Kaspa

Yonatan Sompolinsky, Shai Wyborski and Aviv Zohar framed PHANTOM as a generalization of Nakamoto consensus. Instead of forcing every accepted block into one chain, a blockDAG records references among concurrently produced blocks. The ordering objective is computationally hard, so the paper also introduced GHOSTDAG, a greedy algorithm designed to approximate that structure efficiently.

The paper proves properties inside an explicit model; it does not establish perfect security under every network condition. Kaspa's practical wager is narrower and testable: less honest work must be thrown away when blocks arrive in parallel, allowing a higher block rate while nodes still derive a common order.

Mainnet opened on 7 November 2021. Kaspa's tokenomics record says there was no premine, preallocation, ICO or vesting reserve. The first two weeks used pseudorandom rewards of 1 to 1,000 KAS per block, followed by a 500 KAS-per-second phase. The chromatic phase began on 8 May 2022 and has reduced the per-second rate monthly ever since.

That schedule explains why supply needs careful wording. Code prevents issuance from reaching 29 billion KAS, but the expected total is calculated from integer rewards, the early random phase, DAA-score timing and changing block rates. On the review date, the official API reported 28,704,035,605 KAS as maxSupply. It is an operational estimate beneath the cap, not a contractual promise to the last coin.

The original node, kaspad, was written in Go. A community campaign raised 100 million KAS for a Rust rewrite; another 70 million KAS was assigned to DAGKnight work. The documented process ran through Discord discussion, votes and fundraising. A 3-of-6 committee multisig held those large funds, while the general public development fund used four elected treasurers and a 2-of-4 threshold.

This is meaningful public coordination, but it is not on-chain governance granted to every KAS holder. The process page itself says the scheme is subjective and not formally fixed. Multisig signers can release donated funds at milestones; they cannot make an incompatible node accept new consensus rules. Conversely, absence of a foundation mint key does not erase developer agenda-setting or mining-pool concentration.

Rusty Kaspa became the production implementation and the Go repository was deprecated. Crescendo's v1.0.0 release scheduled a hardfork from 1 to 10 BPS and moved peers to P2P protocol version 7. It was not a settings toggle: operators who wanted to remain on the shared network had to upgrade around a common DAA-score boundary.

Toccata repeated that pattern at a larger scope. The v2.0.0 mainnet release scheduled activation at DAA score 474,165,565, required P2P version 10 and added covenant primitives plus infrastructure for based ZK applications. By 4 September 2026 the official mainnet API reported virtualDaaScore 531,084,014, so the threshold had passed. Toccata belongs in the present-tense feature set, though applications built on it still need their own maturity and security review.

KIP-2 proposes DAGKnight, which would adapt to observed network delay rather than rely on GHOSTDAG's fixed bound. Its status remains Proposed. Research, earmarked funding and test work show intent; they do not make DAGKnight the current consensus or justify advertising its target latency as a live property.

Kaspa therefore has several different power centers. Miners contribute hashpower and choose templates; node and service operators select software; authors write the implementations and KIPs; multisig committees custody donated development funds. KAS holders can transact and miners earn subsidy and fees, but coin ownership alone does not confer equity, redemption, protocol veto or a formal vote. The fair launch removed an insider allocation table. It did not make coordination disappear.