In this issue: Kaspa's Toccata hardfork has been live for five weeks, so we look at what it changed and how it's being used so far. Plus the block reward decrease that came on Wednesday, a standards process opening in public, a language you should read but not build on, and three explainers worth keeping.

Deep-Dive

Toccata Is Live: Kaspa's Covenant Hardfork Has Been Running for Five Weeks

I think everyone already knows this, but... on June 30 Kaspa activated Toccata, the largest consensus change in its history. A pure proof-of-work chain became programmable.

Some vocabulary first, because most coverage skips it. Kaspa is a UTXO chain — as on Bitcoin, your balance is a set of discrete unspent outputs rather than a number in an account. A normal UTXO carries one rule: who is allowed to spend it. A covenant adds a second kind: how it may be spent, and what the outputs it turns into must look like. That single addition is what lets a UTXO chain carry persistent, stateful behavior without bolting on an account model, because the output itself carries the contract forward.

Toccata shipped three things at the consensus layer: 1. Native L1 covenant programming 2. Opcodes for verifying zero-knowledge proofs 3. Sequencing commitments.

It also changed the data model of a transaction, renaming the field previously exposed as mass to storage_mass and adding new computeBudget and covenant fields. Additionally, the minimum fee rate was raised from 1 to 100 sompi per gram. The specifications are KIP-16, KIP-17, KIP-20 and KIP-21.

What matters most here is the line between what shipped and what did not. What shipped is capability. Mature applications, audited contracts, broad wallet support and finished tooling are not shipped — they are downstream work in progress, and some of it is visibly still in flux, as several entries below show. Kaspa now has the primitives upon which smart contracts can be built, and the ecosystem has spent the last five weeks working out what to do with them.

If you run a node or an integration, the operational note is short: be on rusty-kaspa v2.0.1 or newer, and know that the database upgrade is one-way.

Also This Week

Covenant Usage on Kaspa Went From Dozens of Transactions a Day to Thousands

For the three weeks after Toccata activated, transactions creating a covenant output ran at a couple of dozen a day; since July 26 they have run in the low thousands to high thousands, swinging two- and threefold between days — 2,358 on August 5, against a high of 9,837 on August 1. The steadier figure is the covenant UTXO set itself, which has grown every day this month from 1,411 to 2,401, though all of this measures activity rather than distinct users, and a single busy deployer would look identical.

Kaspa's Block Reward Stepped Down on Wednesday, and It Was Not a Halving

The block reward is now 2.31246515 KAS, one step down from 2.44997148, and the ratio between them is exactly 2^(-1/12). Kaspa reduces the reward by that factor every month rather than halving it once a year, so it halves across twelve smooth steps instead of one cliff; the next step is due around September 4.

Kaspa Is Writing Its Post-Toccata Rulebook in Public

kaspanet/kccs — "Kaspa Calls for Conventions" — collects proposals for shared ecosystem conventions such as asset standards and wallet or indexer interoperability, explicitly not consensus changes, and it currently holds seven open ones from six different authors. Nothing has been adopted: the only pull request ever merged is the README, and main has had no commits since July 15, so read this as a process that exists and is being used rather than as standards that are settled.

Argent: An Actor-Model Language for Writing Covenant Apps on Kaspa

Argent treats a stateful covenant UTXO as an actor and a transaction as the message that triggers its next state change, and compiles down to Silverscript, the contract language Toccata-era covenants are written in. Michael Sutton has begun publishing introductory video episodes alongside this minimal starter template — worth reading rather than building on, since the repo states Argent is "not yet release-ready" and Silverscript has not completed its audit.

Can You Build on Kaspa With Just a Node, or Do You Also Need a Server?

Toccata made covenant_id a native field on a UTXO, but no node RPC will list the live UTXOs carrying a given covenant ID, so a covenant app has to run its own block-scanning indexer just to find its own on-chain state. A seven-post thread on kas-smiths.org has turned that gap into a wider architecture argument, and on August 4 it drew a concrete design proposal for a satellite index and two new RPC methods — discussion, not a numbered proposal, with no commitment from the core team behind it.

rusty-kaspa v2.0.1 Is Still the Release Node Operators Need to Be On

Nodes must be on v2.0.1 or newer or risk splitting off from the network, and the database upgrade is one-way — a node cannot be downgraded afterwards, only resynced. It is also still the latest release: nothing new has shipped on the node side since June 15.

Kaspa's Emission Schedule, Explained by the Project Itself

The official wiki's tokenomics page is the primary source for the mechanic almost every secondhand article gets wrong, and it also covers the fair launch, fees, the roughly 36-year emission tail, and an unusually honest section on why different sites report different maximum supplies. One number on it has aged: it projects "about 95% of all Kas" mined by July 10, 2026, where the live figure today is 96.24% — which is where the stale "95% mined" line still circulating comes from.

Why Four Hashrate Trackers Give You Four Different Kaspa Numbers

Hashrate is not measured but inferred from difficulty over a sampling window, so the length of that window and the moment you sample dominate the answer — on July 21 the same network measured anywhere between 221 and 307 PH/s on the same day. Single daily readings over the five days to Wednesday fell from 338.9 to 310.4 PH/s, which looks exactly like a downward trend right up until Thursday came back to 326.3, so treat any hashrate figure, ours included, as an estimate with a date attached.

"Kaspa Explained", a Plain-Language BlockDAG and GHOSTDAG Primer

If you have ever tried to explain Kaspa to someone and lost them at "BlockDAG", this free single page does it without maths: classic proof-of-work chains stay slow because one chain has to avoid honest forks, and Kaspa keeps the work while changing the shape, letting blocks reference several parents and having GHOSTDAG order what a chain would have thrown away. Its headline 10-blocks-per-second figure still holds — we measured 9.9992 blocks per second across five days of DAA scores this week.

Thanks for reading. This Week in Kaspa is new and the shape of it is still settling, so if there is a corner of the ecosystem you want covered — mining, tooling, governance, or the on-chain numbers — say so. And if you know someone who is Kaspa-curious but keeps bouncing off the jargon, forward this to them.