This thread is basically my attempt to explain Tarun's Ohm paper. I'm going to try to make it shorter than 20 tweets. Large sections are directly plagiarized from his paper so please read it

https://people.eecs.berkeley.edu/~ksk/files/Ohm_Liquidity_Management.pdf

Olympus DAO is probably the most controversial DeFi protocol that hasn't straight up gone to zero.

Founded in early 2021 by the pseudonymous Zeus, the protocol made a big impression on retail with its 10,000% apy staking meme, and had a speculative runup characteristic of almost all farming tokens later in the year before collapsing >80% in market cap.

In order to frame what Olympus DAO and the protocols token (the Ω token) do, we must briefly discuss DeFi at large.

DeFi emerged in the late 2010s and early 2020s as a group of protocols that allow two-sided (or more markets) to form around debt issuance, automated market making, and overcollateralized lending.

In contrast to traditional finance, DeFi replaces centralized third-parties who control liquidity flows (e.g., banks and brokers) with a smart contract that manages rules for issuing debt and trading.

This comes at a great cost: namely, DeFi protocols need to incentivize actors to interact and provide capital to the protocol. For example, Aave and Compound rely on liquidators to prevent the protcols debt from exceeding its assets.

For such protocols, incentivizing protocol liquidity is equivalent to reducing protocol risk. <<<< EMPHASIS

Similarly, Curve and Sushiswap rely on incentives to slow down capital outflows and encourage long-duration liquidity. They do so by giving more incentives (in the form of governance tokens, which are equity-like instruments) to LP's who agree to deposit for long periods of time.

In effect, these lock-ups and incentivization boosters allow for protocols to "lease" liquidity for a given time frame. However, these protocols provided such incentives in an ad hoc manner that did not systematically adjust incentives as a function of market activity and volume.

Finally, another concept that became increasingly popular in 2021 is the concept of protocol-owned liquidity. This refers to the idea that a protocol should aggressively use fees and tokens to itself be a liquidity provider on AMM's, in effect "buying" liquidity forever.

Buy, lease, or rent. These are the three basic moves in the game of liquidity provisioning.

In abstract, the goal of every DeFi protocol is to get from an initial state of x capital to a desired state of x+y capital. To do so, they must figure out the most optimal (capital efficient) way to get there.

Considering that we have a problem where we're trying to get from an initial state to a desired state, and there are multiple ways to get there (renting, leasing with various durations, buying), the problem resembles a Linear Quadratic Regulator (LQR).

An example of an LQR is trying to get from your house (initial state) to work (desired state). To do so, there are multiple options. We could take a car, ride a bike, take a bus, or ride a helicopter. Each has an associated cost, and an associated time of arrival.

Which is the most optimal choice? We can't say, because we don't know what a good outcome means. Each one gets us where we need to go, but in different ways. If time is most important, then we'd choose a helicopter. If money is most important, then we'd ride the bike.

Of course, in real life we don't have infinite money or infinite time. We're trying to find a balance between the two. Maybe we'd reason that we need to get to work early today, but we're also not rich. Then we'd take the car.

If we wanted a fancy mathematical way to judge these outcomes, then we could make an equation J = Q * time + R * $, where Q and R are both scalars we use to emphasize the importance of something. We can call this the cost function.

As we can see, the weighting changes the outcome dramatically. If Q is high, we're penalizing options that take more time. If R is high, we're penalizing options that take more money. Based on those scalars, we choose the lowest cost option, which becomes our optimal solution.

This is the exact type of reasoning a protocol must make. Some buy liquidity. Others rent. Some do things in between, like Curve. It all depends on the protocols goals and current market dynamics.

Ω is an iteratively updating LQR that weights the value of various liquidity purchasing/renting operations, and uses a combination of automated execution and on-chain governance in an attempt to create a lower volatility asset.

Sometimes, Ω is put into a steady equilibrium, for example by turning off bonds and turning on inverse bonds. Other times it aggressively expands it's balance sheet by increasing staking apy and turning on a variety of bonds (i.e., issuing debt like instruments).

In doing so, the Ω protocol addresses the tension between safety and exploration as it attempts to create a Ω token with reduced volatility, capable of being used as a reserve asset.

(I don’t think Ohm was actuslly implemented the way Tarun describes the mechanism in the paper and this abstraction is not a good substitute for reading the actual paper)