Learning path
How AI Models Are Built
How ChatGPT, Claude, Gemini and Grok are made: training data, tokens, pretraining, fine-tuning, RLHF and serving, explained for beginners with animated films and runnable Python.
-
1
From Data to ChatGPT: The Life of an AI Model
How are ChatGPT, Claude, Gemini and Grok actually made? A beginner-friendly tour of the six stages every large language model goes through, from raw text to the chat…
-
2
Training Data: Where AI Models Learn From (and How It Is Cleaned)
What text are ChatGPT, Claude and Gemini trained on, why most of it is thrown away, and how engineers remove duplicates, spam and personal data. With runnable Python.
-
3
Pretraining: How a Model Learns by Guessing the Next Word
Loss, gradient descent, parameters and GPUs explained simply. Train a one-number model in Python and see exactly how ChatGPT-style models learn during pretraining.
-
4
Fine-Tuning, RLHF and Safety: How a Model Becomes an Assistant
Supervised fine-tuning, reward models, RLHF, Constitutional AI and DPO explained simply, with runnable Python. Learn why ChatGPT, Claude and Gemini behave so differently from a raw model.
-
5
Serving AI to Millions: Inference, GPUs, Speed and Pricing
What happens between pressing Enter and seeing ChatGPT's answer? Inference, batching, the KV cache, quantization, and why AI APIs charge per token, with runnable Python.
-
6
ChatGPT vs Claude vs Gemini vs Grok vs Llama: AI Model Families Explained
Who makes which AI model, open vs closed weights, chat apps vs APIs, and a simple checklist for choosing the right model for studying, coding or building an…