[marginalia_nu]

The critique is against GPT wrappers where the entire product is just a translation layer that makes API requests to e.g. OpenAI is that this isn't much of a moat: Not only can competitors trivially do the same, your LLM API provider can pull the rug from under you by rolling your business model into the next version of their chat product. This latter scenario isn't hypothetical, but has happened several times.

If you build something more substantial that also incorporates GPT calls (e.g. copilot), that's a different thing, and not typically what people refer to when they talk about GPT wrappers.