TEST DRIVING CHATGPT-4O

Summary from the TLDR Web Dev newsletter

This author tested ChatGPT-4o's ability to solve a math problem using five different approaches. They experimented with text-only prompts, multi-modal prompts with images, and prompts with chain-of-thought prompting. The most successful approach was the text-only prompt with chain-of-thought prompting, which resulted in consistently correct answers. The multi-modal prompts led to incorrect answers because the AI misinterpreted the image details.