karminski/one-small-step
One Small Step
One Small Step is an educational resource that explains core AI and large language model concepts through short, accessible articles designed to be read in under five minutes. It covers the structure and function of key LLM components like attention mechanisms and tokenization, as well as foundational machine learning mathematics such as matrix rank and overfitting.
The project also serves as a guide to the GGUF file format, which packages all model parameters and metadata into a single compact binary file for cross-platform deployment without external dependencies. It explains how this format enables efficient model storage, fast loading through memory-mapped file access, and local inference on consumer-grade hardware including CPUs and GPUs.
Beyond AI education, One Small Step functions as a static site generator that builds a complete website from Markdown files at build time. It uses file-based routing to map each Markdown file directly to a URL path, applies reusable HTML templates with content injection, and bundles CSS and JavaScript assets during the build process to reduce client-side load times. The documentation covers both the AI concept explainer series and the static site generation tooling.
Features
- LLM Concept Explainers - An educational series that explains technical AI and LLM concepts in under five minutes.
- AI Concept Tutorials - Provides short educational articles explaining AI agents, multimodal models, and vector embeddings.
- Application Pattern Explanations - Explains common AI application patterns like retrieval-augmented generation and AI agents.
- LLM Architecture Explainers - Describes the structure and function of key LLM components like attention and tokenization.
- LLM Deployment Format Explainers - Describes file formats like GGUF and Safetensors used for storing and deploying large language models.
- Technique Explainers - Explains techniques for adapting pre-trained models, including fine-tuning, LoRA, and knowledge distillation.
- Format Explainers - An educational resource explaining the GGUF file format for storing quantized LLMs.
- URL Routing - Maps each Markdown file in the content directory directly to a corresponding URL path.
- Model Binary Formats - Stores quantized large language models in a single binary file loadable from multiple programming languages.
- Markdown to HTML Converters - Converts Markdown files into HTML through a build-time processing step.
- Static Site Generation - Generates all HTML pages at build time from Markdown content and templates.
- Static Site Generators - Builds a complete website from Markdown files with reusable templates and asset bundling.
- Technique Explainers - Explains methods like speculative decoding, quantization, and Flash Attention that improve LLM inference speed.
- File-Based Routing Systems - Maps each Markdown file directly to a URL path without manual route configuration.
- Template Content Placeholders - Uses reusable HTML templates with placeholders to inject dynamic content from Markdown.
- Template-Driven Rendering - Uses reusable HTML templates with placeholders to inject dynamic content from Markdown.
- AI Hardware Explainers - Clarifies how hardware components like PCIe retimers and CPU caches affect AI performance.
- Fine-Tuning vs RAG Comparisons - Explains when to apply fine-tuning versus retrieval-augmented generation and how LoRA works.
- Large Language Model Fine-Tuning - Explains how to adjust a large language model for specific tasks and store updated weights with custom prompt templates.
- GGUF Loading Guides - Explains how to load quantized GGUF models for inference using frameworks like llama.cpp and ggml.
- Memory-Mapped Loading - Loads models instantly by using memory-mapped file access for near-instant startup.
- Mathematical Foundations - Explains core math concepts such as matrix rank and overfitting that underpin machine learning.
- Storage Optimizations - Describes how GGUF uses compact binary encoding and efficient data structures to reduce model storage footprint.
- Cross-Platform Model Execution - Explains how GGUF format enables loading models across different programming languages without external libraries.
- RAG and Embedding Explainers - Explains retrieval-augmented generation and how vector embeddings ground LLM responses.
- Content Directories - Organizes the entire site around a content directory where each Markdown file represents a standalone page.
- Markdown Page Directories - Organizes a site around a content directory where each Markdown file becomes a standalone page.
- Model Packaging - Packages all model parameters and metadata into a single compact binary file for deployment.
- Build-Time Asset Processing - Combines and minifies CSS and JavaScript files during the build process to reduce client-side load times.
- Model Binary Packaging - Packages all model parameters and metadata into a single compact binary file for deployment.
- LLM Prototyping - Explains how GGUF format enables rapid loading and testing of different language models during development.
- Memory-Mapped Startup Techniques - Explains how GGUF format uses memory-mapped file access for near-instant model loading and startup.
- Machine Learning Mathematics - Explains foundational mathematical ideas such as matrix rank and overfitting relevant to machine learning.
- Hardware Concept Explainers - Clarifies hardware topics like PCIe Retimer, NVMe SSD DRAM, and system memory metrics for AI workloads.
- Model Packaging - Describes how GGUF packages all model parameters and metadata into a single file for dependency-free deployment.
- Courses and Tutorials - Tutorials on small-scale AI model development.
- Learning Resources - Practical guide to LLM development.
