Common misconceptions

Hyperliquid moves faster than its documentation, so some official pages are out of date. Each heading below is a claim you may meet; the paragraph under it says what is true and where it is explained.

“Hyperliquid is an L2 or a rollup”

True: it is a layer one blockchain, ordered and finalised by its own consensus, HyperBFT, run by validators staking HYPE. The Arbitrum USDC bridge is, per the docs, legacy and deprecated: validators handle its withdrawals, it holds under 10% of HyperCore USDC, and USDC is minted natively through Circle’s CCTP (why its own chain).

“The HyperEVM is a separate chain”

True: the docs say plainly it is not; HyperCore and the HyperEVM share HyperBFT and one state. Wallets list the HyperEVM as its own network (chain id 999) and assets cross through a system address, which looks like a bridge. Some exchanges accept HYPE from HyperCore but not from the HyperEVM (two environments).

“One HyperEVM transaction can trade on HyperCore atomically”

True: CoreWriter only emits a log. HyperCore acts after the EVM block is built, when precompile reads are already stale, and cannot undo the EVM transaction (one transaction).

“A CoreWriter action that did nothing would have reverted”

True: CoreWriter never reverts. A refused action leaves status 1, gas spent, a log emitted and funds unmoved; observed refusals include “Invalid number of decimals” and “Insufficient USDC or HYPE balance for token transfer gas”. Only HyperCore state afterwards tells you (footguns).

“Orders sent from the HyperEVM wait several seconds”

True: the docs list CoreWriter orders and vault transfers as delayed a few seconds. On mainnet 52 orders landed 0.35 to 0.51 s after their EVM block; only vault transfers waited, 4.05 to 4.10 s (timing).

“The HyperEVM cannot write to HyperCore on mainnet yet”

True: the docs’ HyperEVM overview still says write system contracts are not live on mainnet, but one measurement found 5,455 CoreWriter actions in 73,914 mainnet blocks, including filled orders (chain vs docs).

“Outcome markets are free to trade”

True: only the HIP-4 page says so; the fees page and measurement disagree. Across 7,256 mainnet fills every buy was billed 0 and every sell charged: outcome trades pay only when closing or settling, at spot rates scaled by the deployer’s fee scale, with no maker rebates (fees).

“HIP-4 is testnet only”

True: one HIP-4 deployer-actions docs page said so, but outcome markets run on mainnet: first a protocol-run daily BTC market, and on 2026-09-14, 18 validator-approved templates for deployers (HIP-4).

“HIP-3 means anyone can list a perp at no cost”

True: a perp dex needs 500,000 HYPE staked for at least 183 days, slashable by validator vote (up to 100% for the worst cases) and burned if slashed. One dex per deployer; its first 3 assets skip the Dutch auction later ones go through. HIP-4 outcome deployers pay no deploy gas but need the same 500,000 HYPE (deployer economics).

“All volume dashboards measure the same thing”

True: each match writes two fill rows, and trackers sum both, halve the sum, count each trade once or count only fee-paying fills. On one outcome archive, halving gave about $29.27M and counting trades once about $46M (measuring Hyperliquid).

“Open interest from the API is in dollars”

True: openInterest is in base units (coins or contracts); multiply by the mark price (open interest).

“There is a HIP-5”

True: the official HIP list ends at HIP-4; social media mentions of a HIP-5 are speculation (what a HIP is).

For every measured contradiction, see Where the chain and the docs disagree. For the names these claims get wrong, see Naming and style guide.