Tech Glossary

An AI system that can take actions on its own to achieve a goal, using tools, calling APIs, and making decisions across multiple steps without a human approving each one.

A defined way for software systems to talk to each other. Think of it as a menu of operations a service exposes to its callers.

A separate line of development in Git. Lets you work on a feature or fix without affecting the main code until it is ready.

A fast storage layer that holds frequently accessed data so repeated requests do not hit the slower underlying system every time.

A network of servers around the world that cache content close to users so pages and assets load faster.

A practice of automatically building, testing, and deploying code changes. Catches bugs early and ships updates faster.

A saved snapshot of your code in Git, with a message describing what changed. The fundamental unit of version control history.

A program that turns source code (written by humans) into machine code or bytecode that a computer can run.

When environments that should be identical slowly diverge over time because of manual changes. A common source of "works on my machine" bugs.

A lightweight, isolated package that holds an application and everything it needs to run. Docker popularised the concept.

The amount of text an AI model can consider at once. Bigger context windows let the model handle longer documents or longer conversations.

A browser security feature that controls which other websites can access your API. Often the cause of frustrating "blocked by CORS" errors.

A mobile app that shares one codebase across both Android and iOS. Flutter and React Native are popular cross platform frameworks.

Specialised software that lets the operating system talk to a specific piece of hardware like a printer, graphics card, or USB device.

A numerical representation of text, image, or other data that captures its meaning. Similar things have similar embeddings, which makes search and comparison possible at scale.

Scrambling data so only someone with the right key can read it. Essential for keeping passwords, messages, and sensitive data safe.

A system design where services react to events (things that happened) rather than calling each other directly. Often uses queues or message brokers.

Taking a pre-trained AI model and further training it on a specific dataset to make it better at a particular task or domain.

A query language for APIs that lets the client ask for exactly the fields it needs in one request, instead of multiple round trips.

When an AI confidently produces information that sounds correct but is actually wrong or made up. A core challenge in deploying LLMs.

A one way function that turns input data into a fixed length fingerprint. Used to store passwords safely and verify data has not been tampered with.

The protocol that powers the web. Browsers and servers talk to each other using HTTP requests and responses.

Secure version of HTTP that encrypts traffic between your browser and the server using TLS.

Defining cloud infrastructure (servers, databases, networks) in code files instead of clicking through dashboards. Makes environments reproducible.

A data structure that speeds up lookups in a database, similar to the index at the back of a book.

A program that runs source code directly, line by line, without compiling first. Python and JavaScript work this way.

A lightweight text format for representing structured data. The most common format used by web APIs today.

A compact, signed token that proves who a user is. Often used to keep someone logged in across requests.

The core of an operating system that manages memory, processes, and hardware access. Driver development happens at this level.

An open source platform for running and managing containers across many machines. The de facto standard for orchestrating production workloads.

A type of AI model trained on huge amounts of text that can understand and generate human language. ChatGPT, Claude, and Gemini are powered by LLMs.

A protocol introduced by Anthropic in 2024 that lets AI models connect to external tools and data sources in a standardised way.

When a program reserves memory but forgets to release it. Over time the program uses more and more memory until it crashes or slows down.

An architectural style where a system is broken into many small, independently deployable services that communicate over the network.

A software system built as a single, unified codebase and deployed as one unit. Simpler to develop and reason about, but harder to scale parts of independently.

An AI model that can process more than one type of input, such as text, images, audio, or video together.

A mobile app built specifically for one platform (Android or iOS) using its native language and tools. Best performance, but separate codebase per platform.

A family of databases that do not use the strict tabular structure of SQL. Includes document stores, key value stores, and graph databases.

A standard for letting users grant apps access to their accounts on other services (like signing in with Google) without sharing passwords.

The text input you give to an AI model. Good prompts produce better outputs, which is why prompt engineering became a real skill.

A proposal to merge changes from one branch into another, usually reviewed by teammates before being accepted.

A buffer where messages or jobs wait to be processed. Helps systems handle bursts of traffic and decouples producers from consumers.

A technique where an AI system first searches a knowledge base for relevant information, then uses that information to generate an accurate answer. Helps reduce hallucinations.

A storage location for source code, usually managed by Git. Holds the full history of every change ever made to the code.

A style of API design that uses HTTP methods (GET, POST, PUT, DELETE) on resource URLs. The most common pattern for web APIs.

The structure of a database: the tables, columns, and relationships between them. Like a blueprint for the data.

A cloud computing model where you write functions that run on demand without managing servers. AWS Lambda, Azure Functions, and Cloudflare Workers are examples.

The standard language for talking to relational databases. Used to read, write, and modify data in tables.

A group of database operations that either all succeed together or all fail together. Keeps data consistent when many things change at once.

A database designed to store and search through embeddings. Used in RAG systems to find documents related to a query by meaning rather than keyword.

A protocol that keeps a connection open between browser and server for two way real time communication. Used for chat apps and live updates.

A security model that assumes nothing inside the network is automatically trusted. Every request must be verified, regardless of where it comes from.