The rule and the schema are real. The data isn't — yet.

The winner-selection rule this page runs (sort bids by price per CU descending, greedily fill compute-unit capacity, skip a bid that no longer fits rather than partial-filling it, tie-break by earliest submission timestamp) and every field below (slot, auction_index, slot_cus, winners, clearing_price_per_cu, capacity_used) are taken directly from Raiku's own public documentation, not invented for this demo. What isn't real: Raiku's own docs confirm the auction ledger lives entirely inside their private off-chain Engine ("Auction logic location: Marketplace/Engine — not in validator runtime logic"), and raiku-sdk access is currently pilot-gated — so every bid, price, and timestamp below is generated by a seeded random generator standing in for that feed until access opens up.

Auction Radar

for Raiku · AOT / JIT blockspace
replaying synthetic feed slot — seed 42
Auctions observed 0
Avg clearing price —µlam/CU
Avg capacity used —%
Real pruning window 20–33min

Settled auctions — live replay

AOT JIT
Slot#TypeSlot CUs WinnersClearing priceCapacitySettled

Clearing price over time

AOT JIT

Capacity utilization

per slot, trailing window

Backtester

Would this bid have won the most recently settled auction of this type? Runs Raiku's documented winner-selection rule against it directly.

Actual clearing price
—
Actual capacity used
—
Min price to have won
—
Your requested CUs
—

Winner-selection rule

Sort bids by price per CU, descending. Admit bids greedily until the slot's compute-unit capacity is filled — a bid that no longer fits is skipped, never partially filled. Ties break by earliest submission timestamp. This is the exact rule documented for Raiku's marketplace, reimplemented here and in the project's tested Rust backtester crate.

What's real: this selection algorithm, the schema, and the backtest math. What's synthetic: every bid, price, and timestamp on this page.