Software scales with zero marginal costs, but machine learning scales with nonlinear emergent behaviors.

The most important exercise right now is figuring out ML moats is tracking the interface between scaling laws (of emergent behavior in high-quality data) and products.

A truly great business must have an enduring "moat" that protects excellent returns on invested capital.

The problem is, models can easily be replaced with fine-tuning or some other procedural change. The model is the part of the system users will interface with the most, but the dataset, infrastructure, and processes are what will create structural advantages.

Data is the moat for ML systems right now. When training data is well defined and curated over time, it can’t be taken by one leaving employee or a simple leak.

User data is a sure way to get the most diverse data, which is crucial because the data needs to be diverse and not repeated when scaling.

If adding new data leads to new abilities and extremely concentrated usage, then data can provide lasting advantages in a way not seen before.

Companies like Runway and Jasper are crafting moats in verticals, where they are the best-in-class companies and brand names

Lensa is built on Stable Diffusion, so it may not have a moat at all. It won by being first