This video introduces five secret codes or commands designed to enhance the functionality and output of Claude, an AI assistant. The presenter explains each code with its specific benefit and application.

1.  **/caveman**: This code, when used with an open-source add-on, aims to reduce the number of response tokens by 75%, suggesting a more concise output.
2.  **ultraink**: Typing this command anywhere in the prompt is said to make Claude prompt ten times deeper and more efficiently, implying more thorough and effective responses.
3.  **/ghost**: This command is designed to remove AI-specific linguistic markers from Claude's output, such as introductory phrases ('No, here's what I think') or concluding remarks ('No conclusion', 'no preamble'), leading to faster and more direct responses.
4.  **/deepthink**: This code forces Claude to engage in a more rigorous, layered reasoning process before providing an answer. While it may slow down the output, it significantly improves the quality of responses for complex questions.
5.  **by the way**: This command allows Claude to respond to a new prompt without interrupting its current task, enabling multitasking or sequential processing of requests.

The video concludes with a call to action for viewers to follow the presenter and comment 'secret' to receive more such codes.

Nick Sadler

Here are five secret codes for Claude.

Number one is /caveman. Install this open-source add-on and you'll reduce 75% of response tokens immediately.

Number two is ultraink. Just type this anywhere and Claude will prompt 10 times deeper and more efficiently.

Number three is /ghost. This strips away the AI tells from your output. No, here's what I think. No inconclusion, no preamble. You just get ship faster.

Number four is /deepthink. This forces Claude to reason through every layer before answering. Slower output but dramatically better quality on complex questions.

And number five is by the way. Use this when Claude's already working on something and it'll respond without stopping its current job.

If you want more of these codes for your Claude, make sure to follow me and comment secret and I'll shoot them over to you.

Here are five secret codes for Claude that most users don't know.

Here are five secret codes for Claude.
Number one is /caveman. Install this
open- source add-on and you'll reduce
75% of response tokens immediately.
Number two is ultraink. Just type this
anywhere and Claude will prompt 10 times
deeper and more efficiently. Number
three is / ghost. This strips away the
AI tells from your output. No, here's
what I think. No inconclusion, no
preamble. You just get ship faster.
Number four is /deepthink. This forces
Claude to reason through every layer
before answering. Slower output but
dramatically better quality on complex
questions. And number five is by the
way. Use this when Claude's already
working on something and it'll respond
without stopping its current job. If you
want more of these codes for your
Claude, make sure to follow me and
comment secret and I'll shoot them over
to

Get Our Mobile App

Here are five secret codes for Claude that most users don't know.

Transcription