A person is starting to program. The question is small, but important to them: where do I start without getting lost?

In the better-known storefront, the answer seems to come from an enormous machine, built to talk about almost everything. The person tries to adapt their question to the tool's format. What if there were another door?

Two storefronts side by side: an enormous company with a generalist AI behind a line of people and, beside it, a smaller operation with different robots talking up close with diverse audiences.
Size changes the apparent distance; proximity, however, depends on how the service works.

A smaller company could choose models that already exist and operate inference: make the processing available so that someone can use the chosen model. That is different from building its own application, with an interface and tasks designed for an audience, and also from training a generalist model from scratch.

A catalog of models for inference via API is an offering format documented by Together AI. The question, then, is not whether all AI comes from big tech: it is what space exists for operations and services aimed at more specific needs.

Sources and further reading

The name doesn't settle the question

The difference between the offerings describes the product, but does not by itself define its legal classification. In Brazil, when there is a consumer relationship, the Consumer Defense Code (CDC) provides for clear information about the service and its risks (art. 6, III) and addresses the supplier's liability for defects in the service and inadequate information (art. 14). Brazil’s data protection authority (ANPD) explains that the controller defines the purposes and essential elements of processing; the operator, or processor, acts on its behalf, following its instructions. Calling the offering "inference" does not erase privacy and security questions.

Sources and further reading

Size alone does not guarantee listening or care. A smaller operation also needs to maintain the infrastructure, organize support, and cover its costs. If it wants to keep serving, it needs a margin that allows it to maintain and improve the service — without turning profit into its only measure.

Kuaapy's proposal is still under construction. The idea is to start with audiences and concrete tasks, listen to those who use it, and adjust the service when experience shows a gap. It is a bet on proximity, not on competing with the giants' scale.

Why does the conversation about AI usually seem restricted to giant companies? What specific problem could receive more attention from a smaller team?