Delphi Labs presents: a Dynamic Interest Rate Model Using Control Theory
In this piece, we explore an alternative pricing solution which we believe to be more capital efficient and better suited to the dynamic crypto market. delphidigital.io/reports/dynami…
Lending OGs such as @compoundfinance and @AaveAave typically use a fixed price curve where the interest rate (IR) is determined according to the utilization rate of each money market.
While this model has proved useful and was a clever initial approach to the pricing problem, it has some limitations.
Specifically, it can be too rigid for the constantly evolving crypto market and cannot adjust to changes in external market conditions.
This rigidity can translate into 2 undesirable outcomes:
1⃣ Illiquidity in money markets where the IR doesn’t adjust quickly enough (i.e. when new yield farms pop up)
2⃣ underutilization of certain markets
So, how do we propose to solve this?
By incorporating control theory into the pricing mechanism.
Specifically, we propose using a PID controller that dynamically adjusts interest rates to target an optimal utilization within each money market. Let’s explore 👇
Simply put, the PID controller works as follows:
1⃣ It calculates the difference btwn the optimal utilization + the current utilization
2⃣ It adjusts the IR accordingly. All else being equal, the higher the difference, the higher the IR adjustment
3⃣ It repeats 1⃣ periodically
In contrast to the prevalent pricing model across DeFi, the PID model dynamically adjusts to market conditions.
Given that it doesn’t depend on a fixed curve, IRs within this model will keep adjusting whenever the current state is different than the desired one.
This model will be implemented by @mars_protocol, a lending protocol on top of @terra_money that’s currently being incubated by Delphi Labs.
If you’re a DeFi builder or part of a protocol and are interested in experimenting with this model, please reach out to us; we want to hear from you!
Special thanks to @euler_mab from @euler_xyz for his valuable help revising this report. The initial idea of using a PID Controller within the lending context came from his work in Euler XYZ.
Hyperliquid just dodged a $13.5M bullet—but it exposed a critical flaw in decentralized trading.
Here's how one trader almost broke the system and how we can stop it from happening again.
🧵
1/ An attacker opened a large short position on JELLY, then artificially pumped its spot price, forcing liquidation.
This pushed an unrealized $13.5M loss onto Hyperliquid’s liquidity pool (HLP), as the oracle price spiked from $0.0095 to ~$0.50 per token.
2/ Hyperliquid intervened by delisting JELLY perps and force-settling positions at the original price of $0.0095, protecting HLP and leaving the attacker at a loss.
But rather than just reacting, what steps can Perp DEXs take to mitigate future risks?
AI agents are evolving from simple assistants to fully autonomous entities.
@ElizaOS is leading this shift by giving agents the ability to manage funds and operate businesses in Web3.
Here’s how ElizaOS v2 is shaping the future of AI-powered economies.
🧵
1/ AI Independence
ElizaOS started as an AI framework focused on Web3 automation. While v1 enabled AI agents to interact with smart contracts and blockchain data, v2 takes a major leap forward.
AI agents have moved on from simple commands—they’re independently managing workflows, businesses, and financial strategies.
2/ Architectural Upgrades
• Modular Core Framework: Developers can customize AI agents without modifying the core to make deployments more scalable.
• Unified Abstraction Layer: AI agents now handle multi-chain assets seamlessly.
• Event-Driven Architecture: AI agents can react to real-time data, making them more efficient in handling DeFi, governance, and logistics.
These improvements give AI more flexibility, planning capabilities, and the ability to execute more complex tasks.
Imagine an ecosystem where rollups are truly interconnected—sharing liquidity, messaging, and infrastructure without barriers.
@initia is making that future a reality.
🧵
1. Appchains Have Struggled
Appchains are expensive and time consuming to maintain. They require validators, block explorers, wallet integrations, and dev ops.
RaaS solutions (e.g., Caldera, Conduit) offer basic infra but lack key ecosystem components.
2. Initia fixes this by developing a fully integrated ecosystem to ensure rollups can interact seamlessly.
It offers:
• Standardized cross-rollup communication via IBC & LayerZero
• Single Slot Finality for low-latency confirmations
• Gas abstraction via JIT—pay gas with any token
• Native USDC via Noble & Circles CCTP
• Full ecosystem ready before mainnet
• Developer-friendly AgnosticVMs & Opinionated Rollup Framework
AI agents are no longer just experiments. To succeed in real-world applications, they need a framework built for scale and performance.
How does @arcdotfun compare to the competition? Let’s take a look.
🧵
1/ Arc is a Rust-Based Agent platform designed for speed, reliability, and security.
Through hardware-level execution speeds, built in protection against memory errors, and secure concurrency, Arc is able to enhance reliability and scalability for demanding apps in finance, healthcare, and real-time systems.
2/ Arc's architecture allows it to excel in areas where Eliza & ZerePy fall short. Arc's decentralized workflow supports trustless, low-latency operations that are ideal for apps like DeFi arbitrage agents.
Arc's real-time performance ensures sub-millisecond latency for applications like robotics, autonomous vehicles, and IoT systems.
A deep dive into how MegaETH is bringing TradFi speed to DeFi.
🧵
1/ Algorithms dominate financial markets, executing trades in microseconds —giving CEXs a historical advantage.
But that is changing. In Jan 2025, the DEX/CEX volume ratio hit 20%, an ATH.
2/ Demand for high-performance DEX infrastructure is rising, driven by long-tail assets, memecoins, & institutional-grade DEXs like Hyperliquid & Raydium.
MegaETH closes the latency gap, enabling institutional-grade trading on chain.