How can I set the C++ language version? Cursor seems to not know about C++23 features out of the box.
Ahem. No reply. Is there support other than through the forums? I have a paid account so I’d expect some sort of help.
You mean these?
This is technically not an issue with Cursor itself. You may apply the solution from this post:
You may add and reference as many external docs here then use @ in the Agent chat to import them. Are 233 pages enough? It’s just the first search result I found.
I don’t understand the verb “train” here. The language standard is not something to train: it’s something you set. As in “c++17 is now legal”. That should be a setting for the parser.
I opened chat and it told me “If you have access to C++23 and want to keep using println, you’ll need to ensure your compiler supports it and compile with the appropriate C++23 flags (e.g., -std=c++23).”
Sure, when I compile it I’ll set the standard. But meanwhile Cursor is under-tilde’ing it. That comes before the compiler. How do I tell it to only mark stuff that is illegal according to C++23?
The verb trained refers to the LLM model training which is based on an input dataset and provides a set of parameters. The dataset is tens of trillions of bytes and the parameters is a few hundred billions of bytes. This compression effort is the training process and is very expensive so it’s done very far apart in terms of months. This is something the cloud company other than Cursor is responsible for.
After training, test time compute, or inference, is what we are doing in Cursor. We can use the prompt to activate some parameters to fit our purpose better. For example adding the latest documentation in the prompt itself, and of course whatever rules you like.
But the knowledge in the base model is fixed, you can only add new information, every time you start a prompt, for it to understand how the world changed. This is something you are responsible for.
Here is what happens when you don’t provide updated information:
You tell it every time you need to, or create a new Rule