OPEN METRICS · AUTO-GENERATED · NO HAND ON THE SCALE

We publish our own measurements.

BlazePhoenix is, to our knowledge, the only DEX aggregator publishing its own measured costs and its own availability record — a falsifiable claim: find a counterexample and the sentence changes. Every number below is generated by CI from public source files (the daily dataset, the facts register, the BPX register) — no manual step exists between the measurement and this page.

Availability, per chain — the honest series

Share of daily snapshots where the chain's RPC answered our reads, over the 10-day record. rpc_ok=false days are OUR infrastructure failing, and they are counted — an availability page that only counts good days is an advertisement.

Arbitrum
70%
7/10 days
Base
30%
3/10 days
Ethereum
0%
0/10 days
Optimism
90%
9/10 days
Robinhood
20%
2/10 days

Staking solvency: 10/10 days isSolvent() = true — re-read it live on /solvency.

Measured costs — transcluded, not restated

A pool's whole working state packs into one 256-bit word; scoring a candidate costs a single SLOAD (one storage read instead of several).

Reproduce: Verified Hub source; gas traces of warm vs cold swaps.

https://blazephoenix.xyz/learn/the-monoslot

BlazePhoenix discovers Uniswap V4 pools deterministically on-chain with no factory list and no admin allowlist. V4 is a singleton: every pool of every deployer is keccak256(PoolKey) in the one PoolManager. Hookless pools for a pair are derived over the canonical fee grid (hooks=0) and probed in one gas-free extsload (deployer-blind, 99 calls -> 1). Hooked pools carry an arbitrary 160-bit hook address (only 14 low bits constrained) and are unenumerable on-chain by construction.

Reproduce: Derive a real Base pool id and read its slot0: PID=$(cast keccak $(cast abi-encode 'f(address,address,uint24,int24,address)' 0x4200000000000000000000000000000000000006 0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913 500 10 0x0000000000000000000000000000000000000000)); cast call 0xa3c0c9b65bad0b08107aa264b0f3db444b867a71 'getSlot0(bytes32)(uint160,int24,uint24,uint24)' $PID --rpc-url https://mainnet.base.org -- a non-zero sqrtPriceX96 is the pool, derived not looked up.

https://blazephoenix.xyz/learn/deterministic-v4-discovery

Per leg, the BlazePhoenix executor is 2.4× to 10.9× more gas-efficient than the sector's leading commercial aggregation router, measured on the same fork at the same block by replaying that router's own production calldata against its deployed contract (reference compiled at optimizer_runs = 999999). On the clearest pair their single-leg route cost 879,553 gas; four BlazePhoenix legs cost 322,172.

Reproduce: Method self-validates: the reference router's measured delivery matches its API promise to the wei, quotes were pinned to their build block, and API gas estimates were discarded after one measured 148% low. Full method and numbers: whitepaper §16 (The executor, measured against the field).

https://blazephoenix.xyz/whitepaper.md

Source of truth: /facts.json — the proof-carrying register this section transcludes verbatim.

The security ledger

No-exit rate (trending, probed)
14% of 93
Numbered advisories
5
Fixed & regression-tested
3
Credited researchers
9

The register: /security/advisories · the people: /security/researchers

We lint our own prose like our own gas

Every deploy runs a structural lint (27 rules from published GEO research) over all 148 technical articles; the current scoreboard is public at /geo-lint.json — including what fails. A scoreboard that hides the red is not a scoreboard.

Corpus, counted

148 articles · 57 terms · 18 coinedTerms · 39 standardTerms · 34 languageHubs · 5 liveChains

Generated 2026-08-27 · regenerated on every deploy · history in the daily dataset.