1. Crypto Exchange FTX Files Chapter 11
Two weeks ago, proprietary-trading firm Alameda and crypto exchange FTX were notable survivors of the Terra/ Luna ecosystem’s collapse and the subsequent insolvencies of Celsius and Three Arrows Capital. Many investors considered Sam Bankman-Fried’s (SBF’s) companies as crypto industry saviors after he made an effort to, in his words, “protect the digital asset ecosystem” with various bailouts of troubled players like crypto lenders Voyager and BlockFi.
Yet, in just nine days, SBF’s crypto empire collapsed. On November 2, Coindesk published leaked balance sheet information that revealed Alameda’s significant exposure to illiquid assets––including ~$5 billion in FTX’s native exchange token FTT––against ~$8 billion in liabilities. Curiously, Alameda’s purported exposure to FTT was larger than its circulating supply. It quickly became clear that FTX and Alameda were more closely tied than the public knew.
Upon this and perhaps other undisclosed news, FTX competitor Binance announced that it was liquidating FTT, a sizable position from a venture investment in FTX that it had exited the prior year. Thus began the run on the bank. Despite SBF’s assurances last Monday that FTX had ample assets for withdrawals, customers withdrew billions until the exchange halted withdrawals on Tuesday morning, and SBF went silent.
Behind the scenes, FTX was suffering a severe liquidity crunch and SBF was frantic for a deal to save the company. He found few takers. Competitor Binance considered a deal but passed. With all options exhausted by Friday, SBF announced the bankruptcies of the international and US arms of FTX and Alameda.
Apparently, the events leading to these bankruptcies began months earlier. Some reports suggest that, unethically and likely illegally, SBF transferred $4 billion of FTX customer deposits and other assets to Alameda to meet margin calls during the deleveraging around Terra/Luna’s collapse.
Catching most in the crypto industry completely off guard, FTX’s collapse is beginning to reverberate. BlockFi has halted withdrawals; Voyager is looking for new bidders; and Genesis has disclosed ~$175 million trapped in FTX. The full extent of the contagion could take months to play out.
In our view, FTX’s insolvency is one of the most damaging events––potentially worse than the 2014 Mt. Gox hack––in crypto history. Caused by one of the revered leaders of the industry, this collapse has impacted crypto’s reputation dramatically. It could delay institutional crypto adoption by years and perhaps give regulators license to take draconian measures. As Coinbase CEO Brian Armstrong noted in a response to Elizabeth Warren’s call for more aggressive enforcement, “FTX was an offshore exchange not regulated by the SEC. The problem is that the SEC failed to create regulatory clarity here in the US, so many American investors (and 95% of trading activity) went offshore.”
While all of the ramifications are unclear, FTX’s and Alameda’s bankruptcies could cost users and investors as much as $50 billion. In addition, both FTX and Alameda have various exposures to dozens of companies and protocols, including BlockFi, Solana, Skybridge Capital, Yuga Labs, Voyager, and a host of others on a list here.
Now we need the answers to several important questions: How much exposure to FTX and Alameda did these entities have? How many entities that did deals with the now bankrupt companies were obligated to custody their cryptoassets with––and have their treasuries managed by––FTX? Will they face claw back provisions if courts grant liquidity preferences to those with mismanaged customer deposits?
Amid this uncertainty and gloom, we can find several silver linings. Most important, public blockchain networks like Bitcoin and Ethereum have not skipped a beat during this crisis and continue to operate smoothly: their transparency, openness, and audibility have been crucial to their operation.
Second, time and again, the crypto market punishes centralized entities that lack transparency, which is pushing the ecosystem toward more decentralization and transparency. Exchanges, including Binance, have agreed to adopt “Proof of Reserves”––a cryptographically verified proof that assets match liabilities one-to-one, and many more market participants now understand the value of self-custodying their assets.
ARK’s conviction in the long-term promise of public blockchains across money, finance, and the internet is not wavering. While the crypto asset market could labor under selling pressure and liquidity crunches in the short term, we believe this crisis is purging bad actors and will enhance the health of the crypto ecosystem with more transparency and decentralization in the longer term.
Fourteen years ago, the Genesis block of the Bitcoin blockchain included these words:
The call was to move away from trusted third parties and centralized top-down control toward more open, transparent, and decentralized software.
At this time, let us not forget why we are here.
2. Elon’s Got a Case Of The Twitter Blues
Since Elon Musk took the helm at Twitter roughly two weeks ago, he has been telegraphing his plans for the platform’s future. Here are some of the milestones so far: mass layoffs, an exodus of top executives, advertiser defections, record Twitter engagement and growth, a soft rollout of Twitter Blue, then a rollback of Twitter Blue. In this context, the Silicon Valley adage “move fast and break things” has become an understatement!
In our view, Twitter will morph toward a global super app in stages. The first is putting Twitter on firm financial footing with job cuts and direct monetization subscription revenue streams via Twitter Blue. In the second stage, Twitter is likely to build creator-first features that will boost engagement over the long term. The third stage is likely to incorporate payment features. Having founded the precursor to PayPal, Musk has had a history in digital payments and, taking some cues from China’s WeChat Pay, is likely to add digital wallet-like features.
Success like WeChat’s could catapult Twitter into a sustainable future. Launched as a messaging application, WeChat has scaled to more than one billion users. In 2013, the company rolled out a payments product, WeChat Pay, followed by other financial and commercial products including money-market funds and integrations with ride-hailing platforms, food delivery platforms, and other e-commerce offerings. WeChat Pay’s success was the result of a savvy product and marketing strategy. By 2014, WeChat Pay had engaged only 30 million users, or ~10% of WeChat’s then 300 million user base, before it hit its stride by digitalizing a Chinese New Year tradition: digital “red envelopes” that users could send instantly to one other on WeChatPay. That campaign tripled its user base to 100 million in just two weeks and transformed WeChat Pay into one of China’s digital wallet leaders.
What, we wonder, will be Twitter’s “red envelope” strategy? How will Elon Musk incentivize Twitter users to engage with a payment product and turn it into a payments platform?
3. Could Bad Cholesterol Become History, Even In The United States?
Last year at the JP Morgan Healthcare Conference, Verve Therapeutics (VERV), a gene-editing company, presented interesting pre-clinical data on coronary heart disease (CHD) in non-human primates. In the US, CHD is the leading cause of death, one every 36 seconds.
Verve presented data showing that one treatment of VERVE-101, an in-vivo base editor that inactivates the PCSK9 gene in non-human primates, was effective, durable, and well-tolerated. It reduced the average PCSK9 protein level in the blood by 89% for six months and did not deliver any known off-target effects.
This month, Verve presented updated data at the American Heart Association Scientific Sessions showing that a single infusion of VERVE-101 reduced up to 83% of PCSK9 in the blood of non-human primates for 475 days, or more than 15 months.
Verve’s scientists discovered that, by inactivating the PCSK9 gene, they lowered the level of low-density lipoprotein (LDL), or bad cholesterol, and therefore the probability of heart attacks in otherwise healthy non-human primates.
Potentially, gene-editing will be able to inactivate the PCSK9 gene permanently. Specifically, intravenous base editing, a type of gene editing, inactivates the PCSK9 gene by changing a letter in the “spelling” of one DNA base from adenine (A) to guanine (G). Verve harnessed Beam Therapeutics’ (BEAM) technology to make that change, as shown here.
Last week, the US Food and Drug Administration (FDA) placed VERVE-101, Verve Therapeutics’ lead candidate, on clinical hold. The company disclosed the data in October, and the FDA will provide its response with more information in a full letter within thirty days.
In our opinion, the FDA is a temporary setback in the United States, as Verve continues to recruit patients in New Zealand and the United Kingdom. The encouraging pre-clinical data to date suggest that base editing could become a one-time curative therapy in cardiovascular disease.
4. Swarm Learning Could Displace Federated Learning In Training Neural Nets
Neural networks with enormous potential in a range of applications, from clinical diagnostics to self-driving vehicles, are data hungry. When patient medical data train a model, for example, obstacles like data privacy surface that a process called “federated learning” helps overcome. In federated learning, each hospital system trains a model that a central actor distributes with data collected by other actors––“edge data providers”––in different locations. Once a hospital has trained the model, it returns the weights to the central actor who, in turn, aggregates them to derive a finished model. Federated learning requires the central actor to incentivize all the hospitals and edge data providers to participate.
Nature recently published a new paradigm called “swarm learning” that uses smart contracts on the Ethereum network so that participants can share model weights and combine them automatically without centralized coordination. In our view, swarm learning is a significant improvement over federated learning in models, as it solves the incentive problem with results of equal or higher quality.