Simon Willison details the subtle yet impactful changes Anthropic made to Claude Opus's system prompt between versions 4.6 and 4.7. This analysis offers a rare glimpse into how LLM providers fine-tune model behavior.
Sebastian Raschka shares his personal, learning-oriented workflow for effectively dissecting and understanding the architectures of new open-weight Large Language Models. This guide provides a systematic approach for anyone looking to deepen their comprehension of cutting-edge AI models.
Simon Willison shares a clever method to extract and track Claude's system prompts over time using a Python script and Git. This creates a transparent timeline of how LLM instructions evolve.
