

22·
11 months agoThey train grok like other llm Models, but they use system prompts to manipulate answers. That way they can tweak it quickly without retraining.


They train grok like other llm Models, but they use system prompts to manipulate answers. That way they can tweak it quickly without retraining.
You can also edit the search url to default to the web tab, which only shows link results