Category
The vocabulary of modern AI
Every other layer of the stack assumes you know these words. This category covers the core ideas behind modern AI in plain English: what a large language model actually is, how tokens and context windows shape everything it can do, and the techniques - fine-tuning, RLHF, quantization, distillation - that turn a raw model into something useful. Start here if any entry elsewhere on the site uses a term you do not recognize.
Entries in this category
- Context window The maximum amount of text, measured in tokens, that a model can consider in a single request. Read the entry
- Distillation Training a smaller model to imitate a larger one, keeping much of the capability at far lower cost. Read the entry
- Embedding A list of numbers capturing a piece of content's meaning, so similar things sit close together in vector space. Read the entry
- Fine-tuning Further training a pretrained model on your own examples to specialize its behavior for a task. Read the entry
- Hallucination When a model states false information fluently and confidently because it generates plausible text, not verified facts. Read the entry
- Inference Running a trained model to produce output - the phase where every user request is actually served. Read the entry
- LLM (large language model) A neural network trained on huge amounts of text to predict the next token, which lets it generate and understand language. Read the entry
- Mixture of experts An architecture that activates only a few specialist subnetworks per token, so a huge model runs at a smaller model's cost. Read the entry
- Multimodal AI Models that accept and reason over more than text - images, audio, or video - within one system. Read the entry
- Open-weights models Models whose trained weights are published for anyone to download, run locally, and fine-tune. Read the entry
- Prompt engineering Designing the instructions, examples, and context you give a model to get reliable, useful output. Read the entry
- Quantization Storing a model's weights in lower-precision numbers so it needs less memory and runs faster. Read the entry
- Reasoning models Models trained to work through problems step by step before answering, trading time and tokens for accuracy. Read the entry
- RLHF (reinforcement learning from human feedback) Training a model toward answers humans prefer by scoring its outputs and reinforcing the good ones. Read the entry
- System prompt Hidden instructions sent before user messages that set a model's role, rules, and tone for the conversation. Read the entry
- Temperature A setting that controls how random a model's word choices are, from deterministic to creative. Read the entry
- Token The chunk of text - roughly three-quarters of a word - that a language model reads and writes one at a time. Read the entry
- Transformer The neural network architecture, built on attention, that underpins nearly every modern language model. Read the entry
Where this layer fits
Concepts are the foundation everything else stands on. The models category compares the systems these ideas describe, retrieval shows how embeddings and context windows get put to work on your data, and inference covers what it takes to run these models in production.
Not sure where to start? The LLM entry is the anchor the rest of this category builds on. Or browse everything at once in the A-Z index.