Trades settle onchain, but what you're handed is a report — an interface telling you what it says happened. Assay ignores the report and rebuilds the fill from the chain: what left the wallet, what entered it, what the venue kept.
To determine what a thing is actually made of, rather than what it is claimed to be. Assay offices have declined to take sellers at their word since the 1300s: you bring the silver, they test it themselves, and only then is it struck.
Sterling is 925 fine — 925 parts silver per thousand. A fill has a fineness too. You were promised a thousand parts; the chain says how many arrived.
Fineness needs a claim. Reconstruction alone tells you what arrived; it cannot tell you what you were owed. That number lives in the interface that filled you, so you supply it — until venue decoders can read it off the transaction itself.
Every assay you've struck, kept in this browser and nowhere else. Assay has no account system and no server to send it to — clear your site data and this is gone.
An assay office keeps a register of what it has marked. This is yours, and it never leaves your machine.
Four steps, no interpretation. Anyone re-running them lands on the same number or the method is broken.
Fetch the transaction and its receipt. Confirm it settled — an unconfirmed transaction is a rumour, not a fill.
Read every ERC-20 transfer the token contracts themselves emitted, ignoring whatever the venue said about its own work.
Net them against the wallet. That sum is what moved, and no interface gets a vote in it.
Divide by what was promised. Parts per thousand. That is the fineness, and assay offices round down — you never get credit for purity you don't have.
Millesimal fineness is the real standard for silver. These are the actual thresholds, applied to fills.
The honest caveat: step ii trusts each token contract to report its own transfers, which is weaker than diffing raw balances at block N−1 and N. Balance-diffing needs an archive node and the public RPC won't serve it. The method still catches what matters — a venue claiming one output while the token moved another — because the venue and the token are different contracts. A malicious token would defeat it. Written here rather than discovered by you later.
What exists, what doesn't, and nothing in between. Check any Working row by using it.
A row reads Working because the code ships and you can run it yourself. The rest are blank because they are blank. Every project that ends badly had a table like this filled in with numbers months before the numbers were true.
Every settled transaction that moved a token in or out of this wallet, over a recent window. Click any to assay it in full.
Pulled from ERC-20 transfer logs over a fixed recent window, not a full history. A wallet showing nothing here has had no token movement in the window — which is not the same as no history at all.
This is what your dashboard looks like once a Telegram chat is registered. Every number above belongs to that chat and to nobody else.
Assay holds three things about you: the Telegram chat id, the addresses you asked it to watch, and this history. Your display name is whatever Telegram already calls you — it is not asked for. No email, no password, no wallet key; there is nothing here you had to trust us with.
Each line is one thing that happened and one notice that went out. Strikes carry a permalink, and a permalink re-runs the assay from the chain in front of whoever opens it — which is what makes a record worth keeping rather than a claim you have to take from us.
Every settled transaction that moves a token in or out of these gets rebuilt by the same four steps and explained on Telegram. New transactions only — adding a wallet does not scan its history.
Chain data is public, so the agent will watch any address you give it, not only ones you control. The code you registered with proves you own the Telegram chat — never the wallet.
Today you type the promised number yourself. A decoder would read the quote out of the router calldata, so the claim comes from the transaction rather than from your memory of the screen.
quote read from calldata — you did not have to type it promised 98.5 ARB (decoded, venue quote) arrived 96.412031 ARB fineness 978 / 1000 (Sterling, rounded down)
The method reads Transfer logs, so a token that moves value without emitting one is invisible to it. Diffing balances at block N−1 against block N would see the movement regardless.
No ERC-20 transfers in this transaction. ── state diff, block 496,204,453 → 496,204,454 ── ARB +96.412031 (balance moved, no log emitted) USDC −10000 fineness 978 / 1000 (Sterling, rounded down)
One assay is a number. A hundred is a pattern. Drift would track fineness per venue and per token over time, so a venue that quietly gets worse has somewhere to show up.
ARB · your last 6 strikes 999 ▏████████████████████ 4 weeks ago 981 ▏███████████████████ 3 weeks ago 978 ▏███████████████████ 2 weeks ago 934 ▏██████████████████ 9 days ago 921 ▏█████████████████ 4 days ago 906 ▏█████████████████ today median down 73 parts over four weeks
The reconstruction is not specific to Arbitrum — it reads ERC-20 Transfer logs, which every EVM chain emits. What is specific is the router table, and that has to be earned per chain the same way this one was: empirically, and re-checked at runtime.
chain base · 8453 router 0x2626…c43 exactInputSingle arrived 1.4802 WETH fineness 993 / 1000 (Fine, rounded down) router table for this chain: 4 selectors verified, 2 withheld
There isn't one, and there is no date for one. The whitepaper's position stands: if tokenomics ever exist they will be published as real, consistent, on-chain-verifiable numbers — or not at all.
No token exists. Nothing here is an offer. If that ever changes, this panel will carry the contract address and the supply, both readable from the chain rather than from us.
Because the alternative is a page that quietly implies it does. The Standing page lists these as Not built and always has; this is the same list with the output attached, so you can judge whether it would be worth having rather than take the label on faith.
Every panel above is a mock-up written by hand. None of it came from the chain, none of it is anyone's real fill, and no part of it runs.
A fill is a claim until you strike it
Strike