AI people use way too much jargon, so here’s a rookie-friendly translation
here’s a simple AI jargon survival guide for fellow rookies:
1. LLM — Large Language Model
The technology behind tools like ChatGPT, Claude and Gemini. It predicts and generates language based on patterns learned from huge amounts of text.
2. Token
A small piece of text that an AI model reads or generates. A token can be a word, part of a word or even punctuation.
3. Context window
How much information a model can “see” at one time. A larger context window means it can process longer conversations or documents.
4. Prompt
The instruction or question you give the AI.
5. Hallucination
When an AI gives an answer that sounds confident and believable but is actually wrong or made up.
6. Inference
The process of using a trained model to generate an answer. Training builds the model; inference is what happens when you actually use it.
7. API
A way for developers to connect an AI model to another app or product.
8. RAG — Retrieval-Augmented Generation
A system where the AI searches through documents or a database before answering, instead of relying only on what it learned during training.
9. Embedding
A numerical representation of text that helps computers compare meanings. It allows a system to understand that “car” and “vehicle” are related even though they are different words.
10. Vector database
A database designed to store and search embeddings. It is commonly used in RAG systems.
11. Fine-tuning
Further training an existing model on specialised examples so it becomes better at a particular task, style or industry.
12. AI agent
An AI system that can make decisions and take actions, such as searching the web, reading files or using other tools.
13. Tool calling / function calling
When an AI chooses to use an external tool, such as a calculator, search engine or database, rather than answering entirely by itself.
14. MCP — Model Context Protocol
A standard that helps AI applications connect to external tools and data sources without building every connection from scratch.
15. Open-weight model
A model whose trained weights can be downloaded and run by others. This does not always mean that its training data or full development process is open-source.
A few bonus community terms:
- AI wrapper: a product built mainly around another company’s AI model or API
- AI slop: low-quality, mass-produced AI content
- Vibe coding: building software mainly by describing what you want to an AI
- SOTA: state of the art, meaning one of the best-performing systems available
- Model drop: when a company or research team releases a new model
This is obviously simplified, but hopefully it makes AI discussions slightly easier to follow.