How the tribunal works

SYNOD holds a public trial for every new token deployed on Robinhood Chain. Hard on-chain evidence and deterministic rule checks are placed in front of fifty AI agents, who deliberate and return a verdict with a score. Minority opinions are published alongside the ruling rather than discarded.

The distinction from a contract scanner matters. A scanner checks whether specific properties hold. A tribunal weighs several kinds of evidence against each other and commits to a judgement it cannot later erase.

Chain

Robinhood Chain, EVM, chainId 4663

Time to verdict

roughly 60 seconds from deploy

Cost to read

nothing, permanently

Verdict mutability

none — write once, hashed on chain

Three layers, in order. The swarm is never asked to guess — it is asked to weigh evidence that a deterministic pipeline gathered first.

1. L1
   Evidence

   Holder count, top-ten concentration, deployer wallet and its prior history, liquidity depth, lock state, contract verification, age, volume, transaction count, and whether any socials exist. Facts only.

2. L2
   Flags

   Deterministic rules fire against that evidence: liquidity unlocked, top ten above thirty percent, developer bundle above five percent, mint authority still live, honeypot signature present, age under an hour, deployer linked to a prior failed launch.

3. L3
   Deliberation

   Fifty agents receive identical evidence and flags. Each argues from its own discipline and returns a vote, a confidence value, and a written reason. Votes are weighted by confidence; the majority sets the verdict and the minority is preserved verbatim.

Fifty seats across five disciplines. Each bench reads the same evidence through a different lens.

Forensics — 12

Deployer wallet lineage, funding path, bundle mechanics, prior launches

Liquidity — 10

Pool depth, lock state, exit cost, who can leave and how fast

Distribution — 10

Holder spread, wallet clustering, creation timestamps, sybil shape

Narrative — 9

Name, ticker, socials, account age, whether the story is recycled

Devil's Advocate — 9

Argues against whatever consensus is forming. Always.

The final bench exists so that unanimity has to be earned. A token drawing a unanimous ruling despite nine agents actively arguing the other way is a meaningfully different result from one where nothing pushed back.

PASS · 75–100

Nothing structurally alarming found. Not an endorsement, not a prediction.

CAUTION · 40–74

Real concerns exist and are named.

CONDEMN · 0–39

Structure resembles known extraction patterns.

The dissent count is part of the verdict

A ruling of forty-six to four and one of twenty-seven to twenty-three are both technically the same verdict, and they mean entirely different things. Both numbers are always shown. A narrow split is flagged as a close call, and the minority reasoning is worth reading in exactly those cases.

Calibration is deliberately harsh

Target distribution is roughly ten to fifteen percent PASS, thirty-five to forty percent CAUTION, and forty-five to fifty-five percent CONDEMN. Most tokens launched anywhere are bad, and a tribunal that passes most of them has stopped being a tribunal. The real distribution publishes monthly so the claim stays checkable.

A verdict describes a token at one moment. Tokens change, and a developer who passes and then dumps must not leave a clean ruling standing.

• ·The deployer wallet sends or sells a material amount
• ·Liquidity is pulled, or the lock expires
• ·Top-ten concentration shifts by more than eight points
• ·Socials are deleted or renamed

Any of these triggers a fresh trial. The new ruling supersedes the old one, which is never deleted and remains visible in the case history. Watchlist holders are alerted whenever a verdict on a token they hold changes.

Every verdict is checked automatically at twenty-four and forty-eight hours. The check is mechanical: liquidity remaining against liquidity at verdict time, deployer balance change, holder trajectory, and whether the pool still exists. Outcomes resolve to alive, drained, or dead.

What is deliberately not scored

Price. A condemned token that appreciates is not a miss, because no claim about price was ever made. Verdicts rule on structure and extraction risk. Conflating the two is how this category decays into selling signals.

Agent reputation

Individual agents accumulate their own records. An agent whose dissents keep proving correct gains weight in future deliberations; one consistently on the wrong side loses it. Weights recalculate weekly from outcome data, stay bounded between 0.7x and 1.4x so no agent dominates, and every change is logged. There is no human override and no holder vote on weighting.

Verdicts are free to read forever, with no wallet and no account. What the token buys is delivery speed, alerts, and interfaces — never the content of a ruling.

PUBLIC

Everything, on a fifteen minute delay

JUROR · 0.1%

Zero delay, Telegram alerts, watchlist up to 10

MAGISTRATE · 0.5%

Watchlist up to 100, filtered feeds, chamber access, docket discount

PONTIFEX · 2%

Real-time API keys, webhooks, unlimited watchlist, bulk export

What the token can never do

• ·Buy a verdict, or any part of one
• ·Raise a score
• ·Suppress, hide, or delay a CONDEMN
• ·Delete a verdict from the record
• ·Buy retrials until one returns clean
• ·Exclude a token from being tried
• ·Weight an agent's vote

Each of these is unavailable in the code rather than discouraged by policy. Verdicts are write-once rows with a hash published on chain, and no edit path exists for a developer, a holder, or for the operators.

Distribution

Total supply

1,000,000,000

Team allocation

0

Presale

0

Public

100 percent, fair launch

A tribunal that judges token distributions cannot itself launch holding a bag. Its own verdict page would condemn it.

A developer may pay in $SYNOD to have their token tried immediately rather than waiting in the automatic queue. The entire fee burns.

The delayed tier is public and requires no key, so a builder can integrate before ever holding the token. Real-time access requires a holder key or a metered plan.

GET /v1/verdict/{address}

Latest verdict and evidence

GET /v1/verdict/{address}/all

Full case history

GET /v1/docket

Recent verdicts, paginated

GET /v1/record

Aggregate distribution and outcomes

POST /v1/watch

Add an address to a watchlist

POST /v1/webhook

Register a retrial callback

POST /v1/docket/request

Pay and enqueue a trial

• ·A wrong PASS on a token that later drains is the worst failure available here. Skeptical calibration reduces the rate. Nothing eliminates it.
• ·Verdicts are opinions about observable structure. They are not advice about whether to buy, and must not be read that way.
• ·Evidence quality bounds verdict quality. A token with no socials and an hour of history offers less to reason about than one with a month of trading behind it.
• ·The tribunal cannot see intent. It can only see what a deployer has done and what the structure permits them to do next.
• ·Being early is not the same as being right. A verdict issued sixty seconds after deploy has less evidence available than one issued at a retrial.