🎙️ EP 122: GPT-5's Big Math Fail + The 1 Sentence That Killed Prompt Engineering

OpenAI said GPT-5 solved 10 legendary math problems… turns out, it just found old papers. Meanwhile, Stanford researchers dropped a dead-simple trick that breaks the limits of creativity in AI models. No fine-tuning, no hacks, just 1 sentence.

We’ll talk about:

• The brutal backlash over GPT‑5’s “math breakthrough” (and what actually happened)
• Why your AI keeps repeating the same answers—and how Stanford fixed it
• A powerful trick called Verbalized Sampling that unlocks 2× more creativity
• How AI startups are raising rents in San Francisco… and leasing luxury apartments for staff

Keywords: GPT‑5, OpenAI, Verbalized Sampling, Stanford AI, Claude, Gemini, mode collapse, prompt engineering, San Francisco rent, AI job tips, Veo 3.1, AI tools