# 1DollarContext > AI concepts explained like a $1 NYC pizza slice: fast, dense, and satisfying. 1DollarContext breaks down AI concepts, research papers, and prompting techniques into short illustrated explainers called "bundles." Each bundle covers one topic in 4-7 minutes. ## Site - [Homepage](https://onedollarcontext.com): Browse all categories and featured bundles. - [All Bundles](https://onedollarcontext.com/wall): Full grid of every published bundle. ## Fresh From The Oven (Weekly AI News) - [Google AI Cracks 9 Unsolvable Erdős Math Problems](https://onedollarcontext.com/bundles/fresh/google-ai-cracks-9-erdos-math-problems): Some of these puzzles stumped human mathematicians for 56 years. Google DeepMind's AlphaProof Nexus just solved 9 of them autonomously. ## Slice of Paper (Academic Research, Simple) - [SkillOpt: Self-Evolving Agent Skills](https://onedollarcontext.com/bundles/paper/skillopt): AI agents are brittle. They break the moment a database schema changes, an API updates, or an environment drifts. That is not a bug — it is a fundamental design flaw. All their tool-handling logic is hardcoded, and when the world moves, the agent falls over. - [Attention Is All You Need](https://onedollarcontext.com/bundles/paper/attention-is-all-you-need): In 2017, Vaswani et al. published "Attention Is All You Need" — replacing recurrence with a parallel architecture that could finally scale across GPU farms. This bundle unpacks how Transformers work, why the "bank" disambiguation example matters, and why this single paper is responsible for GPT, Claude, and every modern LLM. - [BERT & GPT-1: The Fork](https://onedollarcontext.com/bundles/paper/bert-gpt-fork): In 2017, a team at Google published "Attention Is All You Need" and handed the world a new architecture: the Transformer. It had two halves. An Encoder that reads. A Decoder that writes. - [Language Models are Few-Shot Learners](https://onedollarcontext.com/bundles/paper/language-models-are-few-shot-learners): In 2020, OpenAI scaled a language model to 175 billion parameters and discovered something unexpected: at that size, the model no longer needed retraining to learn new tasks. You could just show it a few examples in the prompt — and it figured the rest out on its own. This bundle unpacks how GPT-3 shattered the fine-tuning paradigm and turned "prompt engineering" into a discipline. - [InstructGPT & Chinchilla: The Optimization](https://onedollarcontext.com/bundles/paper/instructgpt-chinchilla): By early 2022, GPT-3 had shown what a 175-billion-parameter model could do. It was also deeply frustrating. Ask it to write a recipe and it might write three more questions about recipes. Ask it to summarize a paragraph and it might keep writing tangentially related text. The model completed sequences. It did not understand instructions. - [The Sensory Shift: Native Multimodality](https://onedollarcontext.com/bundles/paper/native-multimodality): Before 2024, multimodal AI was a magic trick. You had a language model, a vision model, and an audio model, each trained separately, each speaking its own internal language. To make them work together, engineers stitched them into pipelines. Text went through one model, got translated into a shared format, and passed to the next. Information was lost at every handoff. Latency stacked up at every seam. - [The Reasoning Era](https://onedollarcontext.com/bundles/paper/reasoning-era): The scaling laws of 2020 made a bold promise. Train bigger models on bigger datasets. Watch intelligence compound. Every lab in the world took this as the operating plan. ## How to Make AI at Home (Zero to Hero Tutorials) - [Tokenization: How AI Reads Text](https://onedollarcontext.com/bundles/homemade/tokenization): AI doesn't read words. It reads bricks of text called tokens. - [Embeddings: How AI Understands Meaning](https://onedollarcontext.com/bundles/homemade/embeddings-how-ai-understands-meaning): Embeddings are the silent engine behind ChatGPT, Spotify recommendations, and Google Search. This bundle explains the key insight — turning words into points in space — and walks through the famous King − Man + Woman = Queen vector arithmetic that made the AI world lose its mind. - [Vector Databases: Semantic Search at Scale](https://onedollarcontext.com/bundles/homemade/vector-databases): Vector databases are the infrastructure that makes AI search work at scale. Instead of matching keywords, they find semantically similar results by querying embeddings in high-dimensional space. This bundle explains how embeddings translate meaning into mathematics, how cosine similarity measures closeness, and why every modern AI system needs a vector database to power semantic search. - [Neural Networks: The Engine of AI](https://onedollarcontext.com/bundles/homemade/neural-networks): Neural networks are the core architecture powering modern AI. Inspired by biological neurons, they chain together layers of digital nodes that learn by adjusting the strength of their connections. Stack enough layers deep and a neural network can recognise faces, translate languages, generate code, and reason about the world. - [Attention Mechanism & Transformers](https://onedollarcontext.com/bundles/homemade/attention-mechanism): The architectural secret behind GPT, Claude, and Gemini. - [Weights & Biases: AI's Control Panel](https://onedollarcontext.com/bundles/homemade/weights-and-biases): 70 billion dials. All tuned to perfection. - [Pre-Training vs. Fine-Tuning](https://onedollarcontext.com/bundles/homemade/pretraining-vs-finetuning): Two phases. One brain. Here's how a model goes from clueless to expert. - [AI Alignment & RLHF](https://onedollarcontext.com/bundles/homemade/ai-alignment-rlhf): Why brilliant AI can still go terribly wrong. And how we fix it. ## Secret Sauce (Prompt Engineering Hacks) - [Spec-Driven Context Engineering](https://onedollarcontext.com/bundles/sauce/spec-driven-context-engineering): The longer you run an AI coding session, the worse it gets. Not because the model is bad. Because the context window is a landfill. - [The Autonomous Hill-Climber](https://onedollarcontext.com/bundles/sauce/autonomous-hill-climber): Most AI coding assistants are one-shot tools. You write a prompt. They write code. If it breaks, you rewrite the prompt. You, the human, are the feedback loop. - [Prompt Caching](https://onedollarcontext.com/bundles/sauce/prompt-caching): The cost of an AI coding session is almost never the model. It is the tokens. - [Progressive Context Disclosure](https://onedollarcontext.com/bundles/sauce/progressive-context-disclosure): Most developers set up AI coding rules once and call it done. They create a single file, drop every style guide, architecture pattern and database convention into it, and ship. Then they wonder why the AI still seems distracted. ## Full Content - [Full text of all bundles](https://onedollarcontext.com/llms-full.txt): All bundle content in a single file for LLM ingestion.