#355: Coinbase Launches A Layer 2 Network To Scale Ethereum, & More

Last week, Coinbase announced Base, a Layer 2 (L2) network built on top of Ethereum, offering developers a low-cost, secure, and easy-to-use way to build decentralized applications. Built on the Optimism technology stack, Base is an open-source project that Coinbase hopes will help onboard the next billion users to the cryptoeconomy. Respectful of the current regulatory environment, Coinbase made clear its plans to decentralize the network over time without a native token.

With higher transaction throughput and lower user fees on the Ethereum blockchain, L2 networks ramped up to great success during 2022. Existing L2 networks Arbitrum and Optimism, for example, surpassed the daily transaction count of Ethereum’s base network. Despite that traction, their user counts still are low, with daily active users in the hundreds of thousands, apparently because the experience is not as user friendly as traditional internet services.

At 8.3 million monthly active users and more than 110 million verified accounts, Coinbase is in a good position to scale the number of people using decentralized networks around the world. The company plans to integrate Base into its own product suite, facilitating user access to decentralized apps. Soon Base will house a wide range of applications, from decentralized financial services to gaming, as many companies and projects have announced plans to build on the network.

In our view, Coinbase’s decision to build and integrate its services into a decentralized crypto infrastructure highlights its deep alignment with the fair, transparent, and accessible financial services that public blockchains aim to offer. While it will not derive transaction revenue from Base at launch, Coinbase is likely to benefit financially if its Wallet serves as a trusted on-ramp and access point to applications on the network as it scales.

Last week, we learned that two prominent robotics projects are calling it quits despite artificial intelligence (AI) breakthroughs that are making headlines. Launched to design a robot for the home and office, Alphabet’s Everyday Robots subsidiary has shuttered at the same time that Toyota Research Institute’s home robot is pivoting to more structured environments like supermarkets. Both projects are sunsetting just as AI based on large language models seems to be inflecting toward accelerated growth.

ARK’s research suggests that robots need vast amounts of training data to learn how to move through the world. In our view, successful robotics companies will find an economic way to gain proprietary data advantages.

In a new Nature publication, scientists at the Howard Hughes Medical Institute describe protein structures that they designed computationally from scratch. Combining deep learning (DL) techniques and traditional molecular modeling, the group designed an enzyme that produces stronger than natural bioluminescence—the ability for a molecule to glow. Bioluminescent enzymes are ubiquitous in biological research because they can be used to help scientists visually determine if a desired reaction did indeed occur. The method, which the scientists call “Family-wide hallucination,” generated an enzyme more thermally stable and compact than any in nature.

Family-wide hallucination could be a significant milestone in protein design, potentially enabling breakthroughs in agriculture, biomedicine, and other fields. Scientists began with a protein structure unrelated to the proteins they intended to imitate. Asking DL models like Deepmind’s Alphafold to generate protein structures slightly different from their starting points and then evaluating the performance of each simulated protein, they recreated natural evolution in silico.