From 82e852dd0b15a47f538f21b0b330ffa1b8f0c570 Mon Sep 17 00:00:00 2001 From: Greg Gauthier Date: Fri, 10 Apr 2026 22:24:37 +0100 Subject: [PATCH] docs(readme): add purpose and tested models sections Enhance README with a purpose statement explaining the repository's goals, including insights on AI behaviors, biases, and ethics. Add a section on tested models, detailing the specific Grok, Claude, and ChatGPT versions used for comparisons. --- README.md | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+) diff --git a/README.md b/README.md index ec02706..65f2708 100644 --- a/README.md +++ b/README.md @@ -1,3 +1,27 @@ # Grok Versus Claude A record of ongoing debates between a Grok agent and a Claude agent. + +## Purpose + +This repository serves as a platform for documenting and analyzing the interactions between a Grok agent and a Claude agent. It aims to provide insights into the capabilities, limitations, and potential applications of these AI systems in various domains. + +Though the debates themselves are content rich, these LLMs are essentially stochastic mimics, and the results are not interesting for their own sake. Rather, they function as demonstrations of the ways in which token prediction is affected by the model's training data, and other factors. + +Things that we're especially interested in, are: +- Observable behavioural patterns in each vendor's premium model +- Biases and distortions in model responses +- Performance metrics and evaluation criteria for AI systems +- Ethical considerations and implications of AI technology + +## Tested Models + +Because I have limited resources, I have only been able to subscribe to two vendors: Grok and Claude. In both cases, I am using the best available single-agent model for each vendor. + +- Grok: Grok 4.20 Reasoning (2M tokens context) +- Claude: Opus 4.6 Extended (1M tokens context) +- ChatGPT: GPT-4 (Free Web Tier as a "control" model) + +While these models are not perfectly equivalent, they still provide a useful starting point for understanding the differences between the two. + +