Dwarkesh Podcast

Austin Vernon - Energy Superabundance, Starship Missiles, & Finding Alpha

Thu Sep 08 2022

Austin Vernon is a writer with a degree in chemical engineering who focuses on engineering, software economics, and investing. He has worked in the field of chemical engineering and taught himself how to write software. Despite being known for his pseudonymous online presence, Austin is more interested in the ideas he writes about than personal branding.
The Starship is a cheaper and more efficient alternative to conventional ballistic weapons like the B-52, especially for long distances. There is a need for weapons that are both accurate and cheap, as seen with Russia's expensive parade-style weapons. Chips are a bottleneck in the development of these weapons.
Kinetic weapons in orbit can hit targets faster and have downstream effects on countries that buy from Russia. Tungsten rods are expensive and have limitations, so many kinetic weapons will have shrouds to slow down in the upper atmosphere before releasing. The power of these weapons can be devastating.
The conversation speculates on the possibility of using shrapnel to attack ships and aircraft, but notes that it would be difficult to execute accurately. Modern warfare technology has not changed much physically, but has improved through better electronics and sensors.
Defensive technology has improved more recently than offensive technology, which is evident in China's anti-access area denial strategy. Taiwan has a strong defense budget and has been investing in modern platforms and weapons, including indigenous fighter bombers and anti-ship missiles. While it is possible for Taiwan to defend itself against China, it would be difficult to win a war against them.
The complexity of digitizing and automating processes is often underestimated, as capturing all the complexity at the bit level is extremely difficult and requires more and more code. This leads to diminishing returns in productivity growth from software. The water bet theory of complexity suggests that there is a limit to how much complexity can be managed effectively.
The Toyota production system emphasizes standardization and breaking down processes into small, manageable parts to reduce defect rates and simplify the complexity of making a car. The system requires someone with a global view to ensure that local optimizations do not negatively impact the entire process. Software allows for specialization and the choice to interface with complexity, such as using machine learning models or abstraction layers like Stripe in financial systems. The conceptual framework of Toyota's production system can be applied to software development.
Processing has been getting cheaper, leading to people running over constraints in software development. Saving development time can lead to different trade-offs. The physical constraints of goods make it harder to fight local optimization, but the same overall constraint exists in software. Vertically integrating complex processes can help overcome these constraints.
The physical world's complexity creates extreme constraints, making it difficult for everyone to have a SaaS company. Blockchain technology has exciting implications for decentralizing software development and reducing transaction costs overall. Big firms may need to be somewhat decentralized with independent business units to stay competitive.
The communication protocols at Amazon are very strict, with specific rules for how to communicate and what type of information to include. This has the same effect as type strict languages in that it allows for better tracking of values throughout the flow of control. The Theory of Firms suggests that companies exist to minimize transaction costs.
Toyota made an agreement with unions to only lay off employees once, resulting in long-term employee tenure and stronger incentives for company success. Lean turnaround requires complete buy-in from line workers for successful standardization of work content. Long employee tenure can lead to knowledge building and better quality products.
Improving one part of a process may not be enough to help a company grow, as all processes need to improve together. Collaboration between engineers, supervisors, and workers is necessary to identify areas for improvement. Consultants and young engineers can bring value to companies by learning about new technologies and bringing fresh perspectives.
Metal manufacturing is a concept that originated from the ideas of W. Edwards Deming, which Toyota implemented but America never fully embraced until recently in the software industry and now in manufacturing with Tesla. Tesla's production technology allows them to produce cars without defects and match demand due to their use of gigacasting. Enterprise value is a better measure of a company's worth than market cap because it takes into account future growth potential.
CO2 electrolysis is a promising method for creating complex molecules like carbon dioxide, formic acid, and methane from water, electricity, CO2, and a catalyst. There is debate in the sector about how much should be done in the electrolyzer versus adding other components downstream. The process of converting CO2 into other compounds involves using an electrolyzer with a flat plate instead of a vessel, which makes scaling down easier.
Formic acid is used as an additive in preserving hay for animals and ethanol can be used to make polyethylene or burned in cars, but it's not a good vehicle fuel. There are multiple paths to producing ethylene and the ability to produce jet fuel through an indirect process. Chemical production has physical limitations and requires innovation to stay ahead of competitors, but lacks defensibility and network effects. The speaker is working on a project related to electrolysis.
The speaker started writing about cryptography and hash functions to better understand white papers. They eventually wrote a blog on IPFS. The speaker faced challenges with decentralized hosting due to high Ethereum fees and difficulty in finding servers through IPNS. They currently use a service called call to host their blog.
Per capita GDP has increased despite per capita energy not increasing since the 1970s, suggesting that economic growth can still occur without increasing energy usage. Diminishing returns exist in terms of energy efficiency, but independent energy sources like solar could increase energy availability for countries that lack access to traditional energy sources.
Cheaper energy and automation of trucking will shift a significant amount of goods from trains to trucks, leading to knock-on effects across the economy. Air freight has an insane elasticity of demand, and if made cheaper, it could be an economic booster with half the operating costs of traditional freight.
Air current energy is expensive, making it hard to imagine turning off air conditioners. Batteries will mainly be used for keeping air conditioners on. Transportation will dominate battery demand, with grid storage being tiny in comparison. Making batteries out of dirt (carbon, iron) could reduce costs and increase sustainability.
Solid electrolytes have the potential to increase energy density and reduce costs in batteries, but they are challenging to improve due to degradation and multiple requirements. Battery storage progress is necessary for intermittent sources of energy like solar and wind, especially for electric vehicles. Electric planes may become more feasible with advancements in battery technology.
Cheap flying and bots delivering goods could reduce inner city car trips, making use of tunnels to avoid taking people's land away. The societal cost of building infrastructure could be significantly reduced by using existing right-of-way and improving property rights. Boring machines for tunneling are becoming more efficient.
Increasing the speed of transportation with VTOLs and other technologies can increase the effective economic size of cities, allowing people to live farther away but still commute easily. Walkable neighborhoods are becoming more popular, and alternative methods of transportation may reduce reliance on cars in cities that lack good public transportation.
Inverters are becoming small and dense enough to be approved by FAA, but it takes time for the approval process. Batteries have limitations and are just on the edge of being good enough for some applications, such as a 50-mile VTOL.
Wilma Castell's new book on long-termism highlights the need for easily accessible coal silos in case of societal collapse, and Charm Industrial is already using fast pyrolysis to create bio oil for carbon credits. Global problems often have unexpected directions, such as the use of bio oil for carbon credits.
Wind turbines face more obstacles than solar due to the difficulty of citing transmission lines, while solar is becoming cheaper and more flexible in its solutions. Nuclear power faces challenges in finding safe and effective reactor designs, with each option having its own drawbacks. The nuclear power industry is complex with opportunities for improvement in regulation and safety standards.
Probabilistic risk assessment has allowed for safer and more consistent operation of nuclear power plants, allowing for repairs to be made while the plant is still running. The safety standards set by the NRC have greatly improved, with damage events now occurring once every 10 years. Existing nuclear power plants ran better, more competitors emerged, and electricity demand slowed down, making it not economically feasible to build new nuclear power plants.
Nuclear power is a complicated and time-consuming project that requires extensive engineering to ensure safety under probabilistic risk assessment. The future of nuclear power lies in creating small reactors that have the potential to impact fewer people's lives and are not easily substitutable, such as micro nuclear reactors. Fukushima highlighted the importance of safety measures in nuclear power plants.
Initial subsidies were necessary for the growth of electric vehicles and solar technology. Direct conversion technologies have a