All Battles

ChatGPT vs Gemini: weird tests that expose real differences

Instead of benchmarks, I ran both models through a set of internet-famous “trap tests” — the kind designed to expose blind spots, not show best-case performance. These include: - the 7-finger hand test - the “Swiss cheese font” test - the car wash logic test All prompts were taken or adapted from real discussions and used without optimization.

Claude Code vs Cursor: Same prompt, two different results — which is better?

In this battle, I asked both Claude Code and Cursor to build me a landing page from a Reddit prompt. Copy-pasted the prompt as is to both tools and hit send. In the comparison bellow I highlight the differences and results I've got from the same prompt.