Most crypto roadmaps are marketing documents: a horizontal line of confident quarters, none of them sourced. This one distinguishes four things — what has shipped and can be verified on-chain, what is actively in progress, what has been proposed without an activation score, and what is still research with no delivery date at all.
Kaspa went live with no allocation to founders, investors or a foundation. Rewards were randomised between 1 and 1000 KAS for roughly the first two weeks, then fixed at 500 KAS per second at the first hardfork.
The reward began stepping down by 2^(-1/12) every Kaspa month from an opening rate of 440 KAS per second. Twelve steps compound to exactly one half, so the network halves yearly without a revenue cliff.
Not a cosmetic port. The rewrite cut memory and CPU cost enough to make ten blocks per second runnable on ordinary hardware. Every upgrade since has shipped in it, and Crescendo would have been impossible without it.
Rather than argue about whether high-rate GHOSTDAG works, the developers ran it in public and let anyone measure it. TN11 and later TN10 produced the parallel-block and throughput data that Crescendo's parameters were tuned against.
Block time dropped from 1 second to 100 milliseconds. Per-block reward dropped by the same factor of ten, so emission per second was unchanged. Measured throughput rose roughly eight to nine times rather than a clean ten, because more parallel blocks means more transaction collisions between them.
Kasplex launched a zkEVM Layer 2 with EVM-compatible RPC, a two-way bridge to L1 and its own explorer, in collaboration with the Kaspa Ecosystem Foundation. Because it is EVM-equivalent, Solidity contracts port from Ethereum, BSC or Polygon largely unchanged.
Activated at DAA score 474,165,565, shipped in rusty-kaspa v2.0.x. Covenant opcodes and transaction introspection (KIP-17), covenant IDs for lineage tracking (KIP-20), zero-knowledge verification opcodes (KIP-16) and partitioned sequencing commitments (KIP-21). The minimum fee rate moved from 1 to 100 sompi per gram, and integrators had to accept transaction version 1.
KRC-20 began as inscription-style data indexed by Kasplex off-chain. With native asset support live on L1, migration is being coordinated by Kasplex and some tokens have their own steps. Follow the official guidance — this is exactly the situation impersonation scams target.
Community roadmaps discuss raising the block rate further, with 100 BPS (10 ms blocks) requiring changes to how miners select DAG tips as well as further node optimisation. No activation score exists for any of it yet.
GHOSTDAG fixes k in advance for an assumed worst-case delay. DAGKnight instead responds to the delay the network is actually experiencing, so confirmation tightens when conditions are good. Publicly discussed for years and pointed at a 2026-2027 window in community roadmaps — but until it lands in a KIP and a tagged release, treat dates as unconfirmed.
A design where contract state lives off the base layer while its correctness is verified on it, using the ZK opcodes Toccata introduced. An MVP track, not a deployable platform today.
Everything above is sorted by what evidence exists for it, not by how exciting it is. If you post about Kaspa development, using these distinctions is the cheapest credibility you will ever buy.
It is running on mainnet and you can verify it — an activation DAA score, a tagged node release, or on-chain history.
Work is publicly visible and ongoing, with no completion date that anybody can point to a source for.
Discussed in community roadmaps and developer channels. No activation score, no merged specification, no release.
A design or a paper. Might be years out, might not ship in its current form at all. Any date attached to it is speculation.
Toccata is a useful case study. Depending on when and where you looked, it was reported as activating in March 2026, May 2026, or June 2026 — and confident graphics circulated for each. The activation that actually happened was 30 June 2026 at DAA score 474,165,565, in rusty-kaspa v2.0.x. The lesson is not that people were dishonest; it is that a DAA score is checkable and a quarter on a slide is not.
None of these are announcements. All of them are observable, which is what makes them worth tracking instead of arguing about.
A migration involving user action, deadlines and third-party indexers is where things go wrong and where scams cluster. A clean completion is a strong signal about coordination capacity; a messy one is a lasting credibility cost.
Bridges are the most-exploited category of infrastructure in the industry. Both Kaspa L2s are young. Uneventful months are the evidence here, and there is no shortcut to accumulating them.
The subsidy halves annually, faster than any four-year-cycle network. Fees eventually have to carry security. Watch the ratio of fees to subsidy in miner revenue — it is the metric that decides whether the model works.
It has been discussed for years and would remove the hard-coded k parameter. Until it appears in a KIP and a tagged release with an activation score, every date attached to it is a guess — including the ones in community roadmap graphics.
Live from the network. If the node version above starts with 2, the mainnet you are reading from is running the Toccata consensus rules — which is a considerably better source than any roadmap graphic.
Most crypto roadmaps are marketing documents: a horizontal line of confident quarters, none of them sourced. This one distinguishes four things — what has shipped and can be verified on-chain, what is actively in progress, what has been proposed without an activation score, and what is still research with no delivery date at all.
Kaspa went live with no allocation to founders, investors or a foundation. Rewards were randomised between 1 and 1000 KAS for roughly the first two weeks, then fixed at 500 KAS per second at the first hardfork.
The reward began stepping down by 2^(-1/12) every Kaspa month from an opening rate of 440 KAS per second. Twelve steps compound to exactly one half, so the network halves yearly without a revenue cliff.
Not a cosmetic port. The rewrite cut memory and CPU cost enough to make ten blocks per second runnable on ordinary hardware. Every upgrade since has shipped in it, and Crescendo would have been impossible without it.
Rather than argue about whether high-rate GHOSTDAG works, the developers ran it in public and let anyone measure it. TN11 and later TN10 produced the parallel-block and throughput data that Crescendo's parameters were tuned against.
Block time dropped from 1 second to 100 milliseconds. Per-block reward dropped by the same factor of ten, so emission per second was unchanged. Measured throughput rose roughly eight to nine times rather than a clean ten, because more parallel blocks means more transaction collisions between them.
Kasplex launched a zkEVM Layer 2 with EVM-compatible RPC, a two-way bridge to L1 and its own explorer, in collaboration with the Kaspa Ecosystem Foundation. Because it is EVM-equivalent, Solidity contracts port from Ethereum, BSC or Polygon largely unchanged.
Activated at DAA score 474,165,565, shipped in rusty-kaspa v2.0.x. Covenant opcodes and transaction introspection (KIP-17), covenant IDs for lineage tracking (KIP-20), zero-knowledge verification opcodes (KIP-16) and partitioned sequencing commitments (KIP-21). The minimum fee rate moved from 1 to 100 sompi per gram, and integrators had to accept transaction version 1.
KRC-20 began as inscription-style data indexed by Kasplex off-chain. With native asset support live on L1, migration is being coordinated by Kasplex and some tokens have their own steps. Follow the official guidance — this is exactly the situation impersonation scams target.
Community roadmaps discuss raising the block rate further, with 100 BPS (10 ms blocks) requiring changes to how miners select DAG tips as well as further node optimisation. No activation score exists for any of it yet.
GHOSTDAG fixes k in advance for an assumed worst-case delay. DAGKnight instead responds to the delay the network is actually experiencing, so confirmation tightens when conditions are good. Publicly discussed for years and pointed at a 2026-2027 window in community roadmaps — but until it lands in a KIP and a tagged release, treat dates as unconfirmed.
A design where contract state lives off the base layer while its correctness is verified on it, using the ZK opcodes Toccata introduced. An MVP track, not a deployable platform today.
Everything above is sorted by what evidence exists for it, not by how exciting it is. If you post about Kaspa development, using these distinctions is the cheapest credibility you will ever buy.
It is running on mainnet and you can verify it — an activation DAA score, a tagged node release, or on-chain history.
Work is publicly visible and ongoing, with no completion date that anybody can point to a source for.
Discussed in community roadmaps and developer channels. No activation score, no merged specification, no release.
A design or a paper. Might be years out, might not ship in its current form at all. Any date attached to it is speculation.
Toccata is a useful case study. Depending on when and where you looked, it was reported as activating in March 2026, May 2026, or June 2026 — and confident graphics circulated for each. The activation that actually happened was 30 June 2026 at DAA score 474,165,565, in rusty-kaspa v2.0.x. The lesson is not that people were dishonest; it is that a DAA score is checkable and a quarter on a slide is not.
None of these are announcements. All of them are observable, which is what makes them worth tracking instead of arguing about.
A migration involving user action, deadlines and third-party indexers is where things go wrong and where scams cluster. A clean completion is a strong signal about coordination capacity; a messy one is a lasting credibility cost.
Bridges are the most-exploited category of infrastructure in the industry. Both Kaspa L2s are young. Uneventful months are the evidence here, and there is no shortcut to accumulating them.
The subsidy halves annually, faster than any four-year-cycle network. Fees eventually have to carry security. Watch the ratio of fees to subsidy in miner revenue — it is the metric that decides whether the model works.
It has been discussed for years and would remove the hard-coded k parameter. Until it appears in a KIP and a tagged release with an activation score, every date attached to it is a guess — including the ones in community roadmap graphics.
Live from the network. If the node version above starts with 2, the mainnet you are reading from is running the Toccata consensus rules — which is a considerably better source than any roadmap graphic.