The main AI model providers are in an interesting strategic position.

Their competitive differentiation is they're extremely good at producing models.

All of them were research orgs first that kind of laid on "sell API access" later… and now find themselves in an unexpected consumer-facing role, too.

They started out with a simple demo of their product (the API) in a consumer-facing application.

But the model is so good, that even a barebones demo is quite useful!

As a user, why go to another tool that wraps the model when the UX is the same?

Just go straight to the source, to the default UX.

A small advantage, but a consistent one.

They start attracting a non-trivial amount of consumer usage.

They've kind of stumbled backwards into a situation where they could plausibly create the next generation consumer aggregator… or the next generation B2B aggregator (e.g. Salesforce)

However their sustainable differentiation is not UX or building an ecosystem, but building a model.

And the main labs are acting like they believe there are many orders of magnitude more quality to squeeze out of this model architecture.

If the model will get radically better, why waste time on little details and minor side quests that distract from the main job?

The result is that these companies have an insanely strong position for traditional aggregator style approaches that any other company would kill for… but don't necessarily have the time or inclination to execute on traditional strategies in that position.

For example, as Simon notes, Code Interpreter is insanely powerful, and could be sold today to 100k analysts at a high price… and OpenAI cares about it so little that there's not even a landing page describing what it is that enthusiastic users can evangelize.

The providers all have a consumer UX they host themselves but also an API that allows others to build different experiences powered by the model.

If someone could use their model and build a UX that is significantly better, or harnesses the power of an ecosystem to create a system that is the power of the model combined with an ecosystem, then it could break the proto aggregator advantage of the main labs.

Other entities can use the same basic model and then layer on a more differentiated / powerful UX, improving upon the model.

The importance of an innovative UX or an ecosystem for these labs pales in comparison to the importance of building the next 10x model.