A question
It began with one question. When a loan goes bad on-chain, who cleans it up, and what do they get for it?
The first code written was a few lines that check whether a position can be liquidated. Nothing more.
The story
How the arc began
A liquidation has a shape. A price falls, crosses a line, and someone is paid to be there first. liquid_arc started as an attempt to find out who that someone is, and what they are really paid.
The first one was a loss
The first real liquidation found on Ethereum repaid $16.16 of debt. The bonus was 73 cents. The gas to collect it was about $12.
So the first finding was that most small liquidations lose money. That turned out to be more useful than a win would have been.
-$11.27
what the first liquidation found would have paid
Wrong in a day and a half
A day and a half of Arbitrum data showed two liquidations, both taken by one address. It looked like a monopoly. Thirty days showed 142 liquidations and 25 different liquidators.
The short sample had landed on a quiet stretch, and one field had been misread. The rule came from that day: no number without the command that produced it.
1 to 25
liquidators, once the sample was long enough
Following the money to Base
The search moved from Ethereum to Arbitrum to Base, each time toward where the liquidations actually were.
On Base, thirteen Morpho markets repaid $1.77 million in forty days. Half of it landed on a single day.
$1.77M
repaid on thirteen markets in forty days
The auction
The plan had been to be fast. Then one auction bundle was decoded, and the winner had bid about 99% of the prize. Being fastest would have meant winning almost nothing.
The first estimate of that number had been 10%. Working out why it was wrong became the first report.
~99%
of the prize, bid away by the winner
The cliff that was not there
Large liquidations kept failing in testing, with no error message. It looked like the pool was too shallow to take them.
It was a gas limit, set a few thousand units too low. The pool had the depth all along.
700,000
the limit that looked like a liquidity cliff
Why a handle
liquid_arc is a name, not a person. That is on purpose.
If nobody knows who wrote it, the only thing left to judge is whether the numbers hold up. Every report names its transactions and ships the commands to run it again.
Now
This site is the working notebook, made public: a live feed, the tools built along the way, and the findings as they come.
The arc is not finished.
See what it found
Three pieces of research, a live feed, and two tools. All of it can be checked.