What if you could test your headlines on Hacker News before they ever go live? @hammer_mt built exactly that—an AI simulator that predicts what content will go viral. In his latest piece for @every, Michael reveals how he created a synthetic crowd of 81 AI personas based on real @hackernews users' comments. He explains: - How AI agents can predict which headlines will perform best - Why getting harsh feedback from AI is more valuable than politeness - How different AI models (GPT, Claude, Gemini, Llama) compare when roleplaying - Why Google's Gemini surprisingly outperformed other models - The technique for getting AI to stop being nice and start giving brutal feedback Read more in the full article linked below.