docs: V-02 said "none" and that was wrong - the whole xStock suite is here

Correcting a P0 conclusion I reported as fact.

The claim was "there is no canonical xStock ERC-20 on HyperEVM". The truth is
that all ten of the largest xStocks are deployed on this chain at Backed's
canonical addresses, holding real supply - CRCLx 312,842 units, STRCx 108,351,
NVDAx 104,729, MSTRx 66,795, GOOGLx 59,252, TSLAx 46,467. The three that can
be priced from measured feeds come to ~$63M; the suite is well over $100M.

They are Backed's own deployments and that is asserted, not assumed. Backed
deploys deterministically, so the address is identical across chains, and the
runtime bytecode hash is byte-identical on HyperEVM, Optimism and BNB. On
every asset checked HyperEVM holds SEVERAL TIMES the supply of either - it is
the largest EVM deployment of xStocks, not a fringe one.

How the wrong answer happened, recorded in the row rather than overwritten:
two searches, both empty, both empties read as absence of the asset. HyperSwap
positions came back with no xStocks because xStocks have no pools - a token
can be widely held and never pooled. HyperCore tokenInfo came back with every
equity unlinked because xStocks are EVM-native and were never HyperCore
tokens, so there is no HIP-1 record to link. Both searches answered correctly.
The error was reading a derived index as a census.

The tell was there and was explained away: SPYx WAS found, through a pool
belonging to its wrapper, and was filed as an anomaly rather than the edge of
a suite. One contradicting data point against two confident absences, and the
data point lost. The check never run was eth_getCode on a published address -
one RPC call, which would have settled it on day one.

Recorded as Shape 7 in REVIEW-NOTES: an empty index read as an empty world.
Same shape as Shape 1 and Shape 5 - a reader with no writer, an argument
nobody checked - turned on the verification process itself.

What does not change: they rebase, six of the ten have already moved off a 1.0
multiplier, and Backed's own docs say every EVM deployment does. Changing
chain does not help; only Solana and TON keep raw balances constant.

What sharpens: RebaseDetector is not a safety net catching one bad token. It
refuses the entire intended quote-asset class and is the reason the allowlist
is empty. Correct for balance-based accounting, and it should read as
uncomfortable rather than reassuring. The doc comment now says so, and records
the fix it is holding the door for - booking collateral in shares, the
rebase-invariant unit these tokens already expose.

V-22 is new: no xStock has a HyperSwap pool at any tier, against any pair.
Holders exist; an EVM-side market between them does not. Invisible until
launch day and expensive to discover then.

277 tests still green.

Author: Sent-lab

contracts/src/lib/RebaseDetector.sol

@@ -26,8 +26,42 @@ pragma solidity 0.8.28;
  * "Backed Token Implementation", so this is the real issuer's design and not one
  * venue's wrapper.
  *
+ * THIS IS THE WHOLE ASSET CLASS, NOT ONE BAD TOKEN
+ * ------------------------------------------------
+ * All ten of the largest xStocks are deployed on this chain running byte-identical
+ * code, and six have already moved off a 1.0 multiplier:
+ *
+ *     STRCx 1.0808929977   SPYx  1.0057145603   QQQx   1.0027250297
+ *     GOOGLx 1.0023772501  NVDAx 1.0009180758   MUx    1.0004015986
+ *
+ * So this library is not a safety net catching one unsuitable asset. **It refuses
+ * the entire intended quote-asset class, and it is why the allowlist is empty.**
+ * That is correct for the accounting as it stands, and it should be uncomfortable
+ * rather than reassuring.
+ *
+ * Changing chain does not help. Backed's developer documentation states that on
+ * every EVM chain their tokens "implement the ERC-20 standard with rebasing
+ * logic"; only Solana and TON keep raw balances constant, and neither runs this
+ * codebase.
+ *
+ * THE FIX THIS IS HOLDING THE DOOR FOR
+ * ------------------------------------
+ * These tokens expose `sharesOf(address)`, and shares are the rebase-invariant
+ * unit: `balanceOf = sharesOf × multiplier`. A market that booked collateral in
+ * SHARES would be neutral to the multiplier in both directions — claims scale
+ * together, dividends accrue to whoever holds the claim, and a reverse split
+ * cannot make it insolvent.
+ *
+ * That is an accounting change rather than a workaround, and it is the shape any
+ * real answer takes. Until it exists, refusing is the honest behaviour: the
+ * alternative is a market that looks fine for months and then cannot pay.
+ *
  * WHAT WOULD HAPPEN WITHOUT THIS CHECK
  * ------------------------------------
+ * Every multiplier measured so far has drifted UPWARD, because dividends accrue
+ * into it, and a rising multiplier leaves the market holding more than its books
+ * owe. Surplus, not insolvency. The danger is the other direction.
+ *
  * A reverse split lowers the multiplier. Every holder's balance shrinks, and so
  * does the market's — while `curveCollateral` does not move, because no transfer
  * occurred and no event fired. The market becomes **insolvent against its own

docs/REVIEW-NOTES.md

@@ -356,6 +356,48 @@ deterministically through the same handler the fuzzer drives.
 
 ---
 
+## Shape 7 — an empty index read as an empty world
+
+Every shape above is a defect in the code. This one is a defect in the *verification*, and it
+produced a wrong P0 conclusion that was written into the ledger and reported as fact.
+
+**The claim:** "there is no canonical xStock ERC-20 on HyperEVM."
+
+**The truth:** all ten of the largest xStocks are deployed there, at Backed's canonical
+addresses, running byte-identical code to Optimism and BNB — and HyperEVM holds *more* supply
+than either. Over $100M of assets.
+
+**How the wrong answer was reached.** Two searches, both empty, both empties read as absence of
+the thing itself:
+
+| Searched | Came back | What empty actually meant |
+|---|---|---|
+| HyperSwap V3 positions, sampled for their token pairs | no xStocks | xStocks have **no pools**. A token can be widely held and never pooled. |
+| HyperCore `tokenInfo.evmContract`, 400 HIP-1 indices | every equity `address(0)` | xStocks are **EVM-native**. They were never HyperCore tokens, so there is no HIP-1 record to link. Wrong table. |
+
+Neither search was wrong. Both were answered correctly. The error was reading a **derived
+index** as a **census**: a pool registry indexes what trades, a HIP-1 table indexes what was
+bridged from Core, and neither is a list of what exists.
+
+The tell was there and was explained away. `SPYx` *was* found — through a pool belonging to its
+wrapper — and was treated as an anomaly rather than as the edge of a suite. One contradicting
+data point against two confident absences, and the data point lost.
+
+**The check never run was the cheapest one available.** Take a published address, call
+`eth_getCode`. One RPC call. It would have answered the row on day one and refuted the
+conclusion in seconds.
+
+**Why this belongs in a document about code defects.** It is the same shape as Shape 1 and
+Shape 5 — a reader with no writer, an argument nobody checked — turned on the process that is
+supposed to catch them. An absence is evidence only if you know the index would have contained
+the thing. Both indices here were guaranteed *not* to contain it, for reasons that were
+knowable before the search.
+
+The correction is recorded inside V-02 rather than replacing the wrong text, because a ledger
+that silently repairs itself teaches nobody anything.
+
+---
+
 ## What is deliberately not implemented
 
 These are not placeholders. They are recorded refusals, and each names the