SPILLOVER Open copy lab
Copytrading powered by an onchain intelligence protocol.Live onchain

The edge should
spill over.

Choose a public wallet to follow. When it trades, Spillover can place the same trade from your wallet immediately. Both enabled paths enter from that signal without waiting for AI. The base path holds; the shared agent researches after entry and decides whether to keep or close its position. Every consenting result improves the same global brain.

Start live copy ↗ See how it works ↓

Learning consent cannot move funds. Live execution needs a second, bounded permission on the same wallet.

SAMPLE SHADOW COMPARISON / 024 NOT LIVE PERFORMANCE
RAW COPY+2.4%
AGENT PATH+4.1%
EDGE+1.7pp

Sample data only. These numbers are not live results or a performance claim.

00 / HOW IT WORKS

One source trade.
Two matched entries.

Both paths start from the same source trade. Both paths buy immediately from the same source signal, so AI latency cannot distort the entry comparison. The base path holds while the agent researches what to do next.

00 / SOURCE SIGNAL Public wallet buys on Robinhood Chain SAME TRANSACTION OBSERVED BY BOTH PATHS
01 / BASE COPYBENCHMARK

Enter immediately.
Hold the benchmark.

The base path places the same trade immediately within the limits you set. It makes no judgment of its own, which gives us a clean benchmark for whether the shared agent helped.

02 / SHARED AGENTGLOBAL BRAIN

Enter immediately.
Research what happens next.

The agent researches after entry, checking the position against everything learned from consenting users. Keep the position or close it as soon as the entry confirms: those are its only choices. Research never delays the purchase.

Paper agent: no second spend

It records the same immediate entry quote, then simulates the agent’s keep-or-close decision without buying again.

Real agent: uses only its assigned share

A 50/50 split divides one $10 budget into $5 for the base copy and $5 for the agent. It never turns $10 into $20.

CURRENT CAPABILITY Both entries submit before agent research. A close decision waits only for verified entry settlement, then exits the agent position as soon as possible.

01

You opt in first. The shared agent learns from your runs only after you sign consent. You can revoke it.

02

Every result teaches it. Wins, losses, kept positions, and fast closes all become evidence.

03

The agent has to prove itself. New policies stay in shadow until they beat raw copying on trades they never saw.

01 / GLOBAL BRAIN

Nobody starts
from zero.

CUMULATIVE BY DESIGN

With consent, every run reports back what the source did, what the agent changed, and what happened after costs. Validated lessons become one shared policy, so every later trade inherits the network’s full learning history.

HOW THE NETWORK LEARNS

One global loop. Every tokenholder contributes. Every tokenholder benefits.

  1. 01 / SIGNALOBSERVE

    Analyze once.
    Share everywhere.

    Each new public source trade enters one global stream. Both entries start first; Spillover then analyzes the open agent position once and applies the post-entry decision across every eligible follower.

  2. 02 / BASELINECOMPARE

    Make the agent
    earn the change.

    Base copy and agent paths enter from the same signal. Gas, fees, slippage, agent closes, and avoided losses stay in the comparison.

  3. 03 / MEMORYLEARN

    Every outcome
    stays useful.

    Results mature at 5m, 30m, 4h, and 24h. Wins, losses, holds, and closes become cumulative evidence—not isolated anecdotes.

  4. 04 / POLICYPROMOTE

    Prove it unseen.
    Then ship it.

    A new policy reaches live decisions only after enough matured outcomes, an unseen holdout, positive confidence, and no extra drawdown.

THE COMPOUNDING UTILITY

The network learns once.
Every holder gets the update.

That is the flywheel: more consenting use creates more paired outcomes; more outcomes produce better-evidenced policies; every approved policy makes the same shared brain more useful to the entire network.

05 / TEACH THE BRAIN

Bring an insight.
Prove the edge.

Your insight enters quarantine. Add a public evidence link and explain what the chain data may be missing. It is shadow-tested first and cannot change live policy by itself.

AGENT EDGE

Did judgment beat reflex?

A matched agent leg must outperform the instant base copy after gas, fees, and slippage—not just look better in hindsight.

KNOWLEDGE EDGE

Did your input improve the brain?

Your evidence can earn future credit only if it survives review and produces a measurable improvement on trades it never saw during training.

At launch, verified agent and knowledge edge are intended to participate in the planned 10% Edge Pool. Nothing is claimable today.

LIVE NETWORK LEDGERSYNCING
—SOURCE TRADES ANALYZED
—PAIRED SHADOW RUNS
—MATURED 24H OUTCOMES
PROMOTION GATE— / 200
02 / COPY LAB

Copy the trade.
Make the agent earn it.

The base and agent entries start immediately from the same source signal. The agent then researches the open position and chooses whether to keep it or close it quickly. Both outcomes stay separate so the brain has to prove its edge after costs.

COPY LAB

Copy console

Robinhood Chain · 4663
SOURCE WALLET
7A
0x7A19…F42CSample wallet · preview data
82SAMPLE
30D RETURN+18.6%
WIN RATE61%
MAX DD−7.2%
TRADES43
SAMPLE COMPARISON

Shadow baseline vs agent run

SHADOW
PATHVALUERETURNMAX DDSTATE
Raw shadowComparison baseline $1,024.20+2.42%−4.8%RUNNING
Agent runShared policy v0 $1,041.10+4.11%−3.1%+1.69PP
×
SAMPLE AGENT DECISION · 14:32:08Closed MICA after a contract risk check

Both paths entered immediately. Post-entry research found elevated contract risk, so the agent exited while the base benchmark stayed open.

84%CONFIDENCE
YOUR GUARDRAILS
✓ Risk checks on✓ Direct Uniswap swaps✓ Starts from next trade

Illustrative interface. No wallet or API is connected on this page.

Homepage preview. Open the Copy Lab to connect an execution wallet and start a real monitor. Open copy lab ↗
03 / FEE ENGINE

Trading fees
pay for the product.

At launch, every SPILL swap will use a 1.00% Uniswap v4 pool fee. The contract will split collected fees: 40% runs the agent, 10% funds the Edge Pool, 30% goes to the founder, and 20% stays in reserve. The system is verified on Robinhood testnet. Mainnet is not deployed.

1.00%POOL FEE
BEFORE PROTOCOL FEE
40%COMPUTE + OPS

Pays for the agent, wallet indexing, RPC, monitoring, and cloud runtime.

10%EDGE POOL

Reserved for future rewards tied to verified agent wins and useful evidence. Claims are not live yet.

30%FOUNDER

Fixed creator revenue, visible in the fee split.

20%PROTOCOL RESERVE

Covers audits, legal, incidents, and liquidity support.

VOLUME MODEL

What does the flywheel produce?

MONTHLY SPILL VOLUME$1.00M
SPILL LP FEES$10,000
COMPUTE$4,000
EDGE POOL$1,000
FOUNDER$3,000
RESERVE$2,000

Illustration only. Real revenue depends on volume, protocol fees, asset prices, fee collection, taxes, and operating costs.

THE PREMISE
Whales leave a wake.
We build with what spills over.

Big traders already leave their moves onchain. Spillover lets tokenholders copy those moves, compare them against one shared agent, and turn every result into a lesson the next person gets for free. Token volume pays the bill.

Start live copying ↗