Guides & Explainers
In-depth written guides on AI concepts, from beginner to advanced
Open-Weight AI Models Explained: Open Weights vs Open Source
What open-weight AI models are, how open weights differ from open source, the licences they use, which models publish their weights, and the hardware they need.
Diffusion Models Explained: How AI Turns Noise Into Images
How diffusion models work: from noise to image step by step, text prompts and guidance, latent diffusion, flow matching, and how Stable Diffusion and Sora use them.
Mixture of Experts (MoE) Explained: How Sparse AI Models Work
What a mixture-of-experts (MoE) model is: experts, the router, total vs active parameters, where MoE came from, and the trade-offs, with real model examples.
How to Compare AI Models: Elo, Benchmarks, Price and Speed
How to compare AI model performance: what LMArena Elo, benchmarks, price per token, context window and speed each measure, with a live side-by-side example.
Gemini 3.8 Live Explained: Google's Real-Time Voice AI
What Gemini 3.8 Live and Live Extended Thinking are, how live voice AI works, where to use them, 97 languages, and the price per minute of audio.
Gemini 4 Argon Explained: Availability, Price and Specs
Gemini 4 Argon is generally available since 13 October 2026. Its price ($2/$10 per 1M tokens, then $4/$20), 1M context window, where to use it, and FAQs.
AI Scaling Laws Explained: Kaplan (2020) and Chinchilla (2022)
What AI scaling laws are: how size, data and compute predict model quality, the 2020 Kaplan paper, the Chinchilla 20-tokens-per-parameter rule, and 70B vs 280B.
Attention Is All You Need Explained: The Paper Behind the Transformer
The 2017 transformer paper in plain words: self-attention, queries, keys and values, multi-head attention, the base model, its results, and how it led to ChatGPT.
What Is a Token? Tokens and Context Windows Explained
What an AI token is, how many words fit in one, how images and audio are counted, what a context window limits, and how per-million-token pricing works.
What Is a Model Card? How to Read an AI Model Card
A model card is the document that describes an AI model. What one includes, where to find them, and how to read the context window, cutoff, licence and benchmark fields.
Few-Shot Prompting Explained: Zero-Shot, One-Shot and In-Context Learning
Few-shot prompting explained with examples: zero-, one- and few-shot as defined in the GPT-3 paper, in-context learning, and what research found about examples.
Chain-of-Thought Prompting Explained: Step-by-Step AI Reasoning
What chain-of-thought prompting is, with examples: the 2022 Wei paper, zero-shot "Let's think step by step" (17.7% to 78.7%), self-consistency and thinking modes.
Showing 1–12 of 45 guides