I actually interpret his post as a positive for Cursor (and LLM coding assistants in general).
If you look at his negative reasons:
1. Poor context window/RAG usage --> Will be drastically improved as models get better at large context windows and RAGs get better. We're also severely limited in terms of hardware. As faster chips come out, we can brute force bigger context.
2. AI Overreach --> can be easily soft through more customization/settings and Cursor team figuring out the right balance.
3. Inconsistent performance --> We're limited by hardware. Mostly the same as #1.
4. Setup is a time sink --> just some more UX improvements needed, nothing that can't be overcome easily
If you look at his negative reasons:
1. Poor context window/RAG usage --> Will be drastically improved as models get better at large context windows and RAGs get better. We're also severely limited in terms of hardware. As faster chips come out, we can brute force bigger context.
2. AI Overreach --> can be easily soft through more customization/settings and Cursor team figuring out the right balance.
3. Inconsistent performance --> We're limited by hardware. Mostly the same as #1.
4. Setup is a time sink --> just some more UX improvements needed, nothing that can't be overcome easily