Archived Design Study: DAGKNIGHT and the Former Solidus Concept

Archive and status note

This article preserves a design direction explored under the Solidus name. It is not a description of a current product or network. The Lucrum monorepo's migration plan records a full rebrand from Solidus to Lucrum Finance on March 22, 2026.

Earlier versions said that a Solidus test network had measured approximately 100 ms finality and that DAGKNIGHT, POEM mining, a 16-chain hierarchy, and ForumCore were implemented. We do not have a public testnet explorer, reproducible benchmark harness, deployed-network identifiers, consensus code, or independent audit supporting those statements. They have therefore been removed. The architecture below is historical design context only.

What the DAGKNIGHT paper establishes

DAGKNIGHT, by Michael Sutton and Yonatan Sompolinsky, is a parameterless generalization of Nakamoto consensus over a block DAG. “Parameterless” has a precise scope: the ordering protocol does not hard-code an assumed upper bound on network latency. The client confirmation procedure still requires a locally chosen bound on recent adversarial latency.

The paper analyzes a proof-of-work protocol that tolerates an adversary controlling less than 50% of computational power. It does not specify stake-weighted validators, a two-thirds BFT threshold, or a Solidus-specific adaptive k range of 3–32. Those descriptions appeared in the earlier post without a supporting implementation or specification and should not be attributed to DAGKNIGHT.

DAGKNIGHT evolves the PHANTOM/GHOSTDAG line of research. Its contribution is a responsive ordering rule whose confirmation behavior can adapt to recent worst-case adversarial latency rather than relying on a single protocol-wide latency bound.

What the former Solidus concept proposed

The historical Solidus concept combined several ideas:

A hierarchy of 1 Prime, 3 Region, and 12 Zone chains
A multi-hardware mining concept called POEM
A proposed on-chain trading engine called ForumCore
DAG-based ordering intended to support low-latency settlement

These are design statements, not evidence of a functioning network. In particular, the DAGKNIGHT paper does not imply 100 ms finality, deterministic fee behavior, MEV resistance, or graceful behavior under every partition. Those properties would require a complete protocol specification, implementation, adversarial tests, network-scale measurements, and external review.

Why the original cascade-voting explanation was removed

The earlier article described blocks as stake-weighted validator votes whose descendants cross BFT thresholds. That is not the protocol presented in the DAGKNIGHT paper. DAGKNIGHT is an ordering construction for a proof-of-work block DAG; its security model is framed in computational power and its confirmation procedure is more involved than counting validator descendants.

Likewise, a DAG does not remove MEV by itself. Transaction selection, timestamping, propagation, proposer discretion, ordering rules, and application design all affect the MEV surface. Any claim that reordering becomes a two-thirds-stake consensus attack needs a protocol-specific proof.

What evidence would be required to revive the study

A future project using these ideas should publish:

The exact consensus specification and threat model
Source code and commit hashes for every benchmarked node
A public multi-operator testnet with explorer and network identifiers
Latency distributions under geographic delay, loss, churn, and adversarial load
Safety and liveness tests across supported client versions
Independent consensus and cryptographic review

Until then, the former Solidus performance figures should not be used in investor, customer, or technical claims.

Current direction

DDG's active work in this repository is described under Lucrum. This archived post remains available to document the research path and the distinction between a published consensus paper and a project-specific implementation claim.