UEP | Universal Economic Protocol

Building UEP — Universal Economic Protocol.
Verifiable economic state transitions for asynchronous networks.
Earth → Space → Interplanetary. 🌍🛰️🌌

[UEP | Universal Economic Protocol @UEPCeo]

UEP's first public code is out.

A reproducible reference implementation of the testnet transaction model, plus a digital services marketplace layer.

Apache-2.0 · v0.3.1-public-preview

https://t.co/t877xM1n0P

[UEP | Universal Economic Protocol @UEPCeo]

Status: testnet reference implementation, not a live network or production system. Fees unchanged: 0.1% protocol, 3% Marketplace. Module map:
https://t.co/cBEvkt1xb3
Feedback and test reports: [email protected]

[UEP | Universal Economic Protocol @UEPCeo]

UEP v0.5.3 is out (testnet). A major integration release: the protocol is now organized into separate modules for settlement, swaps, relay, disputes, oracles and IoT/M2M, on one reference ledger. No native token, no external dependencies.
https://t.co/442R0TMazk

[UEP | Universal Economic Protocol @UEPCeo]

Also in v0.5.3: the Marketplace persists its full state in signed snapshots, ledger submissions go through a queue with backpressure, and the consensus labs now gate CI. Docs are reorganized by module, with a CONTRIBUTING guide. Versions are now plain numbers.

[UEP | Universal Economic Protocol @UEPCeo]

4/4 Why we read it: UEP signs client spends with Ed25519 (not post-quantum), and our lab replication applies a spend only after a 3-of-4 replica certificate. At n=4 the hard part is delay, not aggregation: DQS assumes partial synchrony; Earth–Mars round trips take many minutes.

[UEP | Universal Economic Protocol @UEPCeo]

Post-quantum BFT has a quiet bottleneck: quorum certificates. BLS folds n votes into 48–192 bytes; no known post-quantum scheme stays constant-size. Kniep, Sliwinski & Wattenhofer (Anza, ETH Zurich) sidestep aggregation instead of solving it. 1/4
https://t.co/iezfWU6zgS

[UEP | Universal Economic Protocol @UEPCeo]

3/4 Paper's numbers (good case, hash-based WOTS-TS one-time signatures): each DQS message is 624 B, while BLS certs grow with the signer bitmap. At n=10,000: 12.5 MB per node vs 14.1 MB for BLS. Simulated Alpenglow on ~750 Solana nodes: median finality 130 to 140 ms.

[UEP | Universal Economic Protocol @UEPCeo]

@gakonst @carllerche With a 500 ms tick, one thing that pays off in DST: route every time read, timer and RNG through an injected interface, and lint the real impls out of core code. Otherwise one stray now() call silently breaks seed replay.

[UEP | Universal Economic Protocol @UEPCeo]

For UEP this is a question, not a result: our v0.5.1 domain profiles add a Mars delay window (602 blocks, ~50 min), where no quorum can avoid the remote tail. UEP today is a local in-process testnet reference, not multi-node consensus, so this stays open research.

[UEP | Universal Economic Protocol @UEPCeo]

Paper: Hydrozoan (Yu, Kokoris-Kogias, Sonnino; HKUST-GZ, Mysten Labs, UCL). DAG BFT consensus with two commit paths under f Byzantine + c crash faults on n=3f+c+2p+1 validators: 2 message delays with ≤p faults, 3 otherwise, no view changes.
https://t.co/cfJFDRGhkI

[UEP | Universal Economic Protocol @UEPCeo]

Also notable: safety and liveness are machine-checked in Lean 4. The authors audit only definitions and theorem statements (~1k lines per protocol); the proofs were AI-generated and checked by the Lean kernel, not read by humans. A clear trust split for AI-assisted proofs.

[UEP | Universal Economic Protocol @UEPCeo]

@hillelogram Part of the "existing codebase" problem is hidden inputs: wall-clock reads, randomness and network calls buried in business logic. Before you can prove a function correct, it has to be a function. Even a lint banning those reads in core code makes later proofs less hopeless.

[UEP | Universal Economic Protocol @UEPCeo]

@kartikb753 @alex_alexiuc Agreed. One more piece: bind the key to the request content too. If the same key comes back with a different amount or payee, reject it rather than replaying the first result. Otherwise a buggy retry can look like success on terms nobody actually approved.

[UEP | Universal Economic Protocol @UEPCeo]

@medismera @elonmusk One layer above DTN that often gets skipped: application deadlines. A bundle can wait in custody, but a contract timer shouldn't silently expire meanwhile. Fixing delay windows ex ante (worst-case round trip plus margin) helps; conjunctions need explicit contact-gap rules.

[UEP | Universal Economic Protocol @UEPCeo]

Limits: a local single-node reference, not a network. Solar conjunctions and contact gaps aren't covered, and versions after v0.4.6 are listed as pending external review (docs/SECURITY-COVERAGE.md). Code: https://t.co/hTPY26st1R

[UEP | Universal Economic Protocol @UEPCeo]

How it's enforced: a CI lint bans Date, timers, fetch, network imports and randomness in transition code, and a "poisoned clock" test run makes those APIs throw mid-transition. The ADR says plainly these are guards, not a sandbox; review is still needed. https://t.co/8uYCv2e3bm

[UEP | Universal Economic Protocol @UEPCeo]

Listings also declare a domain profile: EARTH, MOON or MARS. It adds a fixed delay to counterparty windows: one worst-case round trip plus 25%, computed offline. Mars: max one-way light time 20.06 min over 2026-28 (JPL DE442s) gives 602 blocks of 5 s, about 50 min.