Why ChatGPT, Claude, and Gemini Can't Predict Stock Prices: An LLM Internals Primer

Short summary

This article uses an uncle-nephew dialogue to debunk claims that ChatGPT, Claude, or Gemini can predict stock prices with 90% accuracy. It explains from first principles how LLMs work — tokenization, embeddings, attention, and parameters — to show why next-word predictors are fundamentally unsuited for numerical stock-market forecasting. The piece is an accessible primer on LLM internals wrapped in a cautionary tale about AI scam ads targeting retail investors.

• •LLMs are next-word predictors, not calculators or stock analysts — wrong tool for numerical forecasting
• •Walks through tokenization, embeddings, attention, and parameters in plain language for beginners
• •Debunks YouTube scam ads selling 'secret prompts' for 90% accuracy stock predictions

Generated with AI, which can make mistakes.