My hypothesis: AI hyperparameters represents (nearly) all our knowledge. But its a noisy, convoluted and approximate representation. To achieve high quality answer to complex questions, the hyperspace needs to be “mined” as single prompts can’t go deep enough. This mining is called “thinking”, and it goes deeper (i.e. delve in its huge “hyperbowels”). These step costs teraflops and than explains raising token needs in "reasoning" agents.
#Token prices fell 98%. Enterprise #AI bills tripled. Now the industry wants a standards body to explain why. - https://thenextweb.com/news/token-prices-fell-98-enterprise-ai-bills-tripled-now-the-industry-wants-a-standards-body-to-explain-why " The Linux Foundation is launching the Tokenomics Foundation as companies from Uber to Microsoft scramble to understand where their AI budgets went. "