I spent three weeks auditing the interest rate model of Aurelius, a lending protocol that raised $50 million in March 2026. The pitch was seductive: dynamic rates driven by a proprietary oracle aggregator, supposedly reflecting real market supply and demand. After mapping the smart contract logic, I found a different truth. The model is nothing more than a linear interpolation between two arbitrary coefficients—coefficients chosen during a two-hour governance vote. No real market data. No on-chain feedback loop. Just a parameterized formula that can be tweaked to favor borrowers or lenders depending on who controls the DAO.
This is the structural flaw that bull markets breed. When capital flows freely, nobody questions the underlying equations. Aurelius’s TVL hit $2 billion in its first month, driven by a liquidity mining program offering 800% APR. The yield was not sustainable—it was a subsidy paid for by the protocol’s treasury, fueled by token inflation. The same pattern I saw in 2020 with the Protocol A collapse, where I simulated impermanent loss scenarios and warned my firm about a yield that was mathematically equivalent to a rug-pull. They ignored the data then. Today, Aurelius’s token is down 70% from its peak, and the treasury is nearly depleted. The interest rate model never adjusted because it was never tied to actual lending demand.
Here is the technical breakdown.
Aurelius uses a two-variable interest rate formula: r = r_base + (r_multiplier * utilization_rate). Utilization rate is defined as outstanding loans / total deposits. R_base and r_multiplier are set by governance and stored as mutable variables. There is no secondary oracle for alternative data, no volatility adjustment, and no time decay. In my audit, I discovered that r_multiplier had not changed in 78 days despite a 40% drop in overall borrowing demand. The formula produces a flat yield curve regardless of market conditions. Liquidity is a mirage; solvency is the only truth. Depositors are now earning 0.3% APY while borrowers pay 12%. The spread embeds the protocol’s hidden insolvency.
Let me walk you through the math. During the first month, utilization was artificially high due to the liquidity mining program—borrowers were incentivized to take loans even at negative real rates. The formula’s r_multiplier was set to 0.04. At 80% utilization, the base rate of 5% plus 3.2% gave 8.2% borrow APR. With the mining reward, effective borrowing cost was negative. Once mining ended, utilization dropped to 35%. Now the rate is 5% + (0.040.35) = 6.4% borrow APR, while depositors get 6.4% 65% (protocol fee) ≈ 4.16% (after fee). But actual lending demand is weak—collateral prices are dropping. The protocol is accruing bad debt disguised as interest income.
I do not trust the pitch; I audit the structure. My 2021 PixelFlux investigation taught me that code is the only truth. The same principle applies here. I traced the governance proposal that set the initial parameters. The proposer was a wallet that held 1.2 million AURE tokens allocated in the pre-sale. The vote passed with 78% approval from a single wallet. KYC? The project boasts about its doxxed team, but KYC is theater—my 2017 ICO audit experience showed that even founders with verified IDs can write broken code. Compliance costs are passed to honest users; the structural risk remains.
The contrarian angle: what did Aurelius get right? The team built a functional UI, maintained good uptime, and actually paid out yields for three months. The code itself is solid—no reentrancy bugs, no integer overflows. The failure is not technical in the execution sense; it is a failure of economic design. The model is arbitrary, but the team argued that simplicity was a feature. They even published a blog post saying "rates should be predictable." Predictable does not mean rational. I can predict a random walk; that doesn’t make it useful. Their core insight—that DeFi users want stability—is correct. But stability without market feedback is a corpse in a suit.
Emotion is a variable I exclude from the equation. The Aurelius community is angry at me for publishing this audit. They accuse me of FUD. I am not predicting a crash; I am identifying a structural guarantee of eventual insolvency. The only question is timing. If a wave of liquidations hits and utilization spikes, the rate model will respond, but too late—the damage will be done. If borrowing demand continues to drop, the protocol’s revenue will become negative. Either outcome is a decline.
Takeaway: DeFi builders must stop treating interest rate models as arbitrary sliders. Real market data is available—chainlink, makerDAO’s peg stability module, compound’s historical utilization curves. Aurelius could have used a weighted average of multiple sources. They chose not to. That choice is a red flag. The market will eventually correct. The only question is who is left holding the bag.