
Kaspa developer Hans Moog, the main author of the vProgs code, has opened 13 bug reports and four proposed fixes in the vProgs code repository since Tuesday, most of them in the code that builds zero-knowledge proofs and stores app state. Six of the reports and two fixes were filed between Wednesday morning and Thursday morning. All were still open as of 8 a.m. ET on Nov. 19.
The wave of reports followed a large merge on Monday, Nov. 16, when 30 pull requests from developer Maxim (GitHub: biryukovmaxim) were merged at 9:33 a.m. ET. The code is a prototype and does not run Kaspa’s mainnet consensus.
What vProgs is
Kaspa is a proof-of-work network built as a BlockDAG. Instead of a single chain where competing blocks are thrown away, it keeps parallel blocks and orders them with a consensus rule called GHOSTDAG. The Toccata hard fork on June 30 added covenants, coins whose spending rules every node enforces, and an opcode that verifies zero-knowledge (ZK) proofs on the base layer.
vProgs, short for verifiable programs, is the planned way to run applications on top of that. Apps execute off-chain, provers produce a ZK proof that the work was done correctly, and a covenant on Kaspa checks the proof and records the new state. The kaspanet/vprogs repository, created in January, describes itself as being “in early development / prototype phase.”
Developers have been careful about the label. Michael Sutton has said the first apps should be called “standalone based zk apps” rather than vProgs, because they will not yet support synchronous composability between apps, according to a KaspaCom roundup that quoted him. Earlier this year Moog described the full design as four steps, with the current work on settling proofs on Kaspa’s base layer.
What the bug reports say
The most detailed report, issue #193, describes a batch of transactions that touch no application state. The prover wrongly treats it as proving an empty state root, then stops when that root does not match the stored one.
“Anybody who can get a transaction accepted on the lane’s subnetwork can do this, for the price of one transaction,” Moog wrote. He added: “As far as we can tell it is a liveness problem only.” A stuck prover stops the app from settling new batches, but the bad root cannot be settled over real state.
Moog opened pull request #201 with a fix on Thursday. Other reports cover a re-root request that “never ends, and nothing reports it above info level” (#195), deleted keys in the state tree that come back after later writes (#199, #200) and pruning that deletes live state nodes after a reorganization (#194). Some are marked as found by reading the code and not yet reproduced.
Community researcher vertex (@KaspaScopio) summarized issue #193 on Wednesday and put it in context: “This concerns experimental vProgs infrastructure, not Kaspa’s mainnet consensus. There is no evidence in the issue of an attack in production.” He noted that regression tests had not yet been added.
Why it matters
For holders, this is not a network risk. Kaspa nodes run the separate rusty-kaspa software, and nothing in these reports touches block production or KAS balances. It is, however, a view of how much work remains before vProgs-style apps carry real money.
For builders, the pattern is worth noting. Many of the bugs sit in recovery paths: restarts, rollbacks, cached proofs and reorganizations. That is where a system on a BlockDAG producing about 10 blocks per second gets stressed, and where an attacker would look. Finding them in public before launch is how the process should work, but the count shows the proving pipeline is not yet stable.
The reports also land as Kaspa’s first covenant apps are going live without that pipeline. Bridges such as Trestly’s kUSDT rely on human signers, and Kaspero Labs’ Kozmic Bridge uses its own ZK design.
What to watch next
The next step is whether the fixes in pull requests #184, #192, #197 and #201 are reviewed and merged, and whether regression tests are added. No release or testnet date for vProgs-based apps has been announced.
On the main node software, DAGKnight pull request #1172 in rusty-kaspa had no new commits since Nov. 17 and remained open on Thursday. Michael Sutton said on Tuesday that he still expects a DAGKnight testnet by the end of the year.
Sources
- https://github.com/kaspanet/vprogs
- https://github.com/kaspanet/vprogs/issues/193
- https://github.com/kaspanet/vprogs/pull/201
- https://github.com/kaspanet/vprogs/issues/195
- https://github.com/kaspanet/vprogs/issues/194
- https://github.com/kaspanet/vprogs/issues/199
- https://github.com/kaspanet/vprogs/issues/200
- https://github.com/kaspanet/vprogs/pull/197
- https://github.com/kaspanet/vprogs/pulls?q=is%3Apr+is%3Amerged
- https://x.com/KaspaScopio/status/2123082775095857538
- https://kaspa.com/learn-kaspa/kasmedia/weekly-knight-development-is-wild
- https://github.com/kaspanet/rusty-kaspa/pull/1172