Benn Stancil's post is useful because it separates model creation from durable business value.
The batch framing is sharp: foundation models can look like expensive films, costly to make but hard to defend once the market moves on. If raw capability advantage decays quickly, labs need other layers to capture value: applications, distribution, customization, user data, and switching costs.
For New Runtime, this matters because it explains why model labs keep moving into products and workflows. The model is the engine, but the economic moat may be the place where the model becomes embedded in repeated work.