Pearl

PEARL

PRL

Pearl (PRL) is the Pearl Research Labs Layer-1 blockchain using Proof of Useful Work: GPU matrix multiplication can serve AI workloads while participating in block production. This profile is the native-chain PRL indexed as pearl-2, not a similarly named application or token. Project documentation describes a UTXO accounting model and Taproot addresses beginning prl1p.Sourceshelp.pearlresearch.ai · What Is PearlGitHub · pearl-research-labs/pearlcoingecko.com · Pearl 2Observed 2026-10-14

EVERY CLAIM IS SOURCE-LINKED · “—” MEANS NOT VERIFIED, NOT ZERO · METHODOLOGY · REVISIONS

[OVERVIEW]

Mission

Build an AI-native monetary network in which matrix multiplications used for inference can also secure a native proof-of-work chain. The commercial thesis is that mining revenue can subsidize inference. PRL is not a contractual claim on GPU capacity, company equity, or inference revenue.Sourcespearlresearch.ai · Aboutpearlresearch.ai · Int Whitepaperpearlresearch.ai · Terms

Founders

Team

Pearl Research Labs identifies Omri Weinstein as co-founder and CEO. The foundational matrix-multiplication PoUW paper is by Ilan Komargodski and Omri Weinstein; paper authorship alone does not establish every author’s corporate role. The inference terms name Pearl Research Ltd. or the Pearl entity on the customer order, last updated August 5, 2026. Funding amounts, a complete founder/team roster, ownership and independently checked incorporation details were not established. NVIDIA Inception membership is a startup-support programme, explicitly not an investment or commercial partnership.Sourceshelp.pearlresearch.ai · Together AI Case Study The Fi…arXiv 2504.09971pearlresearch.ai · Termshelp.pearlresearch.ai · What Is The NVIDIA Inception …

Key milestones

  • 2025-04-14Foundational PoUW researchFirst arXiv submission of Proofs of Useful Work from Arbitrary Matrix Multiplication by Ilan Komargodski and Omri Weinstein. This is the research date, not mainnet launch.SourcearXiv 2504.09971
  • 2026-04-27Mainnet genesisPinned mainnet genesis timestamp is 2026-04-27 09:00:00 UTC. Genesis subsidy is zero; later blocks follow the declining schedule.SourcesGenesis.goValidate.go
  • 2026-05-15Together AI integration announcedGemma-4-31B-it-pearl inference offered at over 25% below the standard endpoint price. The partner describes the discount as funded by expected future crypto emissions, not guaranteed realized mining profit.Sourcestogether.ai · Together Ai Partners With Pea…help.pearlresearch.ai · Together AI Case Study The Fi…
  • 2026-09Floating-point scheme specificationSeptember 2026 whitepaper specifies FP8 useful work and an optional MoE extension. A published specification is not proof that its rules are activated on the current chain.Sourcepearlresearch.ai · Pearl Whitepaper
  • 2026-10-13Node v1.5.0 and desktop wallet v3.0.0Node release published at 14:48:18 UTC; desktop wallet at 16:57:19 UTC. Security hardening and safer rebroadcast handling are explicit release deliverables.SourcesRelease v1.5.0Release pearl-wallet-v3.0.0

[TECHNOLOGY]

CONSENSUS_MECHANISM
Proof of useful work (GPU matrix multiplication), Nakamoto-style longest chain
BLOCKCHAIN_TYPE
Native UTXO Layer 1 (btcd-derived)
BLOCK_TIME
194s
PROGRAMMING_LANGUAGES
Go: full node, wallet and SPV components., Rust: proof system, hashing and Python bindings., Python and CUDA/C++: GPU miner and serving integration., C: XMSS implementation; TypeScript frontend applications.SourcesREADMEREADME

Scalability & architecture

  • Succinct Plonky2/STARKy certificates let validators check the mining claim without repeating the complete GPU matrix multiplication. This optimizes useful-work verification; it does not imply recursive compression of the entire transaction history.SourcesREADMEpearlresearch.ai · Int Whitepaper
  • SPV light wallet uses compact block filters. No independently measured sustained transaction throughput or universal CPU/RAM requirement was established.SourceREADME

[TOKENOMICS]

MAX_SUPPLY
2.10B
EMISSION_SCHEDULE

Maximum nominal supply is 2,100,000,000 PRL. Let h be height and C=floor(4×365×86400/194)=650,226. The ideal cumulative schedule is S(h)=2,100,000,000×h/(h+C). The implemented subsidy is floor(2,100,000,000×10^8×C/((h+C)(h−1+C))) grains, with zero genesis subsidy and 10^8 grains per PRL. No discrete halvings; h=C reaches half the ideal cap. At height 127,874, ideal gross cumulative issuance is 345,116,823.03045881 PRL; summing the integer subsidy gives 345,116,823.02982092 PRL (16.4341% of the cap). This is a schedule-derived maximum subsidy issuance under the pinned rules, not a scanned coinbase total, UTXO balance or circulating-supply measurement. It assumes the schedule applied unchanged to earlier blocks; fees are transfers, not new issuance. Consequently total_supply and circulating_supply remain null.SourcesValidate.goParams.goAmount.goapi.prlscan.com · BlocksObserved 2026-10-14 21:59 UTC

TOKEN_DISTRIBUTION
Block producers
100%
Documented mining reward is the full subsidy plus transaction fees. Genesis subsidy is zero. This describes protocol rewards, not beneficial ownership: miners can sell rewards and pool payout addresses aggregate many customers. No complete launch-allocation or historical coinbase audit was performed.Sourceshelp.pearlresearch.ai · How Mining On Pearl WorksValidate.goGenesis.go
TOKEN_UTILITY

PRL pays native transaction fees and rewards successful miners with the block subsidy plus fees. The inference business can use anticipated PRL proceeds to discount service. The original whitepaper explicitly says mined tokens are not automatically convertible into compute; an on-chain compute marketplace is a proposed extension, not an established redemption right.Sourceshelp.pearlresearch.ai · How Mining On Pearl Workspearlresearch.ai · Int Whitepapertogether.ai · Together Ai Partners With Pea…

[CHART]

No verified TradingView market for Pearl Research Labs PRL yet. The PRLUSDT symbol on TradingView belongs to a different project (Perle), so it is not shown here.

[MARKET_DATA]

PRICE_USD
$1.95
Market Cap
$752.17M
Fully Diluted Valuation
$4.09B
24h Volume
$37.23M
24h Change
+6.48%
All-Time High
$2.46(Oct 21, 2026)
All-Time Low
$0.139936(Jul 23, 2026)

Exchange listings

SafeTrade (PRL/USDT, PRL/USDC, PRL/BTC)BigONE (PRL/USDT)Ourbit (PRL/USDT)

Last updated: Nov 21, 2026, 11:00 PM UTC

[ECONOMICS]

Trading pairs

PairExchangeSource
PRL/USDTBigONEBigONE public API ↗
PRL/USDTOurbitOurbit public API and listing announcement ↗
Exchange metadata identifies the base asset as Pearl. At 21:45 UTC ticker close was 1.9248 USDT and reported base volume 1,730,373.2984 PRL. The response does not supply quote turnover, so no exact USDT-volume claim is inferred. At 21:59 UTC best bid/ask were 1.8982/1.9019, spread 0.195%; returned top-20 orders contained about 4,117.80 USDT of bids and 5,850.97 USDT of asks within 2% of midpoint. This is displayed depth in a bounded response, not guaranteed execution or audited trading.Sourcesbig.one · Tickerbig.one · Asset Pairsbig.one · DepthObserved 2026-10-14 21:59 UTC
Official May 29 listing identifies native Pearl and its explorer. At 21:45 UTC /ticker/24hr returned last 1.9692 USDT, base volume 28029.49 PRL and quoteVolume 53635.6524 USDT. However openTime/closeTime span only 313.32 seconds, so this cannot safely be labelled verified 24-hour turnover. At 21:59 UTC bid/ask 1.9443/1.9632, spread 0.967%; top-20 displayed depth within 2% of midpoint about 139.31 USDT bids / 261.67 USDT asks. Match issuer and native network, not PRLUSDT alone: Perle and other PRL ticker collisions must not be merged.Sourcesapi.ourbit.com · 24hrapi.ourbit.com · Depthourbit.com · 17827791512989Observed 2026-10-14 21:59 UTC

[ECOSYSTEM]

Integrations

Pearl full node and OysterFirst-party node, wallet and light client

pearld, prlctl, oyster and oystercli are published in the monorepo; Oyster can use SPV compact filters. Desktop wallet v3.0.0 supports Windows, macOS and Linux release assets.SourcesREADMERelease pearl-wallet-v3.0.0

vLLM miner and pearl-gatewayFirst-party AI serving and mining

The vLLM plugin instruments compatible open-weight model linear operations. pearl-gateway bridges the miner to pearld RPC; model variants are linked from the Pearl Hugging Face organization.SourcesREADMEhelp.pearlresearch.ai · What Is The Pearl VLLM Miner

Together AICommercial inference integration

Partner announcement confirms Pearl-integrated Gemma inference and an anticipated-emissions-funded discount. NVIDIA Inception is separately a support programme, not evidence of NVIDIA investment.Sourcestogether.ai · Together Ai Partners With Pea…help.pearlresearch.ai · What Is The NVIDIA Inception …

Kryptex / KRigThird-party GPU mining pool and software

Pool publishes PRL Stratum connections and AMD/NVIDIA support. At capture it advertised PPS+ 3%, SOLO 1%, minimum payout 1 PRL. Pool-listed alternatives include SRBMiner-MULTI, PeakMiner, ForgeMiner, RGMiner and others; listing is not an independent binary audit.Sourcespool.kryptex.com · PrlGitHub · kryptex/krig-minerObserved 2026-10-14 21:45 UTC

PearlHash, HeroMiners, LuckyPool and other poolsThird-party pool ecosystem

Recent canonical blocks are attributed to these operators by PRLscan. An official Pearl-operated pool is still described as a waitlist item; old help wording that no pool is available must not erase independently operating pools.Sourcesapi.prlscan.com · Poolshelp.pearlresearch.ai · Is There A Pearl Mining PoolObserved 2026-10-14 21:59 UTC

BigONE and OurbitCentralized spot exchange integrations

Direct exchange APIs returned PRL-USDT and PRLUSDT books. Ourbit’s Pearl listing announcement links the native Pearl explorer, helping disambiguate this asset from Perle. Deposit and withdrawal operation was not tested.Sourcesbig.one · Depthapi.ourbit.com · Depthourbit.com · 17827791512989Observed 2026-10-14 21:59 UTC

dApps

Pearl Inference PlatformOff-chain commercial inference service

Official pricing page publishes serverless input, cached-input and output token prices. Service existence and posted tariffs do not prove audited customer demand, realized revenue, or that every request mines PRL.Sourcespearlresearch.ai · Pricingpearlresearch.ai · TermsObserved 2026-10-14 21:52 UTC

Pearl Compute Platform packagesHosted compute exposure, not staking

Help describes buying a share of compute and receiving variable proportional rewards. This is an operator-mediated service, not protocol staking, a fixed yield, or proof of permissionless on-chain application execution.Sourcehelp.pearlresearch.ai · How Compute Package Rewards A…

Developer metrics

GitHub Stars: 360
Forks: 88

Community metrics

Discord: 7,421

[UNIQUE_VALUE_PROPOSITION]

Core value proposition

The proposed differentiation is one GPU matrix-multiplication workload serving AI inference and blockchain mining, instead of a separate dedicated hashing workload.Sourceshelp.pearlresearch.ai · What Is PearlarXiv 2504.09971Observed 2026-10-14

Competitive edge

Analytical assessment: useful-work reuse could align miners with AI-compute providers. Whether this creates an economic advantage depends on actual workload, proof overhead, hardware efficiency, difficulty and PRL price; it is not proven by the architecture alone.SourcesarXiv 2504.09971GitHub · pearl-research-labs/pearlObserved 2026-10-14

Target audience

GPU operators, AI-inference infrastructure builders, node operators and users of the native PRL network. These are product-use categories inferred from official documentation, not measured user demographics.Sourcehelp.pearlresearch.ai · What Is PearlObserved 2026-10-14

[COMPARATIVE]

Competitive positioning

Pearl’s strongest distinction is attempting to reuse existing inference matrix work for Nakamoto-style block rewards, rather than renting GPUs through an unrelated token layer. Its key test is whether productive AI operators can compete with mining-only GPU operators after proof, numerical and integration overhead. The protocol verifies work, not commercial usefulness or semantic correctness.Sourcespearlresearch.ai · Int Whitepaperpearlresearch.ai · Pearl Whitepapertogether.ai · Together Ai Partners With Pea…

Strengths

Weaknesses

[ROADMAP]

Roadmap

Mainnet and node/wallet stackCOMPLETED

  • Mainnet genesis dated April 27, 2026; public node, wallet and light-client implementations are available.SourcesGenesis.goREADME
  • October 13, 2026 security release: node v1.5.0 and desktop wallet v3.0.0 published. Software delivery is complete; universal network adoption is not measured.SourcesRelease v1.5.0Release pearl-wallet-v3.0.0

Commercial inference integrationCOMPLETED

Floating-point PoUW and model coverageIN PROGRESS

  • September 2026 FP8 scheme paper and ongoing feature-branch work; production activation date was not verified. MoE, quantization correctness and reproducible arithmetic need separate implementation and consensus checks.Sourcespearlresearch.ai · Pearl WhitepaperIssue #424
  • Polymath research includes FP4, training integration, ASIC resistance and zero-overhead proof of inference. Open problems are not completed products or committed launch dates.Sourcepearlpolymath.com

Official pool and wider inference participationPLANNED

Verified compute marketplacePLANNED

  • Original whitepaper proposes a marketplace connecting useful compute supply and demand, potentially with native settlement. It is a design direction, not proof of deployed smart-contract settlement or guaranteed PRL redemption.Sourcepearlresearch.ai · Int Whitepaper

Upcoming milestones

Risk factors

Outlook

Bull case: public useful-work code, functioning mining infrastructure and an independently announced inference partner make Pearl a substantive attempt to connect GPU work with monetary security. Success would mean productive inference remains competitive with mining-only work and integrations survive without relying on rising coin prices. Bear case: a working chain can still become primarily a speculative GPU-mining market; unproven AI demand, numerical/cryptographic changes, issuance and shallow executable books can overwhelm the subsidy story. Verdict: technically credible enough for continued research, but adoption, net mining economics and token value capture remain unproven by these sources. Analysis only, not financial advice.SourcesREADMEtogether.ai · Together Ai Partners With Pea…pearlresearch.ai · Int Whitepaperpearlresearch.ai · Pearl Whitepaperapi.prlscan.com · Blocksbig.one · Depthapi.ourbit.com · DepthObserved 2026-10-14 21:59 UTC

[GOVERNANCE]

Governance model

Public-source, maintainer-led development with consensus changes expressed in node code and releases; operators choose which software to run. A token-weighted DAO, elected council and binding PRL governance vote were not established by the reviewed primary material.SourcesREADMEParams.goRelease v1.5.0

Proposal process

Technical proposals and changes are discussed through GitHub issues and pull requests. The pinned parameters contain explicit activation heights for MoE, dense-only proofs, rank penalties and salted seeds. Release v1.5.0 asks miners and pools to upgrade first because it tightens XMSS script validation. Repository participation is not proof of on-chain voting rights.SourcesREADMEParams.goRelease v1.5.0

[SECURITY]

Bug bounty program

Private disclosure through GitHub security advisories. Only the latest release is actively supported. Indicative awards: low 1,000 PRL, medium 2,000, high 10,000, critical 50,000+; all discretionary, not guaranteed. Eligible production scope is node/protocol, ZK proof/verifier, XMSS, wallet and SPV on master or latest supported release. Miner infrastructure, DNS seeder, frontend apps and unmerged feature branches are outside paid scope. No independent external audit report was verified; audits remains null.SourceSecurity

Security incidents

  • Oct 13, 2026v1.5.0 consolidates reported security fixes for untrusted input, remote crash/resource exhaustion exposure, validation, RPC transport and wallet handling.SourceRelease v1.5.0Impact: The release documents vulnerability classes, not a confirmed theft or successful mainnet compromise. It also changes invalid-length OP_CHECKXMSSSIG handling and mempool limits.SourceRelease v1.5.0Response: All node operators are urged to upgrade; miners and pools first. Desktop wallet v3.0.0 bundles fixes and preserves uncertain broadcasts as pending to reduce double-payment risk.SourcesRelease v1.5.0Release pearl-wallet-v3.0.0

Security practices

  • Proof-of-work certificates use hash-based proof machinery without a trusted setup. The security of arbitrary-input useful work remains dependent on the mathematical assumptions and implementation, not merely the presence of a ZK proof.Sourcespearlresearch.ai · Int WhitepaperarXiv 2504.09971
  • Post-quantum status is partial: the help centre says XMSS validation has been in consensus since genesis and a PQ transaction has been made, but the default Taproot/Schnorr spend path remains elliptic-curve based. No elliptic curves in the mining proof does not mean quantum computers offer no speedup.Sourcehelp.pearlresearch.ai · Is Pearl Post Quantum Secure
  • Open issue #424 concerns potential FP8 miner/verifier rounding mismatch on an experimental branch. The reporter gives a CPU reproduction but says the CUDA probe was not run. It is not established as an active-mainnet exploit.SourceIssue #424Observed 2026-10-14 21:59 UTC
  • Open issue #389 reports an XMSS final-index signing error. The report explicitly says no signing path ships in production and no direct mainnet impact; treat the report as a scoped implementation issue, not proof of broken deployed spending.SourceIssue #389Observed 2026-10-14 21:59 UTC
  • Open issue #425 reports macOS ARM64 wallet code-signing failure even when the download checksum matches. Open issue #364 reports persistent-peer removal reconnecting until restart. These are user reports, not independently reproduced here.SourcesIssue #425Issue #364Observed 2026-10-14 21:59 UTC
  • Keep wallet seeds off pool servers and mining rentals; pool mining requires a receiving address, not signing secrets. Use release checksums and keep RPC authenticated, TLS-protected and local or firewalled.Sourceshelp.pearlresearch.ai · Is There A Pearl Mining PoolREADMERelease v1.5.0

[REGULATORY]

Regulatory classification

Not established by reviewed sources; not a legal determination. Exchange listings and published software are not regulatory approval.Sourcescoingecko.com · Pearl 2GitHub · pearl-research-labs/pearlObserved 2026-10-14

Compliance status

Unverified. No licence, registration, legal opinion or issuer jurisdiction was established for Pearl Research Labs in this review.

[COMMUNITY]

[GITHUB_REPOSITORIES]

pearl-research-labs/pearlPRIMARY

Official monorepo for node, wallet, light client, ZK proving and mining. Stars, contributors and activity counts are intentionally not frozen.SourceGitHub · pearl-research-labs/pearlObserved 2026-10-14

[SOCIAL_MEDIA]

[COMMUNITY_PLATFORMS]

[TECHNICAL_DETAILS]

Technical specifications

Ports
Mainnet P2P 44108, node RPC 44107, wallet RPC 44207. Gateway normally uses a Unix-domain socket; optional TCP transport uses 8337. These are service ports, not public hosted RPC URLs. Keep authenticated RPC private.SourcesREADMERelease v1.5.0
Genesis
Mainnet header timestamp: 2026-04-27 09:00:00 UTC (1777280400). Genesis has no block subsidy.SourcesGenesis.goValidate.go
Vs Kaspa
Official documentation describes a live 10-block-per-second proof-of-work blockDAG that orders parallel blocks. Pearl targets one block per 194 seconds and optimizes reuse of matrix work, not high-rate DAG ordering. Blocks per second are not transaction throughput or a universal safety comparison.Sourceskaspa.org · LoreParams.go
Vs Qubic
Qubic documentation connects useful work to AI miners and Computor ranking by epoch. Pearl uses matrix-work tickets for block production and seeks reuse of arbitrary serving workloads. The shared AI label does not make their consensus, reward claims or hardware economics interchangeable.Sourcesdocs.qubic.org · Upowpearlresearch.ai · Int Whitepaper
Hardware
Official miner prerequisites include Python 3.12, uv, CUDA/NVIDIA GPU and Rust; its GPU test support says sm90 H100/H200. This is not the hardware floor for all PRL mining: KRig advertises AMD and NVIDIA consumer/datacenter cards. CPU-only validation and wallet use must be separated from economically competitive mining. No tested universal RAM, CPU-core, disk or VRAM minimum is supplied here.SourcesREADMEGitHub · kryptex/krig-minerREADME
Vs Monero
RandomX is optimized for general-purpose CPUs and deliberately discourages specialized hardware. Pearl targets GPU matrix multiplication and AI-serving reuse; it is not a privacy substitute for Monero.Sourcesgetmonero.org · RandomxREADMEpearlresearch.ai · Int Whitepaper
Vs Bitcoin
Conventional hash-based PoW, UTXO payments and probabilistic settlement; Pearl changes the work function to matrix multiplication, adding certificate and numerical-correctness complexity rather than merely renaming a hash.Sourcesbitcoin.org · Bitcoinpearlresearch.ai · Int Whitepaper
Difficulty
Target block interval 194 seconds; WTEMA updates difficulty per block with a 168-hour half-life. Actual block times are stochastic. Coinbase maturity is 100 blocks; maturity is not irreversible finality.SourcesParams.goDifficulty.go
Vs Parano1d
NOID v2 targets 30-second blocks and succinct live-state verification. Its pool GPU workers search nonces while the producing node constructs proofs. Pearl certificates prove matrix work; Pearl does not claim the same history-independent state bootstrap.SourcesParametersMiningpearlresearch.ai · Int Whitepaper
Chain Design
Native UTXO Layer 1 derived from btcd, btcwallet and neutrino, with its own useful-work certificates, monetary schedule and address format. It is not an ERC-20 token or an EVM rollup.SourcesREADMEpearlresearch.ai · Int Whitepaper
Light Client
The SPV client uses BIP157/BIP158 compact block filters. It synchronizes block headers and filter headers, fetches filters lazily, and fetches blocks on demand without persisting them. Rescan notifications concern confirmed transactions, not mempool arrival. This is a light-client architecture, not a claim of full-node-equivalent verification.SourceREADME
Live Network
PRLscan returned canonical height 127,874, hash b8c99bd23b93b9c84ebec3eb0c45b2d5d0afff2576c050b34a76658aac4f6a7d, timestamp 2026-10-14T21:54:50Z, difficulty 35,528,259.48849882, bits 1778e30d, 1 transaction and 1 confirmation in the response. Data time trails retrieval time; this is the explorer’s sampled tip, not a guarantee of the latest global head.Sourceapi.prlscan.com · BlocksObserved 2026-10-14 21:59 UTC
Mining Setup
Documented, not executed: obtain a verified release and keep the wallet seed local; generate a receiving address; for pool mining use KRig with the pool’s published TLS Stratum endpoint and your address/worker. Kryptex example: krig-miner --url stratum+ssl://prl.kryptex.network:8048 --user <address>/<worker>. For solo serving, build/install the official stack, run pearld with --miningaddr, connect pearl-gateway to authenticated local RPC, then serve a compatible Pearl model with vllm-miner. Verify accepted shares/payouts or canonical block rewards rather than trusting a local speed display.Sourcespool.kryptex.com · PrlGitHub · kryptex/krig-minerREADME
Address And Unit
Native PRL on a separate UTXO chain; 1 PRL = 100,000,000 grains. Default receive addresses use prl1p Taproot. An EVM chain ID or token contract address is not applicable to this native coin.SourcesAmount.gohelp.pearlresearch.ai · Understanding Pearl Addresses…pearlresearch.ai · Int Whitepaper
Issuance And Age
The observed block is 170.54 days after the code genesis timestamp, not a year-old chain. Current scheduled subsidy is 2,255.34691211 PRL, matching the sampled no-fee coinbase reward field at height 127,874. Target-time daily issuance at this instantaneous reward would be 1,004,443.16 PRL before subsequent decline. Over the next 162,556 target-time blocks, the declining schedule adds 303,263,668.88 PRL, or 87.87% of today’s schedule-derived 345,116,823.02982092 PRL denominator. This is a forward scenario at target cadence, NOT historical annual supply growth, circulating inflation or a price forecast.SourcesValidate.goParams.goGenesis.goapi.prlscan.com · BlocksObserved 2026-10-14 21:59 UTC
Latest Releases
Latest captured node release v1.5.0 published 2026-10-13 14:48:18 UTC; desktop wallet v3.0.0 published 2026-10-13 16:57:19 UTC. The earlier desktop v2.0.7 was published September 30. Release availability is not a measurement of network upgrade adoption.SourcesRelease v1.5.0Release pearl-wallet-v3.0.0Release pearl-wallet-v2.0.7Observed 2026-10-14 21:59 UTC
Pool Versus Solo
A standalone pool miner such as KRig connects to Stratum with a receiving address. It does not need the user to run pearld, oyster, pearl-gateway or a vLLM model server on that rig; the pool provides block-production infrastructure. The GPU still performs Pearl-specific work, not NOID’s generic nonce-only loop. Precise division of certificate construction varies by pool implementation and was not audited. Solo useful inference follows the separate node-plus-gateway-plus-vLLM path.SourcesGitHub · kryptex/krig-minerpool.kryptex.com · PrlREADME
Source Snapshot
Protocol source pinned at 837c69c52a831fff5fa733a358da8b3ae7d8c35c. Live explorer/orderbooks captured at 21:59 UTC; ticker and community captures at 21:45 UTC on 2026-10-14. This is research from public sources, not a locally synced-node attestation.SourcesREADMEapi.prlscan.com · Blocksbig.one · Depthapi.ourbit.com · DepthObserved 2026-10-14 21:59 UTC
Consensus Detail
Nakamoto-style proof of useful work: miners commit matrices, derive challenge-bound noise and obtain lottery tickets from matrix multiplication. Winning work is checked through a block certificate rather than treating a conventional header nonce as sufficient proof. The useful calculation is matrix multiplication, not a proof that an AI answer is true or that a customer paid for it.Sourcespearlresearch.ai · Int WhitepaperREADME
Mining Economics
PRLscan estimates trailing 24-hour gross yield 19.643211 PRL per PH/s-day from observed reward/work data. For illustration only, applying Kryptex’s 3% pool fee and its published 220 TH/s RTX 5090 example gives 4.191861 PRL/day; a 130 TH/s RTX 4090 example (one quarter of its four-card figure) gives 2.477009. This combines non-simultaneous sources and excludes miner developer fees, rejected shares, electricity, cooling, hardware depreciation, payout delays and sales slippage. Future difficulty and PRL price can change the result; it is not a measured rig return or an investment recommendation.Sourcesapi.prlscan.com · Poolspool.kryptex.com · PrlObserved 2026-10-14 21:59 UTC
Network Hashrate
PRLscan’s 24-hour block-implied network estimate is 57.2168 EH/s in its Pearl work units; aggregate reported pool hashrate averaged 55.4963 EH/s. These are different estimates, not independent measured GPU counts, and are not comparable to Bitcoin or Monero H/s.Sourceapi.prlscan.com · PoolsObserved 2026-10-14 21:59 UTC
Contributor Caveat
The repository inherits upstream btcd, wallet and SPV history. Raw GitHub contributor totals therefore cannot be presented as the active Pearl team; no filtered active-contributor count is published here.SourceREADME
Build Prerequisites
The source build lists Go 1.26+, Rust, a C compiler, Task, Python 3.12 and uv; the vLLM miner also needs CUDA. Prebuilt node/wallet installation is documented for Linux, macOS and Windows. These are software prerequisites, not verified minimum CPU, RAM, disk, GPU model or VRAM requirements.SourcesREADMEContributing
Development Process
Contributing guidance requests focused pull requests from master, passing CI, formatting/lint checks, regression tests for bug fixes and certification of the right to contribute under ISC. This describes code contribution, not token-holder voting, governance quorum or treasury control.SourceContributing
Hardware Benchmarks
Pool and miner marketing benchmarks conflict across software/configurations: Kryptex page illustrated RTX 5090 at 220 TH/s and four RTX 4090 at 520 TH/s; KRig README listed RTX 5090 around 385 TH/s and RTX 4090 around 292 TH/s, explicitly warning examples may be outdated. These are operator-published examples, not measured results for this dossier. No power draw or profitability benchmark was performed.Sourcespool.kryptex.com · PrlGitHub · kryptex/krig-minerObserved 2026-10-14 21:45 UTC
Miner Concentration
PRLscan reported 24-hour pool shares with a denominator inferred from its block/share fields of 498 total blocks: Kryptex 255 blocks (51.20%); PearlHash 124 blocks (24.90%); Hero Miners 28 blocks (5.62%); LuckyPool 25 blocks (5.02%); Pearl Fortune 25 blocks (5.02%); Pearlski Pool 19 blocks (3.82%); F2Pool 10 blocks (2.01%); AlphaPool 1 blocks (0.20%). Named pools account for 487 blocks; 11 are outside those named totals. Kryptex’s measured block share slightly exceeds half in this window, but block luck and switchable customer hash matter. Pool labels and payout addresses do not identify beneficial owners. A few dominant pools are common in PoW; this snapshot alone is not evidence of an attack, a single miner controlling the network or an abnormal permanent concentration.Sourceapi.prlscan.com · PoolsObserved 2026-10-14 21:59 UTC
Wallet Architecture
Oyster is a single-user HD wallet daemon using BIP32 key derivation and a BIP44-style derivation path. Its README states that unencrypted private keys are never written to disk and optional encryption also covers public wallet data. It supports pearld over WebSockets or an SPV backend, a default legacy JSON-RPC server and an experimental, feature-gated gRPC API. These are documented properties, not an independent wallet security audit.SourceREADME
Certificate Tradeoff
The INT specification describes a variable-size block certificate with a 65 KB limit and recursive proof compression below 60 KB. These are design figures, not measurements of the sampled block API’s size field. Certificate production is additional software and proving work; succinct verification is not zero-cost production.Sourcepearlresearch.ai · Int Whitepaper
Known Evidence Limits
No independently audited circulating supply, complete historical issuance reconciliation, sustained TPS, global node count, beneficial-owner distribution, corporate funding round, current exchange withdrawal test or installed GPU benchmark was established. The reviewed primary sources support code parameters and dated observations, not those missing conclusions.SourcesValidate.goapi.prlscan.com · BlocksREADMEbig.one · Depthapi.ourbit.com · Depth
Profitability Method
Daily net mining value = actually credited PRL × executable sale price minus electricity (wall watts × operating hours / 1,000 × tariff), hosting, pool/miner fees not already deducted, and depreciation. Shared inference operators should allocate incremental mining overhead separately from their existing serving costs. No tariff or device wall-power measurement was collected, so no net-dollar profit or payback claim is made.Sourcesapi.prlscan.com · Poolspool.kryptex.com · Prlbig.one · Depthapi.ourbit.com · DepthREADMEObserved 2026-10-14 21:59 UTC
Service Availability
The official site advertises serverless OpenAI-compatible inference at inference.pearlresearch.ai · V1, managed dedicated inference and on-premise deployment. The help overview still describes the inference platform as closed beta with a waitlist. These primary sources differ on presentation of availability; no authenticated API access, general availability, model performance or paying-customer counts were verified.Sourcespearlresearch.aihelp.pearlresearch.ai · What Is Pearl
Float Roadmap Boundary
The September paper specifies FP8 quantization after low-rank noise, output-tile tickets and bit-exact reconstruction of GPU arithmetic, with an MoE extension. Pinned mainnet rules separately show MoE activation at 71,935 and dense-only rejection at 91,630, rank penalty at 96,251 and salted seed at 99,000. Do not merge an FP8 feature-branch specification into claims about live dense INT consensus; current FP8 activation was not verified.Sourcespearlresearch.ai · Pearl WhitepaperParams.goIssue #424
Post Quantum Spend Path
The XMSS module documents OP_CHECKXMSSSIG as a post-quantum Taproot spend path based on SHAKE256 assumptions. It is stateful: at most 32 signatures per key, with unique message identifiers; identifier reuse can enable forgery. Signatures are documented as 2340 bytes. Builds without the xmss tag use error-returning stubs. This does not certify every address, deployment or consensus component as quantum-safe.SourceREADME
Supply Reconciliation
Maximum nominal supply is 2,100,000,000 PRL. Let h be height and C=floor(4×365×86400/194)=650,226. The ideal cumulative schedule is S(h)=2,100,000,000×h/(h+C). The implemented subsidy is floor(2,100,000,000×10^8×C/((h+C)(h−1+C))) grains, with zero genesis subsidy and 10^8 grains per PRL. No discrete halvings; h=C reaches half the ideal cap. At height 127,874, ideal gross cumulative issuance is 345,116,823.03045881 PRL; summing the integer subsidy gives 345,116,823.02982092 PRL (16.4341% of the cap). This is a schedule-derived maximum subsidy issuance under the pinned rules, not a scanned coinbase total, UTXO balance or circulating-supply measurement. It assumes the schedule applied unchanged to earlier blocks; fees are transfers, not new issuance. Consequently total_supply and circulating_supply remain null.SourcesValidate.goParams.goAmount.goapi.prlscan.com · BlocksObserved 2026-10-14 21:59 UTC
Developer And Community
GitHub API: 360 stars, 88 forks, 7 watchers. Discord invite API: approximately 7,421 members and 1,503 online. These are dated platform counts, not active developers, unique humans, token holders or decentralization measurements. Contributor count, recent-commit count and X follower count were not verified.Sourcesapi.github.com · Pearldiscord.com · JoinpearlObserved 2026-10-14 21:45 UTC
Transaction Output Rules
The help overview describes Taproot-only addresses, but CheckTransactionSanity at the pinned revision accepts Taproot (P2TR), Pay-to-Merkle-Root (P2MR, BIP 360) and OP_RETURN outputs and rejects other script classes. Preserve the older sourced overview as such; do not treat it as an exhaustive statement of this newer implementation or infer deployed activation from repository contents alone.Sourceshelp.pearlresearch.ai · What Is PearlValidate.go
Useful Work And Verification
In the INT design, miners commit input matrices, derive block-bound low-rank noise and perform noisy matrix multiplication. Hashing intermediate products extracts work tickets; a winning claim includes Merkle authentication and verification of selected work. A recursive hash-based Plonky2 certificate compresses that verification and hides private matrices. The header binds the claim while certificate bytes are excluded from block identity because randomized proof bytes would otherwise make identities malleable. None of this validates the semantic truth of an LLM answer.Sourcespearlresearch.ai · Int WhitepaperREADME
Market Identity And Derivatives
Ourbit separately announced native Pearl PRL/USDT futures for October 13, 2026 at 07:25 UTC, with up to 20× leverage. This verifies an announcement, not current funding, open interest or tradable futures depth. Those fields remain null. Other PRL markets, including Perle-related pairs, are excluded unless linked to Pearl Research Labs/native Pearl.Sourcesourbit.com · 17827791513820ourbit.com · 17827791512989Observed 2026-10-14 21:59 UTC