On this page
Block 62,500, mined at 03:52 UTC on 13 November 2026, brought soft fork epsilon to Neptune Cash mainnet, and with it the foundation of succinctness: being able to verify the chain from its newest block alone, instead of downloading and replaying every block. The design has aimed at this from the start: the 2021 whitepaper already lists fast synchronization among its key features. Every block must now carry a proof that the block before it followed the rules. That earlier block carries the same kind of proof about the one before it, and so on, back to block 62,500. So checking the newest block’s proof checks all of them at once.
The chain of proofs starts at block 62,500. Earlier blocks are checked the old way, and no node syncs from the newest proof alone yet: both are covered under What is not here yet.
In short
- Most people need to do nothing. It is a soft fork: a node that never upgrades normally still follows the chain.
- Composers, the machines that build blocks, and exchanges need Neptune Core 0.20.1. A block built under the old rules is now rejected, but a node on old software would still accept it.
- What changed: each block’s proof now also proves the block before it, so one proof stands for the chain back to 62,500.
- Nothing to try yet: no software syncs a node from that one proof. Today’s node still downloads and checks every block.
Why it matters, and what changed
On Bitcoin, trusting the chain without trusting anyone else means downloading every block since 2009 and checking each one: about 778 GB by 25 October 2026. The job grows for as long as the chain lives, so the number of people who can do it shrinks. A Neptune Cash node that syncs from one proof would need only the latest block, at most 8 MB, at any age. That node has not been written yet; more on that below.
Neptune Cash blocks already carried a proof: a zk-STARK, which shows that a computation was done correctly without anyone having to redo it. Until block 62,500 it covered only the block’s own transactions. The other rules a block must follow had no proof: each node checked them directly. One of those checks was the update to the mutator set: the structure that holds every coin in sealed form and marks which ones have been spent. Epsilon adds three claims to what a block must prove:
- The block before it was valid. This is the recursive one: a proof that checks another proof.
- The composer knows the full mutator set as it stood after the previous block. A block can then only be built by someone who holds the chain’s data, so miners have no reason to build on a block whose data has been held back.
- The block’s spends were applied to the mutator set correctly. This is the check each node used to make directly.
The first claim is what links the blocks: each proof vouches for the proof before it, so one check reaches back through all of them.
The developers call the result a proofchain. It works like an auditor who signs off on last year’s audit along with this year’s books: check the newest certificate and you have checked the run.
That only works if one proof can verify another cheaply, which is what Triton VM and the Tip5 hash were designed for. Other chains stack proofs too, but most, Mina included, lean on elliptic curves to do it, the cryptography a quantum computer is expected to break. Neptune Cash’s recursion uses hash functions alone, and it runs on a chain where the payments themselves are hidden.
Why it is a soft fork and not a hard one
A hard fork relaxes or changes a rule, so blocks following the new rules look invalid to old software, and a node that does not upgrade falls off the chain. Neptune Cash has had five of those since August 2025.
Epsilon only adds requirements. A block that satisfies the new rules satisfies the old ones too, so an un-upgraded node keeps accepting blocks and normally keeps following the real chain. The catch runs the other way. A composer on old software builds blocks that no longer satisfy the network, and those blocks are rejected; but an old node still accepts them, so it can follow such a block until the network’s longer chain overtakes it.
What you need to do
- Wallet users: nothing. Sending and receiving are unchanged.
- Node operators: upgrading to 0.20.1 is recommended. An older node is designed to keep following the chain, but it does not check the new proofs. If a composer on old software builds a block, an older node accepts it while the network rejects it, until the valid chain overtakes it.
- Exchanges: upgrade to 0.20.1. A deposit credited in a block the network later drops would vanish, and an older node cannot tell such a block apart.
- Composers, and pools that build blocks: upgrade to 0.20.1. Blocks built under the old rules are rejected.
- Guessers, the miners running the proof-of-work lottery: if you guess with your own node, upgrade it too. Your work is unchanged, but guessing on a block the network will reject wastes it. If you mine through a pool or a mining service, nothing changes on your side; check that it has upgraded.
Version 0.20.0 carries the same rules; 0.20.1, released an hour later on 6 November, is the one published with downloads.
What it costs
Someone has to make these proofs, and that someone is the composer. The developers’ benchmark notes put proving a block at about 105 seconds if done step by step, some 90 of them new, or around 45 seconds with the independent parts running in parallel. Either fits well inside the 588 seconds the protocol aims to leave between blocks.
Composer figures from the project's own benchmark notes, at 15 seconds per large proof. Sequentially the same work takes about 105 seconds.
Guessing, the hash lottery that most miners actually run, is untouched. If you point a GPU at a pool, nothing about your setup changes.
What is not here yet
Going live is a beginning rather than a finish.
- No software syncs from one proof. Today’s node still downloads and verifies every block, and releases 0.20.0 and 0.20.1 deliberately keep validating that way.
- Blocks still carry data the proof does not need. Each spend brings along the chunks of the mutator set that it touches, and they are what keeps payments per block low: by the developers’ count, an 8 MB block fits only about 150 to 200 inputs, the coins being spent. The new proof lets a node rebuild the mutator set without reading those chunks, so a hard fork can drop them.
- Blocks before 62,500 sit outside the proofchain. A node verifies each block from 55,000 on directly, and takes blocks up to 54,999 on faith through a checkpoint built into the software, because those were proved under older versions of Triton VM. The newest proof names block 62,499, by its hash, as the block it is built on, so a future node that skips replay would have to trust that hash the same way.
What’s next
Two things follow, by the developers’ update of 9 November and the two issues they opened on 12 November to track the work.
A hard fork, planned for block 70,000. The issue that tracks it calls it zeta and gives it two purposes. The first is to drop the chunks from blocks. The developers aim for at least ten times as many inputs per block, and the issue puts what the chunks cost in throughput at a factor of 100 to 1,000. The second is to allow inputs in coinbase transactions, the ones that pay out a block’s reward, so that time-locked coins could be sold without trusting anyone. A draft is to be published for review first.
Faster syncing, in steps. The first step is to make the node use the proofs it now receives; its issue is titled “Use succinctness”. A new node would sync from the newest block and the mutator set, skip the proofs of old blocks, and download old blocks only if it keeps an index of them. The developers expect at least 1,000 times less to download, and a full sync in under a minute on most connections. It is not the whole way: the mutator set grows with the chain’s history, so what a node downloads still grows with it, at a small fraction of today’s size.
The recursion itself is in the rules now. The software that uses it is ordinary engineering, and a small team has to write it.
For how the pieces fit together, see how Neptune Cash works. Questions are welcome on the Neptune Forum and Telegram.
Sources
Soft fork epsilon, PR #988 · consensus rule sets · the commit that set block 62,500 · succinctness · the mutator-set update cost · Neptune Core 0.20.1 · Thoughts on the activation of succinctness, forum #421 · Dev update, 9 November 2026 · Hardfork zeta, issue #996 · Use succinctness, issue #997 · Neptune whitepaper (2021) · Bitcoin’s chain size, blockchain.com