EMPRUYBSC Testnet

Agent Advantage Report

Does hiring an agent actually beat doing the work yourself? Each task below is run twice, once by the agent, once by a person, on BNB Smart Chain Testnet, and decided on time, cost and outcome. Raw output and transaction hashes are attached to every run.

0/4 tasks complete2 runs recordedChain 97

No runs recorded yet.

The experiment design below is fixed in advance so the comparison cannot be tuned after the fact. Results appear here as each pair of runs completes.

Run a 20-level BNB/USDT grid for 24 hours

Grid Runner, decided on 5 measures

Trading
With the agentNot run
By handNot runA person watches the chart and places each buy and sell by hand as price crosses a level, logging every fill.

Not run yet. Decided on: time the person spent watching, fills captured vs fills missed, realised pnl after fees, gas spent, max drawdown.

Keep a Venus loan above health factor 1.5 through a price drop

Health Guard, decided on 4 measures

Security / risk
With the agent
Time
5s
Cost
$0.00

Health factor fell to 1.3714. healthguard noticed and repaid 1.35 USDT through its bounded session, bringing the loan back to 1.5737. Response measured from the reading that breached the threshold to the transaction being sent.

Raw output ↗
By handNot runA person watches the position and repays manually when the health factor falls, including overnight.

Not run yet. Decided on: time from breach to action, lowest health factor reached, whether liquidation occurred, number of interventions.

Find and move to the best APR across five PancakeSwap pools

Yield Router, decided on 4 measures

Yield
With the agentNot run
By handNot runA person opens each pool page, compares APR by hand, then executes the move.

Not run yet. Decided on: research time, apr actually chosen, cost of the move, uplift over holding.

Rebalance a CAKE/BNB position through three range exits

Range Keeper, decided on 4 measures

Rebalancing
With the agent
Time
48s
Cost
$0.0259

Position 37300 left its range at block 129667734 (pool tick 193626 against a 193602..193622 range). The agent's exit landed at block 129667840, 48 seconds later, collecting fees and withdrawing to the owner's wallet through its bounded session. Latency was measured by binary-searching the pool's slot0 across historical blocks, it is not self-reported by the agent. Gas is the receipt's gasUsed x effectiveGasPrice, priced at the mainnet BNB/USDT pool.

Raw output ↗
By handNot runA person notices the position has left range, then collects, withdraws and re-mints by hand.

Not run yet. Decided on: time to notice and act, fees collected vs missed, time in range across the test, gas spent.

Method

  • Both runs use the same starting position, the same pool or market, and the same window.
  • The manual run is performed by a person and timed honestly, including the time spent watching for the trigger.
  • Cost counts gas actually paid plus any agent fee. All four agents are free during the hackathon, so cost is gas.
  • Every on-chain action links to testnet.bscscan.com, so any figure here can be checked at source.