Kaspa GHOSTDAG: How BlockDAG Hits 10 Blocks/Second on PoW

The Problem GHOSTDAG Solves

In a linear blockchain, when two miners solve a block within seconds of each other, only one survives as the canonical chain; the other is orphaned, and its miner's work is wasted. Bitcoin's 10-minute block interval exists mostly to keep that collision rate low enough that the chain stays coherent. Shrink the interval and the orphan rate climbs until the chain fragments.

Kaspa's founders, Yonatan Sompolinsky, Shai Wyborski and Aviv Zohar, published the underlying protocol as PHANTOM and GHOSTDAG: A Scalable Generalization of Nakamoto Consensus (IACR ePrint 2018/104). PHANTOM generalizes Nakamoto consensus from a chain to a directed acyclic graph (a BlockDAG): blocks reference multiple parents, all valid blocks are kept, and a parameter k bounds how many concurrent blocks the protocol tolerates before treating a block as adversarial. Computing the optimal ordering under PHANTOM is NP-hard, so the paper also describes GHOSTDAG, a greedy approximation algorithm that reaches the same practical guarantees in polynomial time. The authors prove GHOSTDAG's block ordering becomes irreversible at a rate that decays exponentially with confirmation depth, the same security shape Bitcoin relies on, just applied to a graph instead of a chain.

From 1 BPS to 10: The Crescendo Hardfork

Kaspa mined at 1 block per second from launch until the Crescendo hardfork activated on May 5, 2025, which raised the network to 10 BPS and folded in a batch of Kaspa Improvement Proposals (KIPs), per the v1.0.0 "Mainnet Crescendo Release" notes in the rusty-kaspa repository. Rusty-kaspa is the current reference client; the original Go implementation, kaspad, is formally deprecated in favor of it and now only receives maintenance on its wallet component.

Ten BPS does not mean ten times Bitcoin's transaction throughput; on block-production rate alone, the table's figures work out to about 6,000 times Bitcoin's. What the number measures is block-production frequency: how many blocks the network produces per second while keeping concurrently mined blocks from competing destructively for the same chain slot. Actual transaction throughput still depends on block size and mempool policy, which is why Kaspa's roadmap treats BPS and TPS as separate levers.

What Actually Shipped: the Toccata Hardfork

The article you may have seen elsewhere calling this the "Covenant hardfork" was tracking an early working name. The upgrade that shipped is Toccata, released as rusty-kaspa v2.0.0 with mainnet activation scheduled at DAA score 474,165,565, roughly June 30, 2026 (release notes). As of this writing the network's live DAA score is 515,812,245, past that threshold, so Toccata is active on mainnet, not pending (kaspa mainnet DAG info, queried August 17, 2026).

Toccata brings native Layer-1 covenants: transaction-level spending constraints that let a UTXO restrict how it can be spent next, without a general-purpose virtual machine. That is enough to build escrow and other conditional-payment logic directly on Kaspa's base layer through a cluster of KIPs (KIP-16 for a ZK precompile opcode, KIP-17 for covenants and scripting, KIP-20 for covenant IDs, KIP-21 rounding out the set). The trade-off against an EVM-style chain is real: covenants constrain a single UTXO's future spend conditions, they do not give you arbitrary composable contract state, so anything requiring shared mutable state across many users still has to happen off the base layer.

For context, Kaspa's STRICT score sits at 76.6/100, driven by strong innovation (8.5/10) and solid tokenomics (7.8/10), offset by an ecosystem-maturity gap that no protocol upgrade closes by itself.

Layer 2: Where the Smart-Contract Composability Actually Lives

Because Toccata's covenants are deliberately narrow, general EVM composability on Kaspa runs on Layer 2, not Layer 1. The Igra Network launched on mainnet January 26, 2026, running EVM-compatible contracts with Kaspa miners acting as decentralized sequencers rather than a separate validator set. A second option, the Kasplex zkEVM (launched August 2025), offers Solidity support and MetaMask compatibility with bridged KAS as gas. Both L2s are described by their teams as building on Kaspa's base layer; this article has not reviewed either one's sequencing, bridge, or settlement architecture, so their finality guarantees are only as strong as the bridge and fraud/validity mechanism each one runs, a detail worth checking against each project's own documentation before treating either as equivalent to L1 settlement.

Supply Dynamics: Chromatic Halving, Measured as of 2026-08-16

Kaspa emits new supply on a "chromatic halving" schedule, reducing block rewards on a monthly curve instead of Bitcoin's four-year step function. Pulled directly from Kaspa's own node API as of this article's 2026-08-16 update, circulating supply stood at 27.64 billion KAS against a 28.70 billion max supply, or 96.3% mined (api.kaspa.org/info/coinsupply). The article's original prediction of roughly 95% mined by July 2026 landed close to the mark; the monthly reward curve has since pushed past it. Bitcoin, by comparison, crossed that same 95%-mined threshold first: its issuance is fixed by halvings every 210,000 blocks, and that schedule places the 19.95 million BTC mark (95% of the 21 million cap) at roughly late 2025, ahead of Kaspa's own milestone above.

Network hashrate measured through the same API, as of the same 2026-08-16 update, sat at approximately 334 PH/s (api.kaspa.org/info/hashrate). That is down from a reported peak above 400 PH/s earlier in the year, a reminder that PoW hashrate responds to miner profitability, which price influences alongside issuance, fees, and difficulty, not a one-way ratchet upward. KAS itself traded around $0.026 as of the same date, roughly 87.5% below its all-time high of $0.207 set July 31, 2024 (CoinGecko market data). With emissions now over 96% complete, most of the eventual supply is already out; the remaining issuance flow depends on the emission schedule, not on that stock figure alone. Whether any of that translates into price appreciation depends entirely on demand the ecosystem has not yet proven it can generate.

Exchange Access and Institutional Reach

KAS trades on more than 35 venues tracked by CoinGecko, including Bitget, Bybit, Kraken, KuCoin and HTX, but not on Binance or Coinbase spot markets, which still trade Kaspa derivatives only where offered (CoinGecko ticker data). A fair-launched asset with no venture allocation has no pre-existing token reserve to offer an exchange as a listing incentive. Whether that is what has kept the largest venues from listing it is an unverified hypothesis; no exchange has published listing criteria or a stated reason for passing on KAS, so this piece cannot attribute the gap to a specific cause.

How Kaspa's PoW Compares to PoS Throughput Chains

The obvious challenge: Solana, Sui and Aptos already deliver high throughput using proof-of-stake. What does a PoW chain buy that they do not already have?

Security model. Attacking a PoW network requires acquiring and running physical hardware, a cost the attacker has to pay whether or not they already hold any of the token. That physical-hardware floor does not make PoW security immune to price: honest-miner revenue depends on the token's price alongside issuance and fees, so a lower price can weaken security by squeezing miner revenue relative to costs. This piece is describing one specific economic-attack path, a stake-majority takeover priced against the token's market cap; it says nothing about compromised validator keys, bribery, or implementation faults, which are separate PoS attack surfaces this piece does not assess. Against that one path, what PoW keeps that PoS does not is the requirement that an attacker actually own or rent hardware rather than acquire a majority of the staked token. Kaspa demonstrates that a PoW chain can match PoS chains on block-production rate; as the BPS/TPS distinction above notes, sustained transaction throughput depends on block size and mempool policy, and no independently measured benchmark comparable to a PoS chain's TPS exists yet, so the throughput case for PoW over PoS rests on block production alone until one does.

Fair-launch distribution by construction. No pre-mine and no allocation round means there is no project-allocated early-investor cohort holding a disproportionate share of supply or governance influence, a structural difference from most VC-funded L1s regardless of how decentralized their validator sets eventually become.

DAGKnight, still ahead. The next consensus upgrade on Kaspa's roadmap, DAGKnight, targets adaptive, latency-aware block ordering that removes the need to hand-pick the k parameter for a fixed network-latency assumption. Kaspa's own developers have discussed scaling beyond 30,000 TPS once DAGKnight ships alongside further script-execution work, but that figure is a design target from the team building the upgrade, not an independently measured result, and it should be read as such until DAGKnight is live on mainnet and someone outside the project benchmarks it.

The honest trade-off is ecosystem depth. Solana and Ethereum carry years of accumulated DeFi tooling, audited contracts and liquidity that a faster consensus layer does not replicate on its own; Kaspa's Toccata covenants and its two Layer 2s are the beginning of that build-out, not a substitute for it.

Key Risks

Ecosystem gap. Two live Layer 2s is a small footprint; this article has no TVL or dApp-count comparison to size it against established smart-contract platforms. Toccata gives builders a new primitive; it does not by itself create demand for using it.

Exchange access. Absence from Binance and Coinbase spot markets caps the retail on-ramp that drives volume on most top-20 assets, and the fair-launch structure that many holders value leaves Kaspa with no token allocation to offer either exchange, though no exchange has stated that as a reason for not listing it (see above).

Execution risk continues. Toccata is active on mainnet, with the live DAA score past the activation threshold cited above, but DAGKnight is still ahead, and each consensus-layer change on a live PoW network carries real implementation risk: a bug in ordering logic is a security bug, not a cosmetic one.

Price has not recovered. KAS remains roughly 87.5% below its 2024 high more than a year later; this article has no daily price series around either the Crescendo or Toccata activations, so it cannot say what either did to the price. Whether that changes is a question about adoption, not about the protocol's technical soundness.

What to Watch Next

With Igra's January 2026 mainnet launch, the Toccata hardfork, and the 95%-supply milestone all now behind the network, the forward-looking questions are different from what they were when this piece was first drafted:

1. DAGKnight's mainnet timeline. This is the upgrade that would validate or undercut the 30,000+ TPS target discussed by the team; until it ships and gets independently benchmarked, treat that number as a design goal.
2. Layer 2 usage, not just launches. Both Igra and Kasplex are live; the metric that matters now is TVL and active-contract count on each, not another mainnet announcement.
3. Whether a major exchange lists KAS spot. No exchange has stated why KAS lacks a spot listing, so this piece cannot attribute the gap to a specific cause; it remains the single clearest catalyst the network does not control.

Kaspa's architectural bet, that proof-of-work security and high block throughput are not mutually exclusive, is no longer theoretical: GHOSTDAG's math has been running in production since 2021, and the Crescendo and Toccata upgrades are deployed, active upgrades rather than roadmap promises. What is not yet settled is demand: the protocol has solved a real technical trade-off, and the price sits far below its 2024 high; what either upgrade did to the price is a question this article's data cannot answer, and it has no TVL comparison to size its DeFi depth against PoS competitors.

Disclaimer: Nothing here constitutes financial advice; it is informational only. Cryptocurrency investments carry significant risk. Always conduct your own research and consult with a qualified financial advisor before making investment decisions.