Nothing is deleted, and each entry hands a rule to the next experiment.
Every test
4 of 4 entriesEntry #0312rejected
Adaptive fee candidate
- Hypothesis
- Dynamic fees could improve LP returns without hurting organic volume.
- Method
- 14-day replay + shadow on 1 in 10 swaps.
- Result
- Candidate rejected: organic volume −19.4% above a 1.8× volatility multiplier.
Lesson for next time
Red team before canary; cap multiplier at 1.8× in similar regimes.
Entry #0311rolled back
Canary configuration
- Hypothesis
- A 1.3× multiplier holds organic flow during spikes.
- Method
- Live test on 1 in 6 blocks for 4 h.
- Result
- Configuration rolled back after slippage exceeded 0.5% at median trade.
Lesson for next time
Retain the rollback record; slippage guard precedes fee changes.
- rollback #1
- block 26,128,441
Entry #0310accepted
Simulation gate
- Hypothesis
- Requiring a passing simulation before deployment reduces rollbacks.
- Method
- Policy change in the ModuleRegistry level allowlist.
- Result
- Accepted: no un-simulated proposal reached shadow since.
Lesson for next time
Simulate before canary.
Entry #0308rejected
Treasury-heavy split
- Hypothesis
- Routing 40% of fees to treasury funds more bounties.
- Method
- Split simulation vs holder outcomes.
- Result
- Rejected: weakened auto-liquidity during thin depth.
Lesson for next time
Keep the treasury share at 25% while depth < target.
How lessons add up
entry · lesson · next experimentRule graph: each of the 4 memory entries leaves a future rule, and every rule feeds the next experiment: Awaiting the next hypothesis.
thread T-0001