Key Takeaways
A nearly $300 million bridge exploit in April 2026 showed how an attacker can borrow against stolen assets across lending markets
Junior capital absorbs losses like these before they reach lenders. How much of it a lending market needs, and where it should sit, depends on which collateral can borrow from which pool.
Today's two standard lending designs both tie up more capital than they need to
Isolated pools leave lending money unused, while shared pools let one failed source reach every pool. In both designs, the extra capital needed grows in step with the number of borrowers.
In this paper, we show that dispersed sparse permissions are optimal: extra capital grows with the square root of the number of borrowers
We call the extra cost of covering an attacker who chooses where to strike the adversarial default premium (ADP). In counterfactual redesigns of real lending markets around three recent incidents, the dispersed design needed up to 34.5% less capital than a shared design, and up to 1.5% less than isolated pools.
Research
View the full presentation
Read the full paper
