Comment by jey

Comment by jey 3 days ago

4 replies

View on Hacker News

This seems to be incorporated into current LLM generations already -- when code execution is enabled both GPT-5.x and Claude 4.x automatically seem to execute Python code to help with reasoning steps.

fsfod 3 days ago

I remember seeing that GPT-5 had two python tools defined in its leaked prompt one them would hide the output from user visible chain of thought UI.

Reply View 0 replies
Esophagus4 3 days ago

Same with CoT prompting.

If you compare the outputs of a CoT input vs a control input, the outputs will have the reasoning step either way for the current generation of models.

Reply View 0 replies
logicprog 3 days ago

Yeah, this is honestly one of the coolest developments of new models.

Reply View 0 replies