unicodesnwrites | Learn AI, explained simply

LESSONS · 14 MIN READAn Image Is Worth 16×16 Words: How AI Actually Sees Your Photos (and What Each One Costs)A transformer only reads tokens, so how does it read your screenshot? We cut a cat photo into 196 patches, rebuild the Vision Transformer's first step in Python, and work out what a 4K screenshot really costs, and why its small text can come back blurry.

RESEARCH EXPLAINED · SERIES · 2 PARTSFrom Shannon to Transformers: The Foundations of Modern AIBefore transformers, there was a long journey of ideas that shaped how machines learn, process information, and understand language. This series explores the foundational papers and breakthroughs behind modern AI, from Claude Shannon's experiments with English letters, bigrams, and information theory to statistical language models, neural networks, and the ideas that eventually led to transformers. We'll work through the original papers step by step, unpack the mathematics intuitively, and reconstruct the key ideas through examples and experiments. The goal isn't just to understand what these papers say, but to discover why these ideas mattered and how they connect to the AI systems we build today.

RESEARCH EXPLAINED · SERIES · 2 PARTSMCP SERVER: Theory and ImplementationYou've seen MCP servers mentioned in AI tool docs. You've wondered how Claude actually calls external functions. This is the guide that explains both — and by the end of this series you'll have a running, production-ready MCP server connected to Claude Desktop, built in Python.

PROJECTS · SERIES · 3 PARTS Build a Modern Flower Classifier App with Flutter, Camera, and TensorFlow LiteAn end-to-end edge device ML classification model

Get one helpful email when there's something new.

Learn how AI really works, explained simply.

Courses and guides

Start with these