OpenAI's latest news is not GPT-5, but "Agent Mode". This reinforces the idea that models are already close to peak, and the next advancement will be in how we use them. We don't need better models, we need cheaper.
People have spent over $3000 on #Shakespere! Wow! You must think we're making a killing. But the truth is we've barely broken even because Claude Sonnet 4 is so fucking expensive. We're just funneling money to Anthropic.