depeg_cascade_usdc_demo

Forked Ethereum mainnet at block 19,000,000, 8 simulation steps (1.6 minutes of mainnet block time). Seed 42.

Borrowers
5
Rebalance actions
5
successful repays/supplies
Horizon
8 steps
Wall-clock
4.5s

Borrower scenario

This run opened 5 borrower positions against a deployed lending Pool (Aave V3 / SparkLend / Compound V3) and observed 5 successful rebalance actions over the 8-step horizon — HF only drifts via interest accrual under nominal conditions so most steps are no-ops by design.

Per-borrower HF / debt / collateral trajectories are not currently captured in data/runs/runs.duckdb; the actions above are the available signal until a per-step borrower-state hook lands.

Methodology

Reinforcement-learning agents

The agents in this run use the hand-coded / scripted baselines (see the Agent column above). Mayavi’s agents are RL-trainable on the same forked-mainnet stack — mayavi train --env aave|vesting|liquidator produces a PPO policy, and VestingRecipient(policy_path=…) loads one into a scenario. Trained-policy-vs-baseline evaluation results (each on a real forked mainnet, $0 marginal cost):