| ▲ | no-name-here 2 days ago | |
The repo would be more helpful if it had, or linked to, any kind of quantified data on how this affects the LLM output quality or task completion. > Why > System prompts are 50-200k tokens. … 1. Are system prompts really the primary purpose for this? 2. As of five months ago, Claude’s prompt was 12-65k tokens depending on enabled tools [1]. Does “50-“ refer to 50K? The repo should link their source for the 50-200K range. | ||