Model provider harnesses might make the undifferentiated LLM API market more sticky. Eg we used to switch models ever time there’s a new one, but now with the C

Model provider harnesses might make the undifferentiated LLM API market more sticky. Eg we used to switch models ever time there’s a new one, but now with the Claude SDK there’s no easy way to switch to OpenAI anymore..