Two mining pools produced 74.2% of Pearl blocks in a 4-hour, 33-minute sample on Oct. 6, a narrow snapshot on a network that aims to pair mining with artificial intelligence computing.
Pearl Coin Guide counted 120 blocks, numbered 124,040 to 124,159, from 8:27 a.m. to 1 p.m. ET (12:27 to 17:00 UTC). The guide attributed 50% of the blocks to Kryptex and 24.2% to PearlHash. The figures are shares of blocks found in that window, not measurements of the pools’ hashrate; miners can shift computing power between pools. Hashrate measures the computing capacity miners devote to finding blocks.
Pearl’s white paper presents the design as a way for operators to earn Pearl (PRL) alongside revenue from paid AI computing. Matrix multiplication is the work miners submit for blocks. The network also accepts calculations built from random inputs, so a reward does not establish that a customer requested the work. Pearl’s Proof of Inference layer, intended to support AI services, remains in early development.
In a June preprint, researcher Abhinaba Basu described a test miner that generated 44 shares from random inputs; a pool accepted each, according to the paper. The result illustrates a limit in using block rewards as evidence of paid AI demand: miners can earn PRL for calculations without showing that a customer paid for them. Customer revenue is therefore a key indicator of whether commercial work contributes to Pearl’s model.
Pearl Research Labs’ release notes classify the Aug. 6 mining-rule change as a soft fork and its Aug. 11 change as a hard fork.
Another possible change appears in draft PIP-3, which proposes moving mining to FP8 calculations and names NVIDIA Hopper and Blackwell hardware. If adopted as written, the specification would leave miners with other hardware architectures outside its scope. Kryptex says the proposal has no activation date.
