Kaspa and proofs of useful work became a research topic in July 2025 after Yonatan Sompolinsky shared a paper on permissionless matrix-multiplication work. The paper proposes a new consensus primitive; it does not announce a Kaspa upgrade, mining-algorithm change, or deployment schedule, so its relevance is conceptual rather than a confirmed roadmap commitment.
Key takeaways
- The underlying paper is Proofs of Useful Work from Arbitrary Matrix Multiplication by Ilan Komargodski, Itamar Schen, and Omri Weinstein.
- Its main construction targets matrix multiplication with
1 + o(1)multiplicative overhead over naive multiplication. - The authors conjecture optimal security; the wording is not a completed proof of every deployment property.
- Sharing the paper connected it to Kaspa research discussion, not to an approved KIP or hardfork.
- Any production proposal would need security analysis, incentives, hardware-access study, code, testnets, and an explicit activation process.
What triggered the Kaspa proof-of-useful-work discussion?
KASmedia’s July 14, 2025 report, Proof-of-Work: The Most Useful of Them All, noted that Kaspa researcher Yonatan Sompolinsky had shared a new cryptography paper. The report placed the paper alongside other ecosystem news, but it did not document a Kaspa implementation or governance decision.
The primary research source is the April 2025 preprint Proofs of Useful Work from Arbitrary Matrix Multiplication. Its authors revisit a long-standing question: can Nakamoto-style consensus use computation that also serves an external workload without giving strategic miners an unfair shortcut?
That is the precise connection. A recognized Kaspa researcher highlighted relevant work, and a community publication discussed it. Neither event changes deployed consensus.
What problem does the paper try to solve?
Conventional proof of work asks miners to perform a deliberately constructed computational task. The result has no independent customer beyond proving that scarce resources were spent, but that proof is useful to the network: it makes identities costly, supports leader selection, and raises the cost of rewriting history.
Proof of useful work tries to reuse some of that expenditure for an external task. The hard part is not finding a computation that people value. It is preserving permissionless security when miners choose their own task inputs, possess different information, and may optimize for the easiest acceptable instance rather than honest work.
The paper frames three requirements together: prescribed hardness, a certificate the protocol can check, and negligible overhead beyond the useful computation itself. If honest miners must spend substantial extra work while an adversary exploits a shortcut, the construction fails as consensus even if its output is commercially useful.
Why did the authors choose matrix multiplication?
Matrix multiplication is central to machine learning and industrial computation, creating a plausible external market. The paper supports arbitrary matrices and claims 1 + o(1) multiplicative overhead over naive multiplication, while considering faster but currently impractical algorithms.
The authors base their security conjecture on the hardness of solving batches of low-rank random linear equations. The word “conjecture” is important. A promising reduction and efficient asymptotic result justify further research; they do not by themselves establish secure economics, robust software, or decentralization under production conditions.
The preprint suggests a separate Layer 1 that reuses matrix work. That proposed system is not Kaspa.
Would useful work automatically improve a proof-of-work network?
No. External value changes incentives. Preferential workload access can lower one miner’s effective consensus cost, while task demand can fluctuate independently of network security needs.
Verification must remain predictable and cheaper than recomputation. Inputs, difficulty, stale work, and duplicates need deterministic treatment. Moving to AI accelerators might broaden one form of access while concentrating mining around scarce GPUs, data centers, or proprietary software.
“Useful” is therefore not a standalone security property. A design must show who supplies tasks, who pays for results, how rewards interact, and how adversarial miners are prevented from selecting favorable instances.
Was this part of Kaspa’s published roadmap?
The available evidence says no. The July report names a shared paper; it does not point to a Kaspa Improvement Proposal, Rusty Kaspa implementation branch, release, testnet activation, or hardfork date for proof of useful work. Kaspa’s deployed proof-of-work rules remained the relevant rules unless an explicit upgrade process changed them.
Readers can use an evidence ladder for future claims:
- A social post shows interest.
- A paper defines a research result.
- A KIP specifies a Kaspa proposal.
- Reviewed code shows an implementation.
- A public testnet supplies operational evidence.
- Release notes and an activation notice establish deployment.
Skipping levels turns research curiosity into misinformation. Builders who want to understand that process can start with the official-style Kaspa devnet workflow and the Rusty Kaspa workshop guide.
What evidence would a Kaspa implementation require?
A serious proposal would specify the useful task, certificate format, consensus verification rule, difficulty adjustment, block-template integration, reward economics, and behavior during absent or excess external demand. It would need adversarial analysis for task selection, withheld work, outsourcing, duplicate results, and specialized hardware advantages.
Engineering evidence would include deterministic implementations, benchmarks on diverse machines, reproducible test vectors, peer review, pool and wallet compatibility, and long-running testnets. Decentralization analysis should measure who can obtain hardware and workloads rather than assume that a general-purpose label guarantees access.
Only after those steps would an activation plan become meaningful. This is a much higher standard than community enthusiasm around a paper, as it should be for consensus code.
Does the research provide a KAS price signal?
No reliable one. Research visibility can attract developer attention, but it does not establish adoption, revenue, network security, or a scheduled feature. Price reacts to many variables that the paper does not measure.
The defensible conclusion is narrower: the Kaspa community was engaging with frontier proof-of-work research. Investors and builders should track primary artifacts rather than convert that engagement into a guaranteed technical or market outcome.
Frequently asked questions
Did the paper propose replacing Kaspa’s mining algorithm?
No. It proposed a general proof-of-useful-work construction and discussed a new Layer 1 design. The cited Kaspa coverage did not announce a Kaspa consensus change.
Is matrix multiplication easy for a blockchain to verify?
The paper develops a certificate and security approach intended to make the construction viable with negligible overhead. Production verification, incentives, and implementation still require separate evaluation.
Does useful work mean ordinary proof of work has no use?
No. Conventional proof of work supplies Sybil resistance and consensus security. “Useful” in this research context means that the same computation also serves an external task.
Source and verification note
The primary technical source is Komargodski, Schen, and Weinstein’s arXiv preprint 2504.09971, also recorded as IACR ePrint 2025/685. The dated Kaspa-community context comes from KASmedia’s July 14, 2025 report. Claims about Kaspa deployment were limited to what those sources demonstrate. This article makes no investment recommendation.
