Learnings From Last Week - Delphi Digital

Last week, Terra’s $UST fell prey to an algorithmic stablecoin death spiral, wiping out nearly $40B in market cap across $UST and $LUNA. It’s unclear at this point whether it was a coordinated attack by a large player or a classic bank run, but it hardly matters either way. This is arguably the most catastrophic event to happen to crypto since Mt Gox.

Our core mission at Delphi has always been to make crypto happen better and faster, which is why last week’s events have had such a disheartening impact on us all. The irony that something we believed in caused the opposite to occur ripped at the very fabric of who we are and what we stand for.

We always knew something like this was possible, and we tried to stress the risks to a system like this in our research and public commentary, but the fact is we miscalculated the risk of a “death spiral” event coming to fruition. We’ve taken some heat for this over the last week, and we deserve it. The criticism is fair and we accept it.

In this post, we’ll provide transparency around our involvement with Terra across the various Delphi divisions, explain what our thesis was and where we think the thesis went wrong, as well as lessons we’re learning from this going forward.

Delphi’s Involvement With Terra

Before we start, it’s important to provide some context on Delphi and how we’re structured. “Delphi Digital” is a collaborative association of independently owned and operated entities sharing the “Delphi” brand and certain shareholders, resources, personnel and values. The “Delphi Ventures” entities, “Delphi Research” entities and the “Delphi Labs” entities collaborate under the “Delphi Digital” brand. Understanding this separation is key since each division has very different levels of involvement with Terra. Below, we’ll go through the facts of each entity’s involvement.

Delphi Ventures

Delphi Ventures Master Fund purchased a small amount of $LUNA (~0.5% of our NAV) on the secondary market in Q1 2021. Net, we only increased our exposure since our original purchase and are currently sitting on a large unrealized loss. Even at LUNA’s peak price this year, LUNA and other Terra assets made up only ~13% of NAV across Delphi Ventures. On a deal-count basis, less than 5% of Ventures’ total number of deals were in companies or protocols related to the Terra ecosystem. This includes Delphi Ventures’ participation in the recent LFG raise (February 2022) – a $10M investment which, based on the current LUNA price, is entirely lost. Delphi Ventures did not sell any LUNA during this event.

Delphi Research

First, a disclaimer. Our core goal with the reports we publish under Delphi Research is to help open source the learnings of our team and give people an in-depth understanding of how these protocols work and the role they play within crypto. As mentioned at the top of every post, these reports are for informational purposes only.

Delphi Research first covered Terra in February 2021. Since then, we’ve released six reports focused on Terra. This is a very small portion of our research output over the relevant time period–for context, Delphi released 30 posts in April 2022 alone (not counting our free Daily Newsletter). Each report included a section at the top disclosing Ventures’ positions and Delphi Labs’ involvement within the ecosystem when applicable. Furthermore, many of the reports had sections focused on the risks of a potential UST depegging. We will neither delete nor edit any of our Terra reports, and instead have made them publicly available so anyone can go back to read our analysis and make their own judgments. We have shared links below in the order of them being published:

2021 Reports focused on Terra or it’s ecosystem:

2022 Reports focused on Terra or it’s ecosystem:

Valuing Layer 1s – Memes, Money, or More? (04/26/2022)

It’s important to note, Delphi Research has not and will never accept payment to publish research. From a financial perspective, the Research entity’s balance sheet was unaffected by this event, except for ~20,000 in UST payments we received for subscriptions through Suberra. We have not sold any of the UST.

Delphi Labs

Unsurprisingly, Delphi Labs, our software research and development company, was where most of our Terra exposure was concentrated. We decided to build on Terra because we believed that decentralized money had the greatest chance of succeeding if it were to be integrated at the L1, focused on real world adoption, and built on a relatively scalable and interoperable blockchain. When we first started researching the Terra ecosystem (Q1 2021), it was growing quickly and seemed to be filling a unique place in the Cosmos ecosystem, but lacked key primitives and had attracted far fewer builders than up-and-coming comparables like Solana. At this time, UST supply was under 1 billion.

Labs spent more than a year of research and development time contributing to joint ventures building the Astroport and Mars protocols on Terra. Delphi Labs has never sold any of the tokens it holds and has made no money selling tokens. Delphi Labs was primarily funded by internal capital contributed by individual Delphi Labs equity holders, other than a no-strings-attached grant of 30,000 LUNA (roughly $250K at the time) and 466,666 UST from TFL for work on the Mars Protocol.

Despite the outcome, we’re extremely proud of the Delphi Labs team for the quality of the protocols they helped build and the way the team handled themselves throughout. In the past year, the Labs team has contributed to many innovations to the space including:

Lockdrop & LBA: An innovative token launch mechanism which has since been used by many protocols across multiple chains.
C2C Lending: Creating a transparent, quantitative risk framework around smart contract to smart contract lending, resulting in the first third party protocol leveraging Mars credit lines.
Open Source Branding: Both the Astroport and Mars brands are open-sourced under creative commons licenses, allowing anyone to use, edit, or fork the brands.
Transparency: Astroport and Mars took an approach of extreme transparency. Initial builder allocations were openly disclosed, including amounts, terms, and voting rights. All legal documents, including token grant agreements and multi-sig agreements, were open-sourced on the projects’ respective githubs. These included unique clauses around enforcing duties to participate in the DAO and vote independently. Heavy disclaimers and disclosures were produced outlining all possible risks faced by users, including the risk of a UST de-peg.
Many other open-source research contributions such as dynamic interest rate models based on control theory, attack costs of TWAP oracles, frameworks for setting AMP parameters on stableswap pools, and much more.

Overall, we’re proud of how the protocols performed despite the extreme conditions. For instance, Astroport’s smart contract system and the Delphi-Labs-run front-end for Astroport withstood incredible load and, with Terraswap (the other AMM on Terra) going down, became the only avenue for people to exit their positions on Terra.

As for the future, after making a big bet on Terra and failing, we want to make sure we learn our lessons and make the right choice on where to focus our efforts. We’ve put together a cross-sectional team of some of our brightest minds across Research and Labs and we’ll be taking our time to ensure we assess all possible options and make the right long-term decision. We’ll be documenting this publicly in order to share our journey and learnings with the community. In the same spirit, we also wanted to spend some time publishing a post-mortem of sorts in which we walk through our history, thesis and thought processes relating to Terra, why we bet big and where we think we went wrong.

Finding Terra

We first began studying Terra in early 2021, with the Research report we linked above on Mirror Protocol. While we thought Mirror was interesting, what really piqued our curiosity was the underlying network Terra – a set of algorithmic stablecoins natively integrated into the L1 economics, focused on real world adoption, and built on a relatively scalable and interoperable blockchain.

Terra had started off building demand for its stablecoins via CHAI, a payments network in Korea which at the time was facilitating $1.5B in annualized transaction volume and had over ~2.5M users. At the end of 2020, Terra had added support for CosmWasm smart contracts, allowing third parties to build applications centered around these stablecoins. With its first application Mirror giving users exposure to synthetic real world assets and gaining significant traction, we realized the promise of an ecosystem focused around creating real world use cases for decentralized stablecoins.

The ecosystem was growing rapidly and the strength of the community was undeniable. The community element was something we especially viewed as a massive strength as it can’t be created simply through incentives. Despite these tailwinds, the ecosystem lacked builders and key financial primitives such as exchange and credit. We understood these primitives deeply through our time spent researching, investing, and building in the space, and decided we would help contribute to building them via Labs.

These key primitives would enable builders to create new products that provide additional forms of utility for UST. The goal was for UST supply to then scale in line with the increased utility.

Anchor and Thesis Drift

Anchor launched in March 2021. Initially, it fit perfectly into our thesis. The idea was to use proof-of-stake assets as collateral to provide a stable yield to depositors. This allowed individuals to help secure proof-of-stake networks while giving them the optionality of forgoing staking yield in exchange for improved borrowing rates.

Anchor began at a time when 3 digit + APY yield farms were more prevalent, so an APY of 20% was thought of as an appealing but reasonable starting point to bootstrap traction. Anchor’s reserve existed to absorb the imbalance between yield paid to UST depositors and yield received from the staked assets from borrowers. When operating at a deficit, the protocol would distribute ANC to incentivize additional deposits of staked assets.

The yield farming era passed and yields compressed, but Anchor’s rate stayed at 20%. This caused UST deposited in Anchor to keep growing, which only accelerated the reserve deficit. In response, the initial rebalancing mechanism was abandoned in favor of direct donations into the reserve from Do. Ultimately, this was the classic algo stable mistake. Using unsustainable incentives to bootstrap supply is obviously effective in good times, but the size of the immediately callable liability during the bad is often underestimated. This is primarily because of the domino effect of a bank run, which dismantles confidence among even some of the strongest hands.

This was why we saw LFG and BTC reserves as a massive step forward to mitigate against that. It effectively converted some of the excess UST demand into exogenous reserves that could be used to defend the peg if necessary. Creating partial exogenous collateralization while bringing down Anchor APY would materially reduce the systemic risk in the network. We believed a high level of external collateralization was a necessity in the long run, and we saw this as a path to get there. Unfortunately it didn’t grow fast enough compared to UST supply, and, combined with a fall in value of the BTC reserves, the liability overhang was too large to be defended.

Looking Ahead

Ultimately, we believed in the Terra ecosystem, we made a big bet on it, and it didn’t turn out as expected, costing us significant money and time. Fortunately, Delphi is fully self-funded, and when we make high-conviction bets like this it’s our own capital at risk. We understood the risks of the algorithmic model upfront and sought to be transparent about them throughout; however, it’s clear we miscalculated the risks. To the vocal critics of Terra’s algorithmic design–you were right and we were wrong.

Anyone who knows us – as a business and as individuals – knows how heartbreaking it is for us to see the space we fight so hard to push forward be set back by events like this. We stand committed to doing whatever we can to leave a positive impact on crypto and the world. We’ve always said actions speak louder than words, so we’ll let our work and efforts do the talking for what comes next.