The thesis · $GRAIL · Robinhood Chain

That token says $2.5B market cap. Its pool holds $260. We put both numbers on the same card.

A marketplace that only counts what you can actually buy. Every listing carries its executable size, probed live against real pools at your order size, then stamped green or grey. The stamp is never for sale. Routing fees buy back and burn $GRAIL.

Positioning · the shelf vs the till

They print the price. We print what fills.

The shelf (them)

The till (Grail)

Headline market cap next to a pool holding $260

Executable size printed beside every headline number

Catalogues compiled once and shipped frozen

Catalogue re-indexed on-chain continuously, with a date

"Trade" buttons that lead nowhere for most listings

A grey stamp says plainly: this one is watch-only

Slippage discovered after the transaction

Slippage quoted before, at the exact size you typed

Issuer claims republished without a check

Claims verified against pool state at the execution block

Listings sold to issuers who pay for placement

The stamp is never for sale, at any price, to anyone

Breadth as the product

Honesty as the product

The problem

One structural lie, repeated by every venue.

A listing shows a headline valuation in the billions and sits one click from a pool holding a few hundred dollars. Both numbers are true. Only one is actionable.

Most catalogue entries are watch-only. The car, the paintings, the rental homes — listed, priced, and not purchasable by the person reading the page.

Buyers find out at the confirmation screen, not the listing page. The cost of that lesson is the spread.

Catalogues age silently. A figure sourced in August is still on the page in December with no date beside it.

Nobody is the referee, because the referee makes no money from the listings it marks down.

Grail’s bet

People do not share catalogues. People share the moment they find out they were being lied to. A grey stamp beside a billion-dollar number is that moment, rendered as a card.

Core mechanism · probe to stamp

Six stages. The receipt prints all of them.

Every venue publishes price. Grail publishes depth at size: what happens to that price when you buy $2,000 of it, right now, through the best route. The gap between those two answers is the product.

1Index
Contract, supply, decimals and pool set read directly from chain.
2Route
Every venue that can fill the pair, including multi-hop paths, ranked.
3Probe
A ladder of sizes simulated; output, impact and route recorded.
4Curve
Five points become a fill curve. Its knee is the max clean size.
5Stamp
Clean size, freshness and route count decide the stamp. Nothing else feeds it.
6Publish
Stamp, curve, block height and route on the page. All four, always.

Why the probe runs at the execution block: a quote taken “now” and a fill landed three blocks later are different trades. Every probe records the block height it simulated at; every published curve carries it. A probe older than its freshness window is not shown as stale — it is not shown at all, and re-runs on demand.

The scoreboard · grading our own calls

Anyone can publish a number. We publish how often ours was wrong.

For every swap routed through Grail, the quoted impact and the realised impact become one signed delta in basis points, aggregated per asset, per venue and per size band, rendered publicly. Positive delta means the fill was worse than quoted. We publish the bad rows because the bad rows are the proof.

Sample board · what it will look like · 2026-09-23 13:00 UTCSwaps just opened · 0 fills graded yet

Asset	Fills	Median Δ	Worst Δ	Within quote
Vaulted gold	412	+3 bps	+41 bps	96.1%
Tokenized equity A	288	+7 bps	+88 bps	92.7%
Tokenized equity B	94	+12 bps	+210 bps	84.0%
Uranium oxide token	11	+38 bps	+402 bps	63.6%

Rows above are illustrative. The real board starts empty on launch day and stays public from the first fill, however thin.

Who this is for

Three people, one number each.

The curious newcomer

›Types a number, sees what it buys in real objects
›Has never opened a DEX and does not intend to
›Converts when one card makes the abstract concrete

The sizing trader

›Already knows the headline number is decoration
›Needs the fill curve before committing capital
›Converts on the first probe that saves a bad entry

The builder

›Wants the stamp as a field in their own product
›Pays per query rather than negotiating a data contract
›Converts because rebuilding the probe costs more

Token utility · $GRAIL

Access scales. Truth does not.

The line that cannot move

The stamp is never for sale. Not for $GRAIL. Not for cash. Not to an issuer, a venue, a market maker or a partner.

There is no tier, no package and no private arrangement that changes a stamp. If that ever stops being true, the product is worth nothing and the token is worth less. Every other decision on this page is negotiable. This one is load-bearing.

Without $GRAIL

With $GRAIL

Full catalogue, free

Same catalogue, same stamps — nothing hidden

Probe capped at $500

Probe at any size

20 API queries a day

Quota by tier

Routing fee 0.30%

Routing fee from 0.225%

No alerts

Stamp-change alerts

No scoreboard exports

Full history export

Note what is not in the right column: better stamps, earlier stamps, or any influence over them.

What the token may never do

Buy a stamp, or change one — not for $GRAIL, not for cash, not for an issuer, venue or partner.
Rank a listing higher.
Vote on methodology by balance. Changes are argued in public, never decided by the largest holder.
Be required to read the catalogue or see any stamp.
Represent a claim on revenue, assets or the treasury.

Governance-by-balance over the methodology would let the largest holder set the standard. That is the failure mode we exist to point at, so it is designed out from the start.

Free

hold 0

✓Full catalogue and every stamp
✓Probe up to $500 per query
✓20 API queries per day
✓Share cards

Bronze

hold 10,000

✓Everything in Free
✓Probe up to $10,000
✓200 API queries per day
✓10 watchlist alerts

Silver

hold 100,000

✓Everything in Bronze
✓Unlimited probe size
✓2,000 API queries per day
✓Unlimited alerts + split-order planner
✓Routing fee 0.225% — 25% off, the cap

Gold

hold 1,000,000

✓Everything in Silver
✓25,000 API queries per day
✓Full scoreboard history export
✓Methodology change notice before publication

Why the discount is capped at 25%. A routing-fee discount is a real cost, paid per trade, forever. Deep discounts quietly invert the model: the heaviest users pay the least while consuming the most probe compute. Base 0.30% · floor 0.225% · never compounds with a promotion · any launch promotion is dated in advance.

Revenue model

Where money enters, and what we refuse to sell.

01the spine

Routing fee

0.30% of notional on every swap executed through Grail, on top of the underlying aggregator's take, disclosed on the confirmation screen before signing. 0.225% at Silver and above.

02the floor

Pro access

$29 per month, or hold 100,000 $GRAIL and pay nothing. Deliberately dual: the subscription smooths revenue, the hold path creates standing demand.

03the compounding one

Query API

Per-call pricing above the free tier for products that embed the stamp. Each embed is also distribution.

04one-time

Launch fees

Creator fees from the token launch, routed to treasury, earmarked for the first six months of probe compute.

Every unit of protocol revenue

Treasury50%

probe compute · node + archive reads · catalogue research · audits

Buyback & burn50%

open-market buy, then burn, batched weekly, every transaction published

Treasury is not a war chest. It is the cost of keeping the probes honest and frequent — the only reason the fee exists.

Planning model · illustrative

not a forecast

Monthly routed notional (assumed): $0
Gross routing fee at 0.30%: $0
After tier discounts (~20% of flow): $0
Pro access, 60 paid seats: $0
API overage: $0
Total: $0
Treasury share → probe + node cost: $2,495 → $1,100
Headroom for catalogue expansion: $1,395

Modelling assumptions for planning, not a forecast and not a promise. The model only works if probes stay cheap — that constraint drives the architecture below.

What we deliberately do not sell

×Listing placement. No issuer pays to appear or rank.
×Stamp upgrades. This is the revenue section and this is exactly where the pressure will arrive.
×"Verified issuer" badges sold as a subscription.
×Order flow. Routes are ranked on fill quality and the ranking inputs are published.
×User data. We do not custody funds and do not broker who probed what.

Each of these is a fast six figures and a dead product. The entire asset is that our number is not for sale.

Tokenomics · fair launch

A product that prints the uncomfortable number cannot launch with a hidden allocation.

Contract facts

Supply: 1,000,000,000 $GRAIL
Distribution: 100% fair launch, no team allocation
Presale / private round: None
Liquidity: Burned at launch, verifiable on-chain
Mint function: None. Supply only moves one direction.
Team holdings: Bought on the open market, wallets published before buying
Chain: Robinhood Chain (EVM, chainId 4663)
Status: Live on Robinhood Chain · 0xb902ec380bafc472bb07010ff90b09d7d6df67af

Demand drivers

Tier access: Probe size, API quota and alerts scale with the balance held. Using the product creates a reason to hold.
Fee relief: Up to 25% off routing, capped so it never inverts the model.
Pro substitution: Holding 100,000 replaces a $29 monthly payment — a soft floor tied to real usage.
Buyback: Half of all protocol revenue buys on the open market, then burns.

Supply reduction. Weekly burn batch from the buyback share, transactions published on the scoreboard page. API overage settled in $GRAIL is burned in full. No mint function. Burn totals rendered live from chain, never from an internal counter. Burned so far: 0 — nothing to burn yet.

Architecture

Cheap probes or no product.

Server-authoritative, and why

The probe: A client-side quote can be spoofed by the client, and the stamp is the product.
The stamp function: One implementation, versioned, with the version stamped onto every result it emits.
Tier resolution: Balance is read from chain server-side at request time, never trusted from the client.
The grader: Realised fills are matched to quoted curves by transaction hash, not by client report.
Not custody: Deliberately client-side. The user signs from their own wallet. We never hold funds.

Keeping probes cheap

Ladder, not continuous: Five sizes, not a sweep. The knee is interpolated between the two bracketing points.
Pool-state math, not simulation: Where the venue's curve is closed-form: reserves in, output out, no node round trip.
Tiered freshness: Open assets re-probe every 15 minutes, thin ones hourly, watch-only daily.
On-demand override: A user typing a size triggers a fresh probe for that asset only; the result warms the cache.

Public API surface

GET /api/v1/assets
list with current stamp, filterable by state
GET /api/v1/assets/{address}
full record: headline, stamp, latest curve, issuer
GET /api/v1/probe?address&size
on-demand probe: output, impact, route, block
GET /api/v1/curve/{address}
the five-point ladder plus the interpolated knee
GET /api/v1/stamp/{address}
the stamp alone, for embedding — the lightest call
GET /api/v1/scoreboard
realised vs quoted deltas, per asset, per window

Every response carries engine_version, stamp_version and the block height it was computed at. A number without those three is not auditable.

Security posture

›Non-custodial by construction. No deposit address exists, so there is nothing to drain.
›No private keys on any server. The burner uses a hardware-backed signer with a capped weekly allowance.
›API routes fail closed. An unauthenticated call to a paid endpoint returns 401, never a degraded free result.
›Read-only database role for anything serving public pages.
›Framework kept on the current patched release line, dependency scanning on every deploy.
›Containers run as a non-root user with no host mounts.

Anti-gaming

Attack, and what it runs into.

Attack

Defence

Wash routing to farm tier status

Fee rebates capped per wallet per day, so round-tripping costs more than it returns.

Probe spam

Free tier rate-limited by address and origin, size capped; heavy use must authenticate.

Liquidity theatre before a probe window

Flagged; the stamp uses a time-weighted read rather than a spot read.

Stamp laundering via stale embeds

Embedded stamps carry a version and timestamp; a stale embed renders as stale.

Self-dealing

Team wallets published before they buy, and graded on the scoreboard like any address.

Roadmap

Six phases. The first one is where we are.

01
The probe
weeks 1–3in progress
- ›Route planner + venue adapters for Robinhood Chain
- ›Five-point ladder probe, closed-form math path first
- ›Stamp function v1, methodology written before it ships
- ›~20 assets indexed
Exit test: A probe on a known-thin asset returns a knee under $500, and a manual swap at that size lands within the quote.
02
The front door
weeks 3–5next
- ›"What does this buy?" page, nine cards, no wallet needed
- ›Share card rendering
- ›Markets table with stamp filters
- ›Public methodology page, versioned from day one
Exit test: A stranger types a number and can explain green-versus-grey back, unprompted.
03
Execution
weeks 5–8planned
- ›Swap from self-custody through the routing engine
- ›Fee collection wired and disclosed pre-signature
- ›Grader online; scoreboard published, bad rows included
- ›Watchlists with stamp-change alerts
Exit test: The scoreboard has enough fills to be embarrassing and we publish it anyway.
04
The token
weeks 8–10planned
- ›$GRAIL fair launch, liquidity burned at launch
- ›Tier resolution live, balance read server-side
- ›First weekly buyback and burn, transactions published
Exit test: Team wallets published before any open-market purchase.
05
The catalogue
months 3–6planned
- ›Issuer dossiers: custodian, jurisdiction, attestation age
- ›Catalogue widened by demand
- ›Query API with per-call settlement
- ›Split-order planner past the knee
06
Beyond one chain
months 6–12planned
- ›Per-chain route sets and venue adapters
- ›Cross-venue comparison for the same underlying
- ›Embeddable stamp widget
- ›Permanent archive of every stamp change

Metrics

What we watch. What we refuse to count.

Primary

Probes run per day
Share cards generated per day
Notional routed through the engine

Secondary

Median realised-vs-quoted delta (lower is the product)
Third-party surfaces rendering the stamp
Assets moving between stamp states per week
Paid seats and API calls above the free tier

Deliberately not a metric

Number of listings. Breadth is how the incumbent lies. A catalogue of 2,000 entries where 40 are purchasable is worse than a catalogue of 20 that are.

Assets stamped

1,936

80 measured against a live router

Fills graded

swaps through Grail are not live yet

Token

Live

0xb902…67af · Robinhood Chain

Key decisions

Decided, and why.

Show every asset, including watch-only ones

The contrast between green and grey is the product. A catalogue of only-green would be honest and unshareable.

Routing fee 0.30%

Above marginal probe cost at realistic volume, below the point where a trader routes around us to save basis points.

Discount capped at 25%

Deeper relief inverts the model: the heaviest users would pay least while consuming the most compute.

Pro is dual — pay or hold

Subscription smooths revenue; the hold path creates standing demand tied to real usage.

~20 assets at launch

Twenty assets probed well is a position; two thousand probed never is a spreadsheet.

Stamp is never purchasable

The only asset is that the number is not for sale. Every revenue line respects this.

No balance-weighted vote over methodology

Letting the largest holder set the standard is precisely the failure we exist to expose.

Publish our own misses

A referee that never grades its own calls is just an opinion.

Non-custodial, always

Nothing to drain, nothing to license, nothing to freeze.

Competitive edge · and the honest take

Six dimensions, graded. Then the part that could kill it.

First mover

moderate

Nobody publishes executable size beside headline valuation in this sector. The window is real but closes the moment somebody copies it.

Network effects

strong, delayed

Every embedded stamp renders our judgment on a surface we do not own. The more surfaces, the more expensive it gets to displace the standard.

Data

strong

The realised-versus-quoted history cannot be backfilled. A competitor launching in a year starts with zero fills against our year of graded calls.

Switching costs

weak, honestly

A trader can use us for the probe and route elsewhere. The defence is that routing through us is the shortest path from the number to the trade.

Brand

strong if earned

"The one that tells you what you can actually buy" survives exactly as long as we keep marking down assets that would rather we did not.

Technical