Two routes to every stock-to-stock trade. We quote both.
Robinhood Chain carries 203 tokenised equities. To move SPY into NVDA you can trade the pair directly, or change through USDG, the dollar stablecoin the chain settles in. Which one pays more depends on the pair and on the size, and it flips. On a large trade the wrong choice has cost as much as 66%. There is no rule to memorise here, which is why every route on this page is quoted against the pools before it is offered. Nothing is deployed yet.
Line thickness is turnover. A gold station has no dollar pool, so you reach it only across the equity network. Every pool drawn here was verified on chain: liquidity read off each v3 pool, and each v4 pool key recomputed against its own id. Pools with nothing behind them are excluded rather than drawn, because a dead pool still quotes a price. One place the map lies: on a real network, changing trains is free. Here it costs a second pool fee, which is why the direct line usually wins and why the exceptions are worth quoting for.
Reading the network from chain 4663…
Reading the map
SPY is the interchange. Six of the nine equity-to-equity pools meet there, so most journeys between two equities change at SPY. AAPL is the only station on both the green and the gold network — the single link between the equity cluster and the metals cluster.
SPCX and QQQ have no dollar pool in this registry, so on this map you reach them by changing onto the equity network. Treat that as the limit of the scan rather than a fact about the chain — we said the same about SLV and it was wrong. SLV has a USDG pool holding $76,024, which this scan had simply not reached yet.
Plan a journey
From one equity to another
Pick two stations. These are the ten stations the map draws — equities sitting on a verified equity-to-equity pool, plus the dollar trunk. The registry holds far more pools than this, but most are the 1% tier with nothing behind them, and routing you through one would be the thing this page spends two screens condemning.
Every route is quoted by asking the pools to fill it, so each pool's fee and the price impact at your size are inside the numbers, and the impact is measured against a reference quote a thousand times smaller. Gas, the protocol fee and transfer restrictions are not in these figures. A price difference on a screen is not a route. HOODFLY does not eliminate impermanent loss, oracle risk, market risk, compliance risk or MEV risk, and does not claim to.
The measurement, and what it actually measured
The same journey, two routes, two prices
SPY to NVDA has a direct line and a change at the dollar trunk. Both were quoted for the same $1,000 trade against real pool liquidity, so fees and price impact are already inside both numbers. These quotes are real. What we first concluded from them was not — the correction is directly below the table, because it is more useful than the table.
| Route | Changes | Ticks crossed | You receive |
|---|---|---|---|
| Direct — SPY line | none | 4 | 4.452553 NVDA |
| Change at USDG — SPY → USDG → NVDA | one | 0 and 1 | 4.469387 NVDA |
Then we re-measured against the venue that carries the market
The pool above holds 7.15 SPY and 6.26 NVDA — about $6,900, so a $1,000 trade is fourteen percent of the book. At that size you are measuring your own price impact, not a routing edge. The real SPY/NVDA market is the Uniswap v4 book, and we have now quoted it directly.
Against the live market the same comparison is a fraction of a percent, and at $1,000 it is not even reliably positive — we measured it three times in twenty minutes and got +0.083%, −0.029% and −0.013%. Both tables stay up: the first is what the thin pool said, the second is what the market says. The gap between them is the entire reason this router quotes execution rather than prices.
The same question, asked of the real market
Executable quotes against the v4 SPY/NVDA book and the dollar trunk, at six sizes. The right-hand column is the one that decides whether this product should exist: what is left after the 0.10% protocol fee.
| Trade size | Direct (v4) | Change at USDG | Edge | After our fee |
|---|---|---|---|---|
| $100 | 0.442120 | 0.442409 | +0.065% | −0.035% |
| $1,000 | 4.419620 | 4.423936 | +0.098% | −0.002% |
| $5,000 | — | — | +0.344% | +0.244% |
| $10,000 | 43.916460 | 44.224173 | +0.701% | +0.601% |
| $50,000 | 211.984566 | 220.786046 | +4.152% | +4.052% |
Ten pairs, four sizes, and no rule
The table above is one pair. Here is every equity pair on the chain with a real market, measured the same way: what the dollar route pays against going direct. Negative means direct wins.
Measured from executable quotes across every equity pair with a real market.
| Pair | $100 | $1,000 | $10,000 | $50,000 |
|---|---|---|---|---|
| SPY/NVDA | −0.05% | −0.03% | +0.57% | +3.82% |
| SPY/MSFT | −0.22% | −0.30% | −0.58% | −66.2% |
| MU/SPY | −0.18% | −0.24% | −0.76% | −2.50% |
| PLTR/SPY | −0.27% | −0.44% | −1.65% | −4.98% |
| AMD/SPY | −0.27% | −0.50% | −3.42% | −58.7% |
| SPY/TSLA | −0.26% | −0.30% | −0.36% | −13.0% |
| GOOGL/SPY | −0.18% | −0.50% | −1.43% | −8.61% |
| TSM/SPY | −1.68% | −1.94% | −4.31% | −59.7% |
| BB/SPY | −10.7% | −22.2% | −90.0% | −97.9% |
| GME/NVDA | +0.89% | +4.09% | — | — |
This is the opposite of what we expected, and we had published the opposite
We led with SPY/NVDA and said the dollar route wins. Across ten pairs it loses on eight. The reason is arithmetic we should have done first: the direct route is one 0.05% hop and the dollar route is two, so it starts 0.05% behind and only wins if the dollar books are deeper. For NVDA they are — a $2.97m dollar book against a $181k equity pool. For AMD, TSM, MSFT and MU it is the other way round, SPY is the deep hub and the dollar books are thin, which is how you get to −58% and −66% at $50,000.
We generalised from the one pair where our assumption happened to hold. SPY is a real hub on this chain, and the dollar trunk is deep for a couple of majors and thin for everything else.
A rule you could memorise would make this product pointless
Which route wins depends on the pair, depends on the size, and flips. At $1,000 on SPY/NVDA we measured +0.083%, then −0.029%, then −0.013% within twenty minutes — at that size the two routes are inside the noise and the sign is not stable. None of this is visible in a price, because a price cannot show you depth.
That is the whole case for quoting both every time, and the whole case against any tool that compares listed prices. Being wrong is cheap on a small trade and costs up to 60% on a large one.
The fee is free on a single hop, and there is no size floor
This said we do not charge below about $1,100. That was not true of the contract. There is no size floor anywhere in the code: what the router actually does is charge nothing on a single hop and 0.10% on a route with a change in it, whatever the size. A $200 trade through the dollar trunk pays the full 0.10%. The claim described a fee model we had discussed and not built, on the page whose whole argument is that a number should be bound to its source.
The underlying arithmetic still holds and is the reason the claim was tempting: on the pairs where routing wins at all, the edge only clears a 0.10% fee above roughly $1,100. Below that, a fee of 0.10% against an edge of 0.065% takes more than it hands over. Either the floor gets built or the fee gets cut. Until one of those happens this page will not describe a floor as though it exists.
Above it, the case gets strong quickly, and for a plain reason: the equity-to-equity pools are shallow and the dollar books are deep. SPY/NVDA holds around $181,000. The USDG/NVDA book holds about $3.4m. So the bigger your trade, the more the direct route costs you and the less the dollar route does — +0.60% net at $10,000, +4.05% at $50,000. This is a router for size, not for pocket money, and pricing it as anything else would be dishonest.
What does survive: listed prices lie
Comparing the two pools' listed prices says direct is 0.46% better. Actually asking the pools to fill the trade says it is 0.38% worse. The answer flips. That flip is real regardless of which pool it was measured on, and it is why nothing here ever reads a listed price to decide a route. Depth is invisible in a listed price, which is exactly how a $6,900 pool passes for a market.
One pair, one moment, and a crossover specific to a 0.10% fee. We have not measured the other eight pairs, and the number moves as liquidity moves. Gas is excluded — two hops cost roughly twice the gas, which is negligible at 0.04 gwei but is not zero.
Every figure we have revised
Kept on the page with the reason, because a number you cannot audit the history of is not a number you should trade on.
| What we said | Why it was wrong |
|---|---|
| “The dollar route beats going direct” | True of SPY/NVDA above ~$5,000. False on eight of the ten pairs we then measured, because the dollar route pays two pool fees against the direct route's one and only wins where the dollar book is deeper. We had generalised from a single pair. |
| “Routing SPY into NVDA was worth 0.38%” | Measured against a v3 pool holding $6,900, where $1,000 is 14% of the book — price impact, not a routing edge. Re-quoted against the v4 book that carries the market, the real figure is 0.098%. Overstated fourfold, and against a 0.10% fee it changes the answer from "worth it" to "not at this size." |
| “Prices differ by 7.7% across pools” | Sliced by pool, the dollar book agrees to 0.3%. The spread was entirely between different quote currencies. |
| “The gap can't clear 1.10% of fees” | An assumed fee tier. All three legs are 0.05%, so the real hurdle is 0.15% — we had inflated it sevenfold. |
| “There's a −5.70% gap” | Priced off a pool with zero liquidity. A dead pool still lists a price; it is a fossil, not a market. |
All three were caught only by making the tools read fee() and liquidity() off every pool instead of trusting their own configuration. An unverified assumption does not announce itself — it produces a confident number.
What can stop you
Two sets of powers apply to any trade here, and they are not the same size. The first belongs to Robinhood and applies to the tokens themselves — it applies whether or not HOODFLY exists. The second belongs to whoever runs this router. Both are listed in full.
Robinhood's powers over the tokens
Whole network may be suspended without notice
A single call halts transfers on all 203 stock tokens at once. It has been used — 30 June 2026, for about 69 seconds. While it is on, nothing here can run, and the planner stops offering journeys rather than letting your transaction fail and take your gas with it.
Any address can be blocked, including yours
The issuer can add any address to a blocklist, and the check covers the sender as well as both counterparties — so a blocked address cannot send, receive, or be routed through. 246 blocks against 177 addresses so far, and we classified every one: all 177 are ordinary wallets. No contract has ever been blocked.
A single key can delete your stock tokens out of your own wallet
adminBurn zeroes any holder's balance, with no pause check and no blocklist check and no warning. This power applies to your wallet whether or not you ever touch HOODFLY — not using this site does not protect you from it, and you should know it exists before you hold these tokens at all. What it means for us is narrower: it is why HOODFLY holds nothing between transactions. There is no balance sitting here to destroy.
The code behind all 203 tokens can be swapped in one transaction, with no delay
One beacon contract holds the implementation pointer for every stock token on the chain. Changing it changes how all 203 behave at once, immediately — it has already been changed twice. Thirteen roles govern this system and every one is a single ordinary key: no multisig, no timelock anywhere. Every guarantee on this page describes how the tokens behave today.
This operator's powers over the router
What this operator can and cannot do
Can: add or remove pools, change the fee up to a ceiling fixed in the bytecode that nobody can raise, pause routing, sweep stray dust.
Cannot: touch your tokens, mint anything, block a withdrawal — nothing is deposited — exceed the fee ceiling, or upgrade the contract. There is no upgrade path.
On venues: the router executes on Uniswap v3 and v4, and on routes that mix them — which matters, because the route that wins here is often a v4 leg followed by a v3 one. A v4 pool is allowlisted by its full key and the contract recomputes the pool id from it, so an entry cannot name one pool and behave as another.
Open items, stated up front: the fee change has no timelock, and there is no external audit. 49 tests pass, 11 of them against live chain state — that is not an audit and we do not call it one.
Tokens we will not route
Contracts exist whose name is a byte-for-byte copy of the official format and which Robinhood never issued. One has more holders than several real ones. Real tokens can also be renamed at will, and one already has been. So nothing is accepted on its name — only on whether Robinhood's factory issued that exact address.
| Claims to be | Address | Holders | Why refused |
|---|---|---|---|
| HOODRAT | 0x807A4ECD29a9a7A3Bb27d589e2ef9BA831387C83 | 5,193 | not in the factory log |
| USAR | 0x38D6254bcEa24cdcD327d99a0D4e50B8403B2Ea1 | 4,877 | not in the factory log |
| GME | 0x1c8a973a8788C4A57371BEd1611830a83FD580F4 | 2,363 | not in the factory log |
| GME | 0x1a27ce5346E873478B0cB337d5Bf7861b14f88E8 | 169 | not in the factory log |
| NVDA | 0x465834D5BA3af2169E49B70A139448e59e3CA492 | 155 | not in the factory log |
| GME | 0xF6A5d9D1e62ad20C8691468500f919C5780F4278 | 109 | not in the factory log |
Holder counts read 27 August 2026 — and treat them as decoration. This chain hosts airdrop shells with 66,522 holders and one and a half cents of liquidity. A holder count proves nothing here, ours included.
And it applies to us
HOODFLY is not deployed. There is no token, no presale, no allocation, no airdrop. When the router exists its address will appear in the bar above and its source will be verified on the explorer. Until then, anything claiming to be HOODFLY is impersonating it.
What you are actually holding
Robinhood Stock Tokens are tokenised debt securities issued by Robinhood Assets (Jersey) Limited, company number 162428. They give economic exposure to the underlying shares and ETFs. In the issuer's own words they do “not grant investors any legal or beneficial rights in, or against the issuer of, those underlying securities.” The issuer states that it “is not regulated,” having obtained certain consents in Jersey.
The issuer's terms state the Products “may not be offered, sold or delivered, directly or indirectly, in the United States or to, or for the account or benefit of, U.S. Persons or to Prohibited Investors,” and that restrictions apply in the United States, Canada, the United Kingdom and Switzerland.
HOODFLY does not issue, offer, sell, distribute or redeem these tokens. It executes a swap you instruct, between tokens you already hold, and never takes custody of either side. Your eligibility to hold these tokens is set by the issuer's terms, not by us — read them at docs.robinhood.com/rhj and take your own advice.
Price feeds for these assets publish 24/5 and, in Chainlink's own words, “do not have heartbeats during off-hours.” The pools trade continuously regardless. HOODFLY does not consult those feeds — every route is priced from the pools themselves, inside the transaction that settles it.