Pearl PRL
Proof-of-Useful-Work L1 — the PoW is a real matrix multiplication, so the same GPU pass runs paid AI inference and secures the chain. Bitcoin-lineage (btcd fork, Taproot-only) with a 2.1B hard cap and smooth polynomial emission instead of halvings.
About Pearl
sources: official materials · verified 2026-10-13Pearl is a Bitcoin-style Layer 1 whose proof of work is a real matrix multiplication — the operation at the heart of every AI model — run on noised inputs and verified with a Plonky2 zkSNARK. The GPU cycles that secure the chain are the same cycles that run inference: Bitcoin turned electricity into security, Pearl turns it into security and useful compute at once. The codebase is a btcd/btcwallet/neutrino fork (UTXO, longest-chain, Taproot-only scripting with OP_CAT enabled, P2MR post-quantum addresses, optional XMSS signatures) plus the PoUW stack: an INT8 MatMul kernel with 1+o(1) overhead (arXiv 2504.09971, co-authored by CEO Omri Weinstein), a ≤65KB block certificate that replaces the nonce, and a vLLM-plugin GPU miner. The protocol is backed by a formal economics paper (Rafael Pass, arXiv 2606.06700) showing majority-attack cost stays tied to the block reward in equilibrium. Pearl Research Labs also runs a commercial inference platform, and its Together AI endpoint (Gemma-4-31B-it-Pearl) sells inference below standard price funded by the PRL mined on the same forward pass. Launched 2026-04-27 with zero genesis allocation; ~16.5% of the 2.1B cap mined by block 128k.
How it works
- Network
- Pearl — own genesis (block 0, 2026-04-27 09:00 UTC, 0 PRL), btcd-lineage Go monorepo (pearld node, Oyster wallet, SPV client), ISC copyfree license
- Consensus
- PoUW Nakamoto consensus — one large INT8 matmul per block on low-rank-noised inputs; BLAKE3 commitments + Plonky2 zkSNARK certificate (≤65KB) in a pouw_meta header field; WTEMA continuous difficulty retarget (168h half-life)
- Mining algorithm
- PearlHash (PoUW matmul + BLAKE3 + zkSNARK) — GPU
- Launched
- 2026-04-27
- Hardware
- GPU
Insights / Thesis
pow.family researchQuick start
tested 2026-10-15~30 min · ~$0.35 to test on a rentalMine PRL (+NOCK merge) on a rented GPU via LuckyPool
miner: SRBMiner (algo pearlhash)- 1
# 1. Install SRBMiner and point it at the nearest LuckyPool stratum:
↳ 18 regions, e.g. pearl-eu2.luckypool.io:3360 (start diff 2M, <500 TH/s) · :3361 (4M) · :3362 (8M, 1000 TH/s+). Restrictive ISP? swap luckypool.io → lproute.com. Source: pearl.luckypool.io/how-to-mine, 2026-10-15.
- 2
# 2. Wallet address format — merge NOCK for free by appending the NOCK address:
↳ PRL_ADDRESS+NOCK_ADDRESS.WORKER — use a NATIVE Nockchain address (not a 0x wrapped-on-Base one); the pool pays NOCK on Nockchain itself. Pearl rewards unchanged. K1Pool instead auto-pays MDL on PRL hash (PPLNS/SOLO).
- 3
./SRBMiner --algorithm pearlhash --pool pearl-eu2.luckypool.io:3360 --wallet PRL_ADDRESS+NOCK_ADDRESS.WORKER --password x
↳ HiveOS: import the LP_PEARL_SRBMINER flight-sheet JSON from the pool page (coin PEARL, srbminer_custom, template %WAL%). Alternatives: PeakMiner, BzMiner. Solo: prefix solo: before the address.
- 4
# 3. Payouts and stats
↳ Min payout 1 PRL (Kryptex) / pool-dependent (LuckyPool). Stats: pearl.luckypool.io (PRL) + nock.luckypool.io (NOCK side). Oyster wallet (pearl wallet v3.0.0, 2026-10-13) for native PRL.
Pitfalls we hit
- Fast blocks: chain averages ~116s vs the 194s target (difficulty WTEMA chasing hashrate hypergrowth) — emission runs ~1.7× ahead of the nominal calendar; '4 years to 50%' lands earlier on the clock.
- Supply numbers can disagree by ~0.58%: prlscan's address ledger sums ~347.7M while the issuance formula (and CoinGecko/CMC) say ~345.7M at block 128k — the ledger over-counts (indexer artifact, audited 2026-10-15: genesis 0, subsidy formula-exact, fees negligible). This site displays the audited issuance figure.
- PoUW 'useful' revenue is real but small: the Together AI endpoint discount is '25%+ below standard price' on one model — the inference-side subsidy is early. Pool mining income remains the dominant economics today.
Tokenomics — token & supply
- Ticker
- PRL
- Maximum supply
- 2,100,000,000 PRL hard cap (100× Bitcoin) — verified in code (emission_test.go, 2026-10-15)
- Emission
- Smooth polynomial decay R(t) = H/(t+H) with H = 650,226 blocks — no halvings, issuance tapers continuously. Test vectors: 3,229.6 PRL/block at h=1 → 807.4 at 50% (h=650k) → 26.7 at 90.9% (h=6.5M). ~16.46% mined at h≈128k (2026-10-15). Fees mint to miners (tiny: thousands of PRL chain-wide).
- Reward now
- ~2,254 PRL per block (h≈128k, 2026-10-15) ≈ 1.13M PRL/day
- Halving
- None — continuous polynomial decay replaces halvings by design
- Premine
- 0%
PRL network
mining share
24h pool sharesupply
issued = 2.1B × h/(h+650,226) · audited vs codePools
pool compare tool → P1| pool | fee | scheme | min payout | merge | stratum |
|---|---|---|---|---|---|
| Kryptex | 1% | PPS+ 3% · SOLO 1% · min payout 1 PRL · ~54% of network blocks/24h (2026-10-15) | 1 PRL | — | see pool page — PPS+ 3% / SOLO 1% |
| LuckyPool (merge PRL+NOCK) live tracked | 1% | PPLNS · NOCK merge free · solo: prefix · 18 regions + lproute.com mirrors | — | auxpow | pearl-eu2.luckypool.io:3360 (start diff 2M) |
| K1Pool (merge PRL+MDL) | 0% | PPLNS + SOLO · auto-pays MDL as merged second coin · 0% fee promo (2026-10-15) | — | auxpow | see pool page |
| F2Pool | — | PRL pool since 2026-07-06 · RTX 50/40/30 series per official guide | — | — | see pool page |
Hardware & miners
full hardware db (971 devices) → P1Exchanges & liquidity
Reference material
cited on this page · links verified 2026-10-13Who to watch
sensors, not shills — verify before actingRisks
Pool concentration: Kryptex found ~54% of blocks/24h (275/505, prlscan 2026-10-15) — one operator outage or censoring policy moves majority hashrate.
Novel security model: PoUW matmul consensus is peer-reviewed on paper (arXiv 2504.09971 / 2606.06700) but only ~6 months in production; the FP8 expansion (PR #424, 2026-10-13) is still fixing miner/verifier boundary mismatches — consensus-critical code is actively changing.
Backers' amounts undisclosed: RootData links backers to Multicoin Capital co-founders and CoinFund's CEO; how much PRL they hold (OTC ledger labels exist on prlscan) is unknown. Team is public at CEO level (Omri Weinstein) but thin beyond.
Fast blocks: chain averages ~116s vs the 194s target (difficulty WTEMA chasing hashrate hypergrowth) — emission runs ~1.7× ahead of the nominal calendar; '4 years to 50%' lands earlier on the clock.
Supply numbers can disagree by ~0.58%: prlscan's address ledger sums ~347.7M while the issuance formula (and CoinGecko/CMC) say ~345.7M at block 128k — the ledger over-counts (indexer artifact, audited 2026-10-15: genesis 0, subsidy formula-exact, fees negligible). This site displays the audited issuance figure.
PoUW 'useful' revenue is real but small: the Together AI endpoint discount is '25%+ below standard price' on one model — the inference-side subsidy is early. Pool mining income remains the dominant economics today.
The usefulness gap is measured, not alleged: an independent ~$50 cloud-GPU study of 8,012 pool workers (arXiv:2606.04819) found 100% of miners generating random matrices and zero AI inference — the protocol verifies multiplication correctness, not usefulness, and the team has not publicly answered.
Upbit and Bithumb deferred listing over circulating-supply concerns; with ~175M circulating against a ~2.1B cap (~50x FDV dilution), supply transparency is the open question tier-1 venues are waiting on.
NoisyGEMM ships INT8 only — FP8/BF16 support (where most real AI workloads live) is not yet on the network; until it is, the 'useful work' narrative runs ahead of what the chain can actually verify.
tokenomics
- emission
- Smooth polynomial decay R(t) = H/(t+H) with H = 650,226 blocks — no halvings, issuance tapers continuously. Test vectors: 3,229.6 PRL/block at h=1 → 807.4 at 50% (h=650k) → 26.7 at 90.9% (h=6.5M). ~16.46% mined at h≈128k (2026-10-15). Fees mint to miners (tiny: thousands of PRL chain-wide).
- reward now
- ~2,254 PRL per block (h≈128k, 2026-10-15) ≈ 1.13M PRL/day
- halving
- None — continuous polynomial decay replaces halvings by design
Don't confuse with
PRL — Pearl Research (Base chain token): A separate 'Pearl Research' token on Base with 100B supply and no max supply appears on some aggregators/Coinbase listings — not the Pearl PoUW L1 (pearlresearch.ai). This coin: 2.1B hard cap, PoW matmul.
catalysts
Not financial advice. Verify before deploying capital.