Proof of Work
Miners spend real computation to propose blocks. Kaspa keeps this Bitcoin-like foundation intact.
Glossary
Kaspa glossary
The Kaspa terms you'll hit first, defined in a sentence or two and linked into the longer explainers.
Core mechanics
Live-network terms
UTXO
An unspent transaction output: a spendable coin chunk. Kaspa spends UTXOs; Ethereum edits account balances instead.
BlockDAG
A graph of blocks where multiple honest blocks can exist in parallel, instead of every block forced into one line immediately.
GHOSTDAG
Kaspa's live consensus rule for ordering the blockDAG so parallel work still counts toward consensus.
Selected parent
The parent block that defines the main ordering path through the DAG. Not the same thing as Bitcoin's longest chain.
Mergeset
The set of past blocks a new block pulls into view: how parallel work gets folded into the ordering.
Anticone
Blocks that sit in neither a block's past nor its future. Nodes need an anticone rule to order these consistently.
Blue / red blocks
GHOSTDAG marks well-connected blocks blue and less-compatible ones red, for ordering and security accounting.
Blue work
Accumulated proof-of-work weight carried by blue blocks. It's how Kaspa compares competing DAG histories.
Blue score
A blockDAG progress marker built from blue blocks in the selected-parent chain. Explorers and APIs use it for ordering and range queries.
DAA score
Difficulty Adjustment Algorithm score: roughly, how many blocks would have been produced to reach this point at the current difficulty. It counts accumulated work rather than wall-clock time, which is why hard forks are scheduled at a DAA score instead of a date. Toccata activated at 474,165,565.
Virtual block
The conceptual current tip nodes use to reason about the present state of the DAG.
Speed and storage
Scaling terms
BPS
Blocks per second. Kaspa runs at 10 BPS in the current era, which is a different number from transactions per second.
Fast inclusion
How quickly a transaction lands in some block at all. At 10 BPS that is usually a fraction of a second, which is the part of "fast" people notice first.
Fast confirmation
Confidence that a transaction won't be reversed rises quickly. Not the same thing as appearing in a block quickly.
Finality depth
The threshold past which old history counts as settled under the protocol's assumptions.
Pruning
Discarding old block data that ordinary nodes no longer need for current validation.
Archival node
A node that keeps full historical data for explorers, audits, and historical queries.
MuHash
A compact commitment tool used to verify UTXO-set state and keep pruning safe.
Transaction mass
What a transaction costs the network, measured on three separate dimensions: compute (script and signature work), transient storage, and permanent UTXO-set growth. Each is normalized against its own block limit and the transaction is charged the largest of the three, not their sum. Blocks cap total mass, not transaction count, so one heavy transaction can crowd out several light ones.
Storage mass
A cost model that makes UTXO/state growth pay for the burden it puts on the network.
Sompi
The smallest unit of KAS. One KAS equals 100,000,000 sompi. APIs return amounts as strings so software doesn't lose precision.
Hosted API
An API run by a provider outside your own node or indexer. Fine for quick reads and prototypes; production apps need key handling, rate-limit handling, and a reliability plan.
API key
A credential sent with hosted API requests, usually in a header like x-api-key. Keep it out of client-side code and public repos.
Rate limit
A provider rule that caps request volume. Apps should read rate-limit headers, cache stable data, batch where possible, and back off on 429 responses.
Cursor pagination
Fetching long lists in stable chunks. Address history endpoints use it so apps can walk through many transactions without one huge response.
Accepting block
The block that accepted a transaction into ordered history. Acceptance data is evidence, but apps still need confirmation and reorg-handling rules on top of it.
Roadmap terms
App and research terms
Toccata
Kaspa's live hard fork for spend rules, asset rules, ZK proof checks, sequencing commitments, and vProgs groundwork. It activated at DAA score 474,165,565.
Covenant
A rule that limits how a coin or asset can be spent later: the basis for vaults, escrow, and simple app state.
Silverscript
A higher-level language that compiles down to Kaspa script, so a covenant can be written as readable rules instead of hand-assembled opcodes. The repo is kaspanet/silverscript.
ZK verification
Checking a proof that computation was done correctly, without re-running that computation on L1.
Sequencing commitment
Lets an app track the activity lane it actually cares about, instead of forcing every proof to depend on the whole DAG stream.
vProgs
Apps that prove their own logic around Kaspa, while L1 stays focused on ordering, commitments, verification, and metadata.
DAGKnight
A parameterless/adaptive consensus research direction. Not current mainnet behavior.
RTD
Real-Time Decentralization: Hashdag's framing for Kaspa's core edge, Bitcoin-style PoW security and censorship resistance operating in real time. Oracle and TangVM ideas are downstream extensions of this.
Proof of Useful Work
A speculative research direction where mining work might also perform useful computation. Not live Kaspa functionality.