Public demo. Live NEAR mainnet data, read only. The demo portfolios are simulated and shared by every visitor; what you ask the agent to do runs as a dry run and is never signed or sent. Operator controls are disabled here: run the project locally to use them.

Why the autopilot does not chase APR

We followed 16 of the largest NEAR pools on Rhea and 18 Burrow lending markets for 90 days, block by block, and compared the yield a pool list advertises with what people actually kept.

14 of 16

The advertised APR was higher than what liquidity providers kept

Pool lists show yesterday's fees, annualised. What a provider keeps is fees minus the loss from prices moving apart (impermanent loss) and the cost of getting in and out.

13 of 16

Pools left providers worse off than simply holding the two tokens

Fees did not cover impermanent loss. With a $500 position, 14 of 16 were negative after costs. That is why the autopilot ranks by the result after costs, not by APR.

7 of 16

Of the biggest pools hardly traded at all

A big pool is not a better pool: size barely predicts what providers kept (rank correlation 0.33, where 1 would be a perfect match). Much of that capital is parked, not earning.

11 of 18

Burrow's displayed rate mostly matches what lenders earned

Our weakest point, stated plainly: lending rates are far more honest than pool APRs. In the other 7 markets the rate moved enough that the number shown today was not what accrued.

Where the money went, by pool type

Medians, per year. “Earned in fees” is what a pool list shows; “after price moves” subtracts impermanent loss; “after costs” also subtracts getting in and out with $5,000.

Pools	Earned in fees	After price moves	After costs	Lost money
Still in the top 15 at the end (15)	0.7%	-2.2%	-3.8%	13 of 15
In the top 15 at the start only (shrank) (1)	20.4%	-27.0%	-28.1%	1 of 1
WNEAR / stablecoin (4)	11.0%	-25.5%	-26.7%	4 of 4
WNEAR / volatile token (12)	0.3%	-0.9%	-2.0%	10 of 12

What this means for you

Judge an opportunity by what is left after costs and impermanent loss, never by the advertised APR.
Position size matters: fixed costs eat small positions.
The highest fees came with the biggest losses: NEAR/stablecoin pools earned the most in fees and lost the most once NEAR's price moved.
Lending is simpler to read than liquidity provision; it still has rate and smart-contract risk.

How it was tested

Each claim had a rule, fixed in advance, that would prove it wrong. Every number is recomputed from raw mainnet data committed with the code (one block per day). Verdict: SUPPORTED on all three counts. To avoid looking only at the survivors, the sample also includes the pools that were among the largest at the start but not at the end (1).

Full method, every pool and the raw data →