When you ask two AI models the same basic math question and get completely opposite answers. Deepseek correctly identifies that 9.9 > 9.11 (treating them as decimals), while ChatGPT somehow thinks 9.11 > 9.9. This is why we still have jobs. For now.
Nothing says "trust me with your critical systems" like failing elementary school math. Somewhere, a software engineer is using this screenshot in their slide deck titled "Why Human QA Still Matters".