#409: Regulation Can Kill Innovation, As Detailed In ARK’s New Research On The Nuclear Industry, & More
1. Regulation Can Kill Innovation, As Detailed In ARK’s New Research On The Nuclear Industry

Last week, ARK Research Associate Daniel Maguire published a paper detailing1 how regulation has stifled the development of nuclear energy. Before anti-nuclear sentiment and more stringent regulations led to massive cost increases, nuclear facilities in the US and Japan had been following a learning curve2 characterized by a ~20% decline in overnight construction costs (OCC) for every cumulative doubling in megawatts of power deployed,3 as shown below. Informed by Wright’s Law, we believe nuclear energy would have cost less than natural gas today had stringent regulations not caused ~$345 billion in overruns on the construction and development of US nuclear plants.
Source: ARK Investment Management LLC, 2024, based on data from Lovering et al. 2016; IAEA 2023; EIA 2022; and World Nuclear News 2022 as of October 2023.4 For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security. Past performance is not indicative of future results.
2. Open-Source AI Agents Are Moving Full Steam Ahead

Less than a month after releasing Devin, an AI coding agent that apparently solves5 software bugs better than any other to date, Cognition Labs is facing competition. Researchers at Princeton, who created the original SWE-bench benchmark, have released an open-source coding agent called SWE-agent.
On SWE-bench, a collection of real-world software bugs reported on GitHub, SWE-agent scored 12.29%—nearly as high as Devin’s 13.84%. Notably, Devin’s performance was evaluated on a subset of cases in the dataset, while SWE-agent's relatively impressive score was based on the entire dataset. Based on the entire dataset, the third best model—scoring only 3.79%—was Anthropic’s Claude 3 Opus, as shown below.
Source: “Princeton-nlp/SWE-agent—Overview.” GitHub. Data as of April 6, 2024. For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security. Past performance is not indicative of future results.
Importantly, SWE-agent is not a model per se, but a wrapper around any large language model (LLM) that provides access to tools like code editors and command lines and can iterate with chain-of-thought prompting to solve complex problems.
That the performance of agentic AI frameworks like SWE-agent surpassing out-of-the box models like GPT-4 and Claude 3 Opus demonstrates how rapidly the open-source community is creating and distributing meaningful and novel approaches to solving problems. It also highlights that existing generation models have significant room to improve. The race is on.
3. EV Battery Costs Have Declined Significantly

For years, high battery costs have been the main barrier to manufacturing affordable long-range electric vehicles (EVs). Prior to 2015, battery costs were so prohibitive that no EV in the US offered more than 300 miles of range. In 2016, the price of the first EV to exceed that range was $100,000.
In the last eight years, technology learning curves have cut the price in half to $50,000. ARK’s research suggests that manufacturers should be able to scale profitably into mass market production of EVs with ample range during the next few years, thanks to improvements in battery cost and vehicle efficiency, as shown below.
*Model Year 2016-2023. Sources: ARK Investment Management LLC, 2023, based on data from Kane 2022.6 Forecasts are inherently limited and cannot be relied upon. For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security. Past performance is not indicative of future results.
The ability to manufacture for the mass market profitably should separate winners from losers in the race to dominate the battery electric vehicle (BEV) market. Based on concerns about range anxiety and perhaps politics, some companies are shifting their production focus back toward hybrid-electric vehicles. In our view, companies prioritizing plug-in hybrid electric vehicles (PHEVs) will be thrown off the learning curve for BEVs. We believe Tesla is shaping the future of affordable autonomous BEVs, while BYD is focusing on affordable EVs.
4. BitVM Could Make Bitcoin More Programmable And Scalable

The Bitcoin community is delighted by BitVM,7 a new technology that could enable Ethereum-like smart contracts on Bitcoin by allowing any computation to be verified on the Bitcoin blockchain, thus offering Turing completeness through Bitcoin's raw scripting language.
Criticism of the Lightning Network—Bitcoin's original layer 2 scaling solution—has mounted over the years because of its slow development and limited capabilities. As a result, since early 2022 the number of active channels on the Lightning Network has declined and, since inception, its capacity has never surpassed 300M USD, as shown below.
Source: Bitcoin Visuals 2024.8 For informational purposes only and should not be considered investment advice or a recommendation to buy, sell, or hold any particular security or cryptocurrency. Past performance is not indicative of future results.
BitVM’s fresh approach to scaling is sparking enthusiasm among users, developers, and investors. In the six months since Robin Linus first introduced BitVM, a new wave of developers, researchers, and venture capital firms has shown keen interest in exploring its potential. Numerous independent projects have surfaced, aiming to leverage BitVM with their own layer 2 solutions on top of Bitcoin. Disillusioned for the last few years with Lightning network-related companies and focused increasingly on Ethereum-based infrastructure and applications, investors now are supporting and funding new BitVM-related ventures.
In early days of promising Bitcoin scaling solutions and smart contracts, BitVM has ignited a spark that we hope will inspire more entrepreneurs to address the most challenging technical problems with the first and most renowned blockchain.
1 Maguire, D. 2024. “Nuclear Energy Could Contribute Meaningfully To Securing Net-Zero Emissions” ARK Investment Management LLC.
2 As theorized by Wright’s Law. See Winton, B. 2019. “Moore’s Law Isn’t Dead: It’s Wrong—Long Live Wright’s Law” ARK Investment Management LLC.
3 “Overnight Construction Cost” refers to the cost of a construction project if no interest was incurred during construction—in other words, as though the project had been completed overnight.
4 Lovering, J. et al. 2016. “Historical construction costs of global nuclear power reactors” Energy Policy. International Energy Agency. 2022. “Cost and Performance Characteristics of New Generating Technologies, Annual Energy Outlook 2022” World Nuclear News. 2022. “AP1000 remains attractive option for US market, says MIT.”
5 Soja, J. 2024. “Cognition Labs’ Devin AI Agent Outperforms LLMs On Complex, Real-World Engineering Tasks” ARK Disrupt Newsletter.
6 ARK Investment Management LLC. 2023. “Big Ideas 2023” Kane, M. 2022. "US: The Number Of EV Models With 300+ Miles Of Range Tripled in 2022" InsideEVs.
7 Robin Linus discussed the breakthrough in a whitepaper published in October 2023. See Linus, R. 2023. “BitVM: Compute Anything on Bitcoin.” bitvm.org.
8 Bitcoin Visuals. 2024. “Lightning Network Channels.”