Would you trust a lower-cost OpenAI-compatible API layer for a side project?

We’re exploring a pretty specific product shape:

A lower-cost, OpenAI-compatible API layer for indie developers and early AI products.

The idea is not “more features.”
The idea is:

lower cost
minimal migration
familiar workflow

In other words: keep the OpenAI-style integration pattern, reduce friction, and improve economics for smaller teams.

The biggest open question isn’t whether cost matters.

It clearly does.

The bigger question is trust.

If you were building:

a side project
an MVP
a content-heavy AI app
a low-sensitivity workflow

Would you use a third-party OpenAI-compatible layer to reduce API cost?

If not, what would stop you first?

My guess is the top concerns are something like:

reliability
latency
data handling
long-term stability
“why not just stay direct?”

Would love honest answers, especially from people who’ve actually shipped AI products.