I'm a bit of a madman for this. Claude 3.7 Sonnet is out now with an extended thinking mode, but it still only thinks in plain text.
So... I added a symbolic layer to the reasoning process with a Unicode Thinking Protocol that buckets different styles of thinking into different "fonts" like and and .
There are four buckets:
INITIAL THOUGHTS⠀⇢ (Mathematical Italic)⠀⇢ Example: ℎ ...⠀⇢ First impressions, problem framing, initial approaches
KEY DEDUCTIONS⠀⇢ (Mathematical Bold)⠀⇢ Example: ...⠀⇢ Critical insights, important patterns, main conclusions
UNCERTAINTY⠀⇢ (Mathematical Fraktur)⠀⇢ Example: ...⠀⇢ Expressions of uncertainty, alternatives, limitations
CONCLUSIONS⠀⇢ (Mathematical Script)⠀⇢ Example: ...⠀⇢ Final syntheses and preparation for response
Pros: Makes sequential thinking more explicit by mapping modes to unicode styles. Also, potentially lets the model access different / novel latent space when generating responses because it's not thinking in plain text anymore.
Cons: Token consumption is HIGH, output is slow. Don't use this via API.
I'm a bit of a madman for this. Claude 3.7 Sonnet is out now with an extended thinking mode, but it still only thinks in plain text.
So... I added a symbolic layer to the reasoning process with a Unicode Thinking Protocol that buckets different styles of thinking into different "fonts" like and and .
There are four buckets:
INITIAL THOUGHTS⠀⇢ (Mathematical Italic)⠀⇢ Example: ℎ ...⠀⇢ First impressions, problem framing, initial approaches
KEY DEDUCTIONS⠀⇢ (Mathematical Bold)⠀⇢ Example: ...⠀⇢ Critical insights, important patterns, main conclusions
UNCERTAINTY⠀⇢ (Mathematical Fraktur)⠀⇢ Example: ...⠀⇢ Expressions of uncertainty, alternatives, limitations
CONCLUSIONS⠀⇢ (Mathematical Script)⠀⇢ Example: ...⠀⇢ Final syntheses and preparation for response
Pros: Makes sequential thinking more explicit by mapping modes to unicode styles. Also, potentially lets the model access different / novel latent space when generating responses because it's not thinking in plain text anymore.
Cons: Token consumption is HIGH, output is slow. Don't use this via API.
The Github repo is here: https://github.com/elsheppo/unicode-thinking-claude. This was inspired by "Thinking Claude" and a lot of explorations into symbolic reasoning methods, such as SynthLang.
Initially I wanted to fine-tune an LLM to speak in Unicode, but this proved to be a much quicker way to see the idea in action.
Full deep dive: https://www.shepbryan.com/blog/unicode-thinking-claude-3-7-s...