#332: ARK Publishes A New Study On The Therapeutic Potential Of Psychedelics, & More

As the fields of psychedelic neuroscience and pharmacology break through decades of cultural stigma, pharmaceutical companies seem eager to push psychedelic formulations into clinical trials.

Investors exploring the opportunities in this innovative space could be confused by the debate around psychedelics’ true potential. A recent article published in the Journal of the American Medical Association (JAMA), for example, highlighted that Psilocybin – the active ingredient in “magic mushrooms” has the potential to curtail alcohol abuse. On the heels of that article, JAMA Psychiatry featured a comment by Roland Griffiths, a prominent researcher in psychedelic neuroscience, warning that the hype surrounding psychedelics could exaggerate their utility.

Last week, ARK published a new study ––How “Far Out” are Psychedelic Therapeutics?––that delves into the potential merits of psychedelics as therapeutic agents. We offer what we believe is an objective analysis to help investors assess the potential, distinguishing between psychedelics that might improve mental health from those that will not.

This summer, OpenAI’s DALL·E 2 delighted the world with its ability to generate creative images from text prompts. In the last few months DALL·E 2 has passed some important milestones: commercial availability, pricing tiers, and more restrictive content mediation.

During the same time, other research groups have developed similar text-to-image models. In our view, Stability.AI has debuted the most provocative model, called Stable Diffusion. Indeed, its image-generation model seems superior to DALL·E 2 in certain domains, particularly face generation.

Among the important differences between the two models, Stable Diffusion is open source and places few constraints on generation requests. Users can self-host it on their desktops, for example, and can generate any images they desire. In contrast, OpenAI curbs the generation of offensive and other controversial images.

The proliferation of open-source DALL·E 2-like models raises an important question for investors: will these models be commoditized? We believe large models trained on publicly available data will be commoditized, while models trained on proprietary data will remain differentiated and difficult to duplicate. Tesla, for example, collects massive amounts of data from millions of vehicles, including Autopilot interventions, also known as “corner cases”. Almost impossible to replicate, those data will be critical to the success of its autonomous vehicle program. In our view, the most successful models will combine propriety data assets with publicly available data.

On August 24, Coinbase launched a liquid staking product, cbETH. Like Lido’s liquid staked ether, stETH, cbETH creates a tokenized deposit that offers redemption rights to the ether staked on its platform. As the name suggests, liquid staking will offer more liquidity to assets staked in Ethereum’s Proof-of-Stake network by allowing users to transfer, trade, or use the token in decentralized finance (DeFi) without disturbing the underlying assets.

As the first major exchange to offer a liquid staking product, Coinbase is attempting to gain share in a market characterized by highly profitable, less volatile, recurring revenue that could be an important diversifier to its base trading business. To date, investors have wrapped more than 650,000 ether of the estimated 2 million staked ether on Coinbase. In other words, cbETH’s outstanding supply is worth ~$1 billion today.

Coinbase’s liquid staking product makes several distinctive design decisions compared to Lido. While Lido’s stETH is based on the “aToken” model, which updates holder balances daily to reflect interest accrued, cbETH uses the “cToken” model, which reports holder balances based on the original deposit while the redemption value of their tokens increases over time. In practice, it appears that the latter model is less intuitive for users, as 1 cbETH does not equal 1 ETH over time. That said, the cToken model does reduce the complexity associated with integrating tokens into downstream DeFI, which could accelerate the adoption of cbETH.

The two models also may differ in their tax treatment. While we are not offering any advice or opinion on the matter, some crypto tax professionals suggest that cToken appreciation could be treated as capital gains, while aToken interest is ordinary income. Coinbase takes a 25% cut on cbETH staking income, while Lido charges just 10% on stETH staking income.

Perhaps the most important and controversial difference between the two services is that Lido distributes its validator pool across a global network of 29 node operators and plans to decentralize its operations even more over time. In contrast, Coinbase will be the sole issuer and controller of cbETH.

This difference creates an interesting decision point for end users. Will they choose the stability and convenience that Coinbase offers, or the decentralized and censorship-resistant approach that Lido has created?

Importantly, because Coinbase will be able to freeze cbETH assets when required by law, will its proliferation increase crypto centralization and the risk of censorship? After OFAC’s sanctioning of Tornado Cash in August, this hot topic is getting hotter by the minute.