Korva is a scroll of the internet’s best ideas.
Compute fleets are the durable advantage
Brilliant intelligence can migrate from any product to any other product. Codex, OpenAI's coding tool, is winning mostly because it is the best product and the best model, with only a very tiny boost from bundling with ChatGPT. A product advantage like that is fragile: if someone builds something better, they can pull people away from Codex in the same way.
Other advantages hold up better. Network effects still count, and so do economic scale and the ability to build the cheapest compute fleets. Workflows, integrations, complex processes, and the ability for teams to collaborate together are pretty powerful, and so is brand preference and familiarity.
Why do network effects protect a product?
Ethernet co-inventor Robert Metcalfe argued that a network's value grows roughly with the square of its number of users. Each new user makes the product more useful for everyone already there, which makes leaving costly. Telephones and online marketplaces gained their grip this way. Metcalfe's observation dates to around 1980, when he used it to pitch Ethernet.
Why does scale cut the cost of compute?
Chip design and factory costs are fixed, so spreading them over more units lowers the price of each. In 1936 aeronautical engineer Theodore Wright reported that each doubling of cumulative aircraft production cut labor cost per plane by a steady percentage. The relationship, called Wright's law, has since been fitted to products including semiconductors and solar panels.
Why do brands win when products match?
A 2004 study led by Samuel McClure, published in the journal Neuron, scanned people's brains while they sipped Coca-Cola and Pepsi. In blind tastes, preferences split roughly evenly. When the Coca-Cola brand was revealed, more people preferred it, and brain regions linked to memory and emotion became more active.
More from Invest Like the Best with Patrick O'Shaughnessy
