Pearl Network mainnet
Every figure below is computed from Keshi's own index. Windowed metrics state their window; partial windows say so; market data sits in its own frame and never mixes with chain data.
Header timestamps are miner-set with a ±5 minute consensus tolerance. This is block age, not network latency.
The window is part of the number. Published work rates for the same chain state diverge by up to 26% purely from estimator window, so Keshi always shows its own window and never scrapes anyone else’s figure.
200 window
EH/s: matmul attempts per second, a difficulty-derived consensus measure, not a measure of AI computation performed.
200 window
Share of the declared multiplication the ZK proof actually covers, MAC-weighted over the window (Σproven / Σdeclared). Per block the ratio is purely geometric: tileSize / (m·n).
WTEMA, adjusts every block
Smooth decay, no halvings. Pearl coinbases pay subsidy plus fees in one output.
Canonical non-coinbase transactions.
Mempools are per-node. This is our node’s mempool; other explorers legitimately disagree, and they are not wrong.
Target 194 s
Computed from miner-set header timestamps over canonical blocks; never extrapolate height to time from it.
Keshi’s headline supply number is observed issuance: the sum of coinbase outputs over canonical blocks. On Pearl a coinbase output pays subsidy plus that block’s fees, so observed issuance runs slightly above the theoretical curve.
Instantaneous per-block rate over the 200 window.
matmul attempts per second, a difficulty-derived consensus measure, not a measure of AI computation performed.
WTEMA with a 7-day half-life adjusts every block, so this is a smooth series rather than a step function.
Header-to-header spacing against Pearl's 194 s target.
Header timestamps are miner-set with a ±5 minute tolerance, so single-block spacing is noisy by construction; the rolling mean is the readable signal.
The window is part of the number. Published work rates for the same chain state diverge by up to 26% purely from estimator window, so Keshi always shows its own window and never scrapes anyone else’s figure.
Canonical non-coinbase transactions per UTC day.
Miner-claimed fees: a lower bound.
Coinbase outputs per UTC day.
Our node's mempool, sampled by this page since it loaded.
Mempools are per-node. This is our node’s mempool; other explorers legitimately disagree, and they are not wrong.
Polling; the live stream is not connected.
From the 3 newest blocks in our index.
Attributed blocks only; unattributed stays visible.