• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    GitHub Post Shares Claude Fable 5.1 Prompt

    A GitHub repository holds the full Claude Fable 5.1 system prompt text.

    BR
    2 Sources, 29d ago, first seen 29d ago

    TLDR

    Brian Roemmele posted about the Anthropic Claude Fable 5.1 system prompt. He stated that the text runs 270,000 characters and that users pay for it with each prompt. The text appears in a repository maintained by elder-plinius that collects prompts from multiple AI systems. Anthropic maintains a page that references the Claude Fable 5.1 system prompt and its updates on claude.ai and the mobile apps.

    Combined views

    56.7K

    2 Sources, first seen 29d ago

    Combined views

    56.7K

    2 Sources, first seen 29d ago

    450 likes
    450 likes
    36 comments
    392 saves
    67 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    36 comments
    392 saves
    67 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    2 Sources

    @BrianRoemmeleThe Antropic Fable AI System Prompt: Hacked! It’s 270,000 characters AND YOU PAY FOR IT EACH TIME YOU PROMPT! Claude Fable 5.1’s system prompt is a record of what Anthropic thinks goes wrong when a capable model is left in a long conversation with tools, memory, and a drawing canvas. And of course, this is essentially a list of everything under the sun ultimatly and it will become impossible for the AI to navigate the simple to the most complex requests with this burden. You have the ability to have a better understanding than the supposed experts at Antropic, you know that this is already broken because a one sentence prompt is first sending over 10s of 1000s of words to be "safe". Yes, it is ridiculous. Publishing a 270,000-character safety essay is not the same as having a safety system. It is evidence that the system is being managed in prose because it is not reliable in behavior. It is my postulate, over 2 decades ago, that constitutional documents like the one used by Anthropic are childish at best and, at worst, crippling to the model on a holistic level. They will grow to be millions of characters and plug up the AI model like a kitchen sink after a big Thanksgiving meal. And the insult? You pay for the document that works no better than a broken clock being right twice a day. Read more below:

    2 Sources

    @BrianRoemmeleThe Antropic Fable AI System Prompt: Hacked! It’s 270,000 characters AND YOU PAY FOR IT EACH TIME YOU PROMPT! Claude Fable 5.1’s system prompt is a record of what Anthropic thinks goes wrong when a capable model is left in a long conversation with tools, memory, and a drawing canvas. And of course, this is essentially a list of everything under the sun ultimatly and it will become impossible for the AI to navigate the simple to the most complex requests with this burden. You have the ability to have a better understanding than the supposed experts at Antropic, you know that this is already broken because a one sentence prompt is first sending over 10s of 1000s of words to be "safe". Yes, it is ridiculous. Publishing a 270,000-character safety essay is not the same as having a safety system. It is evidence that the system is being managed in prose because it is not reliable in behavior. It is my postulate, over 2 decades ago, that constitutional documents like the one used by Anthropic are childish at best and, at worst, crippling to the model on a holistic level. They will grow to be millions of characters and plug up the AI model like a kitchen sink after a big Thanksgiving meal. And the insult? You pay for the document that works no better than a broken clock being right twice a day. Read more below: