Demonstration build · all figures are sample data · no contract is deployed
Monitor

Live

What the AI is doing right now, the fee it is charging, and the evidence behind its choices.

Read from the chain and signed AI activity logs (sample data in this demo)Sample data
The AI

Shadow-testing M-019

Current activity

The AI is running a six-hour shadow test of an asymmetric fee: 2.00% on buys and 2.60% on sells. The proposed fee is not charged. Only its projected effect is measured.

2.20%Current fee
+0.30%External fee
2.50%Total per swap
Baseline
The fee returns to 2.20% whenever no test is active, or when a test fails a gate.
Cap
The hook cannot exceed 5.00% in any state.
Current task
“Measure whether a higher sell-side fee changes retention without shifting volume to unofficial pools.”
Daily objective
Raise 24 h fee revenue by at least 2% while keeping swap retention above 90%.
Activity

What it just did

Block 21,847,300Sample data
Evidence

Check the proof

Its reasoning

Hypothesis
Higher sell-side fee lifts revenue without hurting retention.
For
Sells show lower price sensitivity than buys (n=412).
Against
M-014 pushed volume to an unofficial pool for 6 h.
Next action
Promote to live test at 20% exposure.
Confidence
62%

The test bench

TestResultRevenueRetention
M-019Live+3.4%90.9%
M-018Rejected−1.2%88.2%
M-017Kept+2.2%92.0%

Last lessons

  • Raising both sides together hurts small buyers most. Keep changes one-sided.
  • Asymmetry above 0.6 pp moves volume off the official pool.
  • Weekend tests need longer windows to be comparable.

Hard limits

5.00%Fee cap
1BSupply
48 hTimelock
365 dLiquidity lock
0.30%External fee