillment receipts, correction handling, publication review, and an idempotent state transition before it can be advertised as open.

37.Launch remains behind public evidence. No flag day is defensible before named economic services, actual work owners, executable safety rules, and measured low-cost validation cross public thresholds.