Pearl launched a UTXO chain where miners replace bare hashes with matrix products. The certificate verifies difficult arithmetic, while customer demand, inference provenance and a PRL holder's rights sit outside that proof.
On April 27, 2026, Pearl Research Labs made its code public and started the chain. A miner commits to an integer matrix product, reveals a selected tile and supplies a succinct certificate; nodes validate that evidence alongside Bitcoin-like UTXOs. The native asset is PRL, with no EVM contract or relationship to the similarly named vePEARL DEX token.
The protocol's strongest distinction is also its unresolved boundary. It verifies matrix multiplication, an operation used by neural networks, but does not require the matrices to originate in a customer's inference job. Pearl's whitepaper acknowledges non-useful work and no automatic exchange of PRL for compute. A June study reported that random matrices earned valid pool shares. Together AI's Pearl endpoint supplies one documented useful route, not provenance for the whole chain.
The code fixes a 2.1 billion PRL ceiling and 194-second target spacing. Rewards decline through a polynomial formula rather than sudden halvings; near block 650,226, roughly four years after genesis, about half the ceiling is still unissued. Miners collect subsidy and fees. Holders do not stake to validate, lock into ve voting or receive protocol revenue.
PIP-2 made grouped GEMM a Final hard-fork design. PIP-3 proposed floating-point certificates but was still Draft on September 5, 2026. The team controls the main repository and proposal publication; actual incompatible consensus depends on node and miner adoption. No reviewed code exposed a remote freeze or discretionary mint key for native balances.
Pearl Research Ltd.'s service terms reserve coins mined during a customer's hosted inference to the company unless the order changes that allocation. This narrowly defined commercial contract should not be projected onto public mining. Nor does it turn PRL into shares, revenue rights, compute credits or a legal redemption claim.
Pearl publishes the client and starts the PRL UTXO chain.
The partnership documents a Pearl-backed model service subsidized by expected PRL emissions.
PIP-2 later reaches Final status as a hard-fork specification for mixture-of-experts work.
PIP-3 introduces a hard-fork design but remains Draft at review.
Last evidence review: 2026-09-05
Pearl is a Bitcoin-derived UTXO network that replaces hash-only mining with verifiable matrix multiplication. PRL is its native coin, not a token contract: blocks mint PRL to miners and transactions spend it in units of 10^-8 PRL, called a grain. This identity excludes Pearl Exchange's DEX, RWA and vePEARL model and other projects that use Pearl or PRL names.
The reviewed protocol has no staking, ve-lock or token-weighted vote. A PRL balance is a transferable ledger balance; it is not documented as equity in Pearl Research Ltd., a claim on inference revenue, or a right to redeem AI compute.
Pearl asks whether proof-of-work energy can produce a reusable computation instead of a disposable hash. Its certificate lets nodes verify a large integer matrix multiplication without repeating all of it. That proves the arithmetic, but usefulness also depends on who requested the matrices and why.
Pearl's own whitepaper says the coin is not automatically convertible into useful compute and that the network can attract computation without useful work. An independent study of cuPOW reported accepted pool shares made from random matrices and argued that the current protocol does not authenticate a paid inference request. Together AI's Pearl-powered endpoint shows a real inference integration, but it does not make every accepted block commercially useful.
Miners compute matrix products and build a certificate from commitments and an opened tile. Nodes check the certificate, UTXOs, scripts and the chain's heaviest-work rule. Mainnet targets one block every 194 seconds. The supply ceiling is 2.1 billion PRL; subsidy follows a smooth polynomial curve rather than step halvings, with roughly half of the ceiling scheduled to remain unissued after 650,226 blocks, about four years. Miners also receive transaction fees.
Consensus changes are coordinated through Pearl Improvement Proposals and software releases. PIP-2 is a Final hard-fork specification for grouped GEMM, while the floating-point certificate in PIP-3 remained Draft on the review date and would also require a hard fork. The Pearl team can author proposals and reference code, but miners and nodes must adopt incompatible rules. The reviewed code did not reveal a contract-owner key that could freeze native PRL UTXOs or mint outside consensus.
Pearl Research Ltd.'s inference-service terms say coins mined while providing a customer's hosted service belong to the company unless an order says otherwise. That clause governs those customers, not every public miner. The same terms do not create equity, revenue share, compute redemption or governance rights for an ordinary PRL holder.
No. CoinGecko id pearl-2 identifies Pearl Research Labs' native PRL chain coin. It has no token contract, ve-lock or RWA claim.
It proves the specified matrix multiplication. The current certificate does not by itself authenticate a customer or economic use, and an independent study demonstrated accepted random-matrix shares.
No automatic conversion is documented. The whitepaper explicitly separates coin ownership from access to useful compute.
Consensus mints a declining block subsidy under a 2.1 billion ceiling and adds transaction fees to miner compensation. There are no step halvings.
The team can publish code and PIPs, but incompatible consensus changes require miners and nodes to adopt a fork. PRL balances do not cast protocol votes.
It provides control of spendable native coins. Reviewed materials do not grant company equity, revenue share, hosted-compute entitlement, guaranteed income or redemption.
Choose a tracking site for Pearl:
