Intercloud: Eventual Consistency for Decentralised Economies via Chilling-Effect Consensus

Gregory Magarshak · 2026

Read Intercloud: Eventual Consistency for Decentralised Economies via Chilling-Effect Consensus by Gregory Magarshak (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-22830.txt.

Introduction

Magarshak presents Intercloud, a decentralized economic network in which watcher swarms secure private data streams by observing only cryptographic hashes, never plaintext. Security rests on two ideas: ripple identifiers that stop any ripple repeating through a node in one epoch, and consensus by attesting that no conflicting evidence exists. A swarm of about 35 watchers, the paper argues, suffices to block double-spending, and conflicting attestations yield self-certifying proof of corruption.

Catalog record

CREATORGregory Magarshak
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE63969
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.22830

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Intercloud: Eventual Consistency for Decentralised Economies via Chilling-Effect Consensus Gregory Magarshak Abstract We present Intercloud , a decentralised economic network in which streams of private data are secured by Watcher swarms that observe only cryptographic hashes, never plaintext. Intercloud requires no global consensus beyond a single shared random seed per epoch. Security is provided by two mechanisms inherited from the Magarshak Machine [ 9 ] : (i) ripple deduplication via epoch-stamped identifiers, guaranteeing that reactive execution terminates without global coordination by preventing any ripple from rippling through the same node twice per epoch; and (ii) chilling-effect consensus , in which a swarm reaches finality by attesting to the absence of conflicting evidence , not by voting between alternatives. Any conflicting signed attestation automatically yields a self-certifying Proof of Corruption, making misbehaviour irrefutably detectable. We prove that execution ripples terminate in bounded time; that a swarm of approximately 35 Watchers assigned by a verifiable random function suffices for high-probability double-spending prevention, matching Hoepman’s lower bound [ 6 ] ; that two correct clients can hold conflicting finality attestations only if the adversary both compromises a supermajority of the assigned swarm and eclipses both clients from all honest nodes; and that Buridan’s Principle [ 8 ] does not apply because the swarm is finite and the consensus question is absence of evidence. Stream states are coloured Green, Yellow, or Red; end users choose their own finality threshold. The single global random oracle required costs O ⁡ ( N ) O(N) messages per epoch , not per transaction. Beyond the consensus layer, we develop a full economic model in which local coins are issued and retired by currency streams with pluggable exchange-rate formulae, and we prove that security weight tracks economic value automatically via a dot-product identity. The coin layer and content layer are strictly separated. I Introduction I-A The Blockchain Cost Problem Consider transporting $1 by armoured truck. A rational security model assigns one truck to one dollar. A blockchain does the opposite: every validator in the network must verify every transaction, regardless of its value. The global security budget — billions of dollars of electricity in proof-of-work systems, billions of staked tokens in proof-of-stake systems — is deployed uniformly for every transaction. A one-dollar transfer and a billion-dollar transfer consume the same validation resources. Beyond cost, the global ledger model has a privacy problem. Every transaction, every balance, every smart-contract invocation is visible to all participants. Transaction graph analysis can de-anonymise users even when addresses are pseudonymous. Intercloud addresses both problems simultaneously. Its central architectural choice is that Watcher nodes — the nodes that provide security — see only hashes of stream states, never plaintext. Security is allocated proportionally to the square root of economic value : a stream holding one dollar uses approximately 35 Watchers; a stream holding one million dollars uses \sqrt{10^{6}}\approx 1000 times more Watchers — roughly 35,000. This is the armoured-convoy model made precise: the cost of security scales sub-linearly with the value transported. I-B The Two Core Mechanisms I-B 1 Ripple deduplication The Magarshak Machine [ 9 ] prevents cyclic reactive execution by tagging each ripple with a unique ripple identifier ( \mathit{rID} ). Nodes store \mathit{rID} s in a hash table indexed by (\mathit{rID},\mathit{ep}) , retaining entries for two epoch durations before eviction. Because execution ripples carry fees at each hop, the table remains compact. This mechanism is proven correct for single-node Magarshak Machines; we extend it to distributed Intercloud in Section IV . I-B 2 Chilling-effect consensus A Watcher swarm for stream S_{i} reaches finality when a supermajority of its members has each independently verified via gossip that a supermajority of the swarm has attested to the same hash h and seen no conflicting hash during this epoch. Critically, this is not a vote between two candidate values. It is an attestation to the absence of conflict . If any two attestations conflict, both signers automatically produce a Proof of Corruption ( \Pi ) — a self-certifying evidence package requiring no trusted third party to verify. The chilling effect is the deterrence: a Watcher that produces conflicting attestations is immediately and irrevocably convicted. I-C Comparison with Existing Approaches I-C 1 Global-consensus systems Nakamoto [ 10 ] and proof-of-stake variants require every validator to process every transaction. The validation cost is O(N) per transaction where N is the network size. I-C 2 BFT protocols PBFT [ 3 ] achieves safety with f<n/3 Byzantine nodes but requires O(n^{2}) messages per consensus round. Tendermint [ 2 ] reduces this but still requires global participation per round. I-C 3 Hoepman’s result Hoepman [ 6 ] proved that using coin identifiers to assign clerk sets, double-spending can be prevented with clerk sets whose size is independent of the total number of nodes N , depending only on security parameters. For standard parameters this yields sets of approximately 35 nodes. Intercloud’s Watcher swarms are precisely these clerk sets, with stream identifiers replacing coin identifiers. I-C 4 Buridan’s Principle Lamport [ 8 ] proved that a binary discrete decision on a continuous-valued input cannot be made in bounded time. We prove in Section VIII that chilling-effect consensus is not subject to this principle because the swarm is finite and the decision is not binary. I-D Paper Organisation Section II reviews relevant prior work. Section III defines the Intercloud model. Sections IV – X state and prove the main results. Section XIII formalises the junior-node corruption detection and PoC lottery reward mechanism. Section XIV develops the vesting and rational security argument. Sections XV – XVII develop the local coin economy: emission, retirement, exchange rates, the dot-product security theorem, and the coin–content separation. Section XII presents the zero-knowledge extension. Section XIX discusses limitations and future work. Section XX concludes. II Background II- 1 Magarshak Machine A Magarshak Machine [ 9 ] \mathcal{MM}=(S,P,A,C,E,R) — the spacer model — is a formal model for governed, append-only distributed state. Streams S_{i} are append-only message logs under a publisher’s authority; R is a bidirectional relation index — two tables ( \mathit{relatedTo} , \mathit{relatedFrom} ) updated atomically when streams post relateTo / unrelateTo messages; actions A are policy-checked transformations declaring their read and write sets; capabilities C are the sole mechanism for side effects; policy P governs all transitions; and execution engine E is the push-only reactive scheduler driven by four reduction rules ( Create , Trigger , Execute , Retry ). The MM paper derives 22 structural theorems from these rules, including append-only safety, embarrassing parallelism scaling linearly with publishers, causal consistency, ripple termination via local deduplication, minimal cache invalidation, deterministic replay, and consensus freedom. The Magarshak Machine proves ripple termination for single nodes via the local ripple-log deduplication mechanism. We extend this to distributed networks in Theorem 2 . II- 2 Lamport’s causal ordering Lamport [ 7 ] established that in a distributed system without synchronised clocks, the happened-before relation \to defines a consistent partial order on events. Within each stream in Intercloud, message ordering follows Lamport timestamps; competing transfer requests are resolved by this causal ordering at the Executor. II- 3 Lamport’s Buridan’s Principle Written in 1984 and published in 2012 [ 8 ] , this principle formalises the arbiter problem: any physical mechanism that

Read the complete text