5/5 From Story to Stage: Creating Presentations with Marp
The fairy tale is written. The diagrams are drawn. Now turn it all into a presentation -- without opening PowerPoint, without leaving VSCode, without losing your Git history.
Archives: 2025
Found 14 posts
December 2025 (8 post s)
4/5 AGENTS.md: Teaching AI How You Work
README is for humans. AGENTS.md is for AI. 40,000+ projects use it to give AI persistent instructions that survive across sessions. Here's how to write one.
3/5 GitHub Copilot Beyond Code: Your AI Writing Partner
20 million developers use Copilot for code. Almost nobody uses it for content. Here's how to turn a fairy tale outline into a full story with diagrams -- without writing a single line of code.
2/5 VSCode Isn't a Code Editor. It's an AI Workspace.
100 million developers use VSCode. Most of them only use it for code. Here's how to set it up as an AI content creation workspace.
1/5 Markdown Is the Interface Between You and AI
Markdown isn't a formatting tool. It's a communication protocol between you and every AI model on the planet. Here's why that matters.
Markdown Essentials: Write Once, Render Everywhere
Markdown is the universal language of structured text. Learn the essentials without the bloat.
Marp Presentations: Markdown Becomes Slides
Presentations as code. Write Markdown, get slides. No PowerPoint, no Keynote, no drag-and-drop.
Mermaid Diagrams: When Text Becomes Visual
Diagrams as code. Version-controlled, LLM-friendly, and no drag-and-drop tools required.
August 2025 (5 post s)
5/5 Your First Model: Navigating a Hugging Face Repo
A model is a folder. config.json is the architecture. generation_config.json is the sampling defaults. vocab.json is the tokenizer. Here's how to read them — and where to change the knobs.
4/5 Be the Language DJ: Temperature, Top-k, and Top-p
The model produces 50,257 scores. Sampling decides which one becomes the next token. Temperature, top-k, and top-p are your mixer sliders — here's what each one does, with numbers.
3/5 The Transformer in 90 Seconds (Then the Other 900)
Attention is a weighted mix. Multi-head is a filter bank. The causal mask means no spoilers. Here's the transformer architecture without the math — then with just enough of it.
2/5 Why Can't LLMs Spell 'Raspberry'? It's Tokenization.
Tokenization is at the heart of every weird LLM behavior. Why they can't reverse strings, why Japanese costs more, why 'SolidGoldMagikarp' breaks them. Here's why — and what you can do about it.
1/5 Tokens In, Logits Out: What's Actually Inside ChatGPT
You use ChatGPT, Gemini, Claude every day. But what's inside the box? Assistants are not models. Once you see the difference, you unlock controls most people don't know exist.