DAGKnight
A collection of notes & links on DAGKnight, the proposed parameterless successor to GHOSTDAG.Overview
PHANTOM GHOSTDAG, Kaspa’s current consensus protocol, depends on a hardcoded assumption about network latency. The k parameter is derived from an assumed upper bound on block propagation time. If real latency is better than the assumption, confirmation speed is left on the table; the parameter must be chosen in advance, network-wide.
DAGKnight is a parameterless proof-of-work consensus protocol. Instead of assuming a latency bound, it continuously infers actual latency from the structure of the DAG itself and adapts. Confirmation times scale with the network’s real, current latency rather than a worst-case constant. The protocol tolerates attackers with up to 50% of hashrate.
The practical implication: as internet infrastructure and block propagation improve, confirmation speed improves automatically - no hard fork or parameter change required.
Status
DAGKnight is proposed in KIP-2. As of mid-2026 it remains under research and development and is not live on mainnet - Kaspa runs GHOSTDAG today.
Resources
- The DAG KNIGHT Protocol: A Parameterless Generalization of Nakamoto Consensus by Michael Sutton and Yonatan Sompolinsky (paper).
- KIP-2: Upgrade consensus to follow the DAGKNIGHT protocol.
- Kaspa Research Forum.