Software Engineering Vocabulary: 1,143 Key English Terms

A named block of code that does one job. It can take input values, and it often returns a result to the code that called it.

A name in a program that stores a value, such as a number, text or list, so that the code can use or change that value later.

All the source code that makes up a project or product, seen as one whole, including its structure, style and history.

A collection of ready-made code that other programs can use, so that developers don't have to write common things like date handling or HTTP requests themselves.

To make a function, method or API run from another part of the code. As a noun, 'a call' is one such request, for example 'an API call'.

A piece of code that repeats the same steps several times, for example once for every item in a list or until a condition becomes false.

In object-oriented programming, a template that defines the data and behavior of a type of object, such as a User or an Invoice.

A named input that a function or method accepts, listed in its definition. The actual value you pass in when you call it is called the argument.

Short for integrated development environment: an application for writing code that brings together an editor, a debugger, search and other tools in one place.

To turn source code written by people into code that a computer can run, using a program called a compiler.

A piece of text in a program, such as a name, a message or a password, stored as a sequence of characters and usually written inside quotes.

A data type that has only two possible values, true or false, used for yes-or-no decisions such as whether a user is logged in.

A unit in a program that groups related data and the functions that work on it, for example a user object with a name, an email and a login method.

A function that belongs to a class or object and usually works with that object's data, for example user.save() or list.sort().

The actual value you give to a function when you call it. In add(2, 3), the numbers 2 and 3 are the arguments.

To end a function and send a value back to the code that called it, for example when a password check gives back true or false.

A piece of code that runs only when a condition is true, often followed by an else part that runs when the condition is false.

The rules for how code must be written in a programming language, such as where brackets, commas and keywords go, so that the computer can read it.

A note inside the code that the computer ignores, written to explain to other developers why something is done. Remarks left in a code review are also called comments.

The human-readable instructions that programmers write in a programming language, before they are turned into a program the computer can run.

Describes software whose source code is public, so anyone can read, use, change and share it under the terms of its license.

To turn a plan, design or feature idea into working code. In object-oriented code, it also means to provide the methods that an interface requires.

In programming, a list of methods that a class promises to provide, without saying how they work. More generally, the point where two systems, or a user and a system, meet.

A style of programming that organizes code into objects that combine data and behavior, built on ideas such as classes, inheritance and encapsulation.

The time when a program is actually running, as opposed to when it is written or compiled. Also the software environment that runs the code, such as Node.js.

A tool that downloads, installs, updates and removes the libraries a project depends on, such as npm for JavaScript or pip for Python.

A relationship in which a class receives behavior or structure from another class and can add or change selected parts.

The ability to use different implementations through a shared interface while each implementation provides its own behavior.

Keeping an object's internal state behind a controlled interface so that other code does not depend on its internal representation.

A special operation used when creating an object to establish its initial state and any required dependencies.

A special operation associated with the end of an object's lifetime, commonly used to release resources owned by that object.

A function together with access to variables from its surrounding lexical environment, even when it is called from elsewhere.

A function passed to another piece of code so that the receiving code can invoke it when needed.

A named value or binding that cannot be reassigned after initialization under the rules of the programming language.

A numeric data type representing whole numbers without a fractional part, with a supported range that depends on its representation.

A number stored using a finite-precision representation with a significand and an exponent, allowing a wide range of magnitudes.

A type or declared set of named alternatives that represents a limited collection of possible values.

A value that refers to a memory location, allowing code to access an object or data stored at that location.

A way for code to refer to an existing value or object without necessarily making an independent copy of it.

Describing a value or object whose state cannot be changed after it has been created.

Describing a value or object whose state can be changed after it has been created.

A sequence of execution within a process that can run alongside other sequences while sharing some process resources.

A synchronization mechanism that allows only one cooperating execution context at a time to enter a protected section.

A situation in which a program's result depends on the timing or ordering of operations that may run concurrently.

Automatic reclamation of memory occupied by objects that a runtime determines are no longer reachable or otherwise needed.

Converting structured data or an object's state into a representation that can be stored or transmitted and read back later.

Reading a stored or transmitted representation and reconstructing the data structures or values it describes.

The structure that keeps track of active function calls, usually with the most recent call on top.

A function-like routine that can pause and resume later, allowing cooperative multitasking or simpler asynchronous code.

The memory block created for a function call, holding local data, parameters, and return information during execution.

Program state that can be changed after creation, such as variables or objects whose stored values are updated over time.

A runtime mechanism that waits for events and schedules tasks or callbacks, so many operations can be handled without blocking the program.

A rule that resolves names from the surrounding code structure, rather than from the sequence of function calls.

A situation where a name declared in an inner scope hides the same name from an outer scope.

Evaluating only as much of an expression as needed to determine its result, skipping an operand when possible.

Determining the type of a value or expression from code, without requiring the programmer to write that type explicitly.

An automatic conversion from one data type to another while an expression or operation is evaluated.

Choosing a code branch by checking a value's structure or contents, often also giving names to its parts.

A function that returns the same result for the same inputs and has no observable effects outside its calculation.

A function that accepts another function as an argument or returns a function as its result.

A function call in the final position of another function, so its result becomes the caller's result without further calculation.

A synchronization mechanism with a count that controls how many tasks may access a resource or proceed at once.

A managed group of reusable threads that execute tasks submitted by a program.

Program behavior for which the language specification sets no requirements, typically because the code violates a rule of the language.

A clear, step-by-step method for solving a problem or doing a task, such as sorting a list or finding the shortest route between two points.

A way of organizing data in memory, such as an array, list, tree or hash map, so that certain operations are fast and simple.

An ordered collection of items stored one after another, where each item can be reached by its position number, called an index.

To go through the items of a collection one by one, or more generally to repeat a process and improve the result each time.

A data structure that stores key-value pairs and finds the value for a given key very quickly, usually in constant time.

A technique where a function solves a problem by calling itself on a smaller part of the same problem until it reaches a simple case.

A way of describing how the running time or memory use of an algorithm grows as the input gets bigger, for example O(1), O(n) or O(n²).

How the number of steps an algorithm takes grows as the size of its input grows, usually written in Big O notation.

The position number of an item in an array or list. In most languages counting starts at zero, so the first item has index 0.

To put items in a particular order, for example numbers from smallest to largest or names alphabetically. Many algorithms exist to do this efficiently.

A data structure where the last item added is the first one taken out (LIFO), like a pile of plates. You push items onto the top and pop them off the top.

A data structure where the first item added is the first one taken out (FIFO), like people waiting in line. Also a system that holds tasks or messages until they are processed.

A data structure made of nodes, where each node holds a value and a pointer to the next node, so items are connected in a chain instead of stored side by side.

A data structure with a root at the top and branches of child nodes below it, used for data that has levels, such as folders, menus or a page's HTML.

A fast way to find a value in a sorted list by checking the middle item and repeatedly throwing away the half where the value cannot be.

A simple approach that tries every possible option until it finds the answer. It is easy to write but often too slow for large inputs.

How much extra memory an algorithm needs as its input grows, usually written in Big O notation, for example O(1) or O(n).

An ordered group of values treated as one value, often allowing elements of different types in fixed positions.

A collection that stores distinct values and supports operations such as membership checks, union and intersection.

A tree-based data structure with an ordering rule between parents and children, often used to retrieve the smallest or largest item efficiently.

An object or protocol that provides values from a sequence one at a time while tracking progress through that sequence.

An ordering of a directed acyclic graph in which every dependency appears before the item that depends on it.

A tree data structure kept roughly even in height, so search, insert, and delete operations stay efficient.

An optimization technique that stores results of expensive function calls and reuses them when the same inputs appear again.

A graph or tree traversal method that visits all nearby nodes first, moving level by level before going deeper.

A graph or tree traversal method that follows one path as far as possible before backtracking to explore others.

A problem-solving technique that breaks a task into overlapping subproblems and stores results to avoid repeated work.

A matrix representation of a graph where each cell shows whether, or how strongly, two nodes are connected.

An algorithm that chooses the best immediate option at each step, hoping those local choices lead to a good overall result.

A tree data structure that stores summaries for intervals of an array, supporting efficient range queries and updates.

A tree structure that stores all suffixes of a string, allowing fast searches for substrings and repeated patterns.

A tree data structure in which each node has at most two children, usually called the left child and right child.

A data structure that stores the starting positions of all suffixes of a string in sorted order for fast text search tasks.

A way to represent a graph by storing, for each node, a list of the nodes directly connected to it.

A collection that removes items according to their priority rather than simply in the order they arrived.

A tree that stores sequences by sharing their common prefixes along paths from the root.

A compact structure that tests set membership. It may report an absent item as present, but a standard correctly used filter does not miss inserted items.

A structure that tracks non-overlapping groups and supports finding an item's group and joining groups efficiently.

Studying total cost across a sequence of operations to find a cost per operation, even when individual operations vary greatly.

A sorting method that preserves the original relative order of items whose sorting keys are equal.

A case where two different inputs produce the same hash value.

A fixed-capacity buffer whose read or write position wraps back to the beginning when it reaches the end.

A request to merge your code changes from one branch into another, so that teammates can review and discuss them first.

A saved snapshot of changes in a repository, with a short message that explains what was changed and why. Also used as a verb: to commit your changes.

A separate line of development in a repository, where you can work on changes without affecting the main code until you are ready to merge.

To combine the changes from one branch into another, so that both sets of work end up in the same code.

A storage place for a project's code and its full history of changes, usually managed with Git and often hosted online.

To send your local commits to a shared remote repository, so that other people and automated systems can see them.

A situation where Git cannot combine two branches automatically because both changed the same lines, so a person has to choose the right version.

In Git, to switch your working files to a particular branch or commit so that you can see or work on that version of the code.

To download a full copy of a remote repository, including its history, to your own computer so that you can work on it.

In Git, to move your branch's commits so that they start from a newer point, usually the latest main, which gives a cleaner, straight-line history.

A system that records every change to a set of files over time, so a team can work on the same code, see who changed what and go back to earlier versions.

The most widely used version control system: a free, open-source tool that tracks changes in code and lets many developers work on the same project.

To download the latest changes from a remote repository and merge them into your local copy, so your code is up to date with your team's.

A view or file that shows exactly which lines were added, removed or changed between two versions of code, usually in green and red.

The short text that describes what a commit changes and why. Teammates read it later to understand the history of the code.

The central branch of a repository that holds the official, up-to-date version of the code. Other branches are created from it and merged back into it.

A separate branch created to work on one feature or fix, so the work can be developed and reviewed without affecting the main branch.

To undo a change by creating a new commit that does the exact opposite, so the earlier state returns while the history stays intact.

To make your own copy of someone else's repository on a platform such as GitHub, so you can change it freely and later propose your changes back.

In Git, to save your unfinished changes in a temporary place and clean your working copy, so you can switch tasks and bring the changes back later.

In Git, to copy one specific commit from one branch onto another without merging the whole branch. In general English, to choose only the best or most useful items.

A Git collaboration model where each contributor works in a personal fork, then submits changes back to the main repository.

A Git rebase mode that lets you reorder, combine, edit, or remove commits while rewriting branch history.

A best common ancestor of two commits, used by Git when calculating a three-way merge.

A Git merge that simply moves the target branch pointer ahead because it has no new commits of its own after branching.

To combine several commits into a single commit, usually to make branch history cleaner before merging or review.

A checkout state where HEAD points directly to a commit rather than to a branch name.

A local history of changes to references, such as where HEAD or a branch pointed at earlier times.

A checkout that places only selected repository paths in the working tree while keeping other paths out of it.

A repository copy with a limited portion of commit history rather than the full history available from its source.

A merge that compares two versions with a shared ancestor to identify and combine the changes made on each side.

To locate a change in an ordered series of revisions by testing a midpoint and repeatedly narrowing the remaining range.

The practice of teammates reading and commenting on each other's code changes before they are merged, to catch problems and share knowledge.

To improve the structure of existing code, making it cleaner or easier to change, without changing what it does for the user.

The future cost of shortcuts taken in code today: quick solutions that save time now but make later changes slower and riskier.

Short for 'looks good to me': a quick written comment a reviewer leaves to say that a code change is fine and can be merged.

A very small, optional comment in a review about details like naming, spacing or wording, which should not block the change. Often shortened to 'nit'.

To deal with the comments people gave you, by making the requested changes or explaining why you chose a different approach.

A tool that scans code without running it and warns about style problems, suspicious patterns and simple mistakes, such as unused variables.

Written directly into the source code instead of coming from settings, a database or user input, so it cannot change without editing the code.

A sign in code that suggests a deeper design problem, such as a very long function, duplicated logic or too many parameters, even if the code works.

A teammate who reads and checks someone else's code changes, usually in a pull request, before they are allowed to be merged.

To officially say that a change, plan or request is acceptable. In code review, approving a pull request allows it to be merged.

Easy for another person to read and understand quickly, thanks to clear names, a simple structure and short functions.

A way of working that experienced people widely agree gives the best results, such as writing tests, reviewing code and never storing passwords in plain text.

Code that is simple, clearly named, well organized and easy for others to read, test and change, rather than code that only works.

Easy to understand, fix and extend over a long time, even by developers who did not write the original code.

An agreed set of rules for how to name things in code, such as variables, files and branches, so that names look consistent across a project.

Code that is never run or whose result is never used, for example an old function nobody calls. It only makes the codebase harder to read.

Short for 'Don't Repeat Yourself': the principle that each piece of knowledge or logic should exist in one place in the code instead of being copied around.

A note left during review to question, suggest, or explain a specific part of code, text, or a change.

A shared set of coding rules for style, structure, naming, and other practices to keep code consistent.

A build process that produces the same output every time from the same source, tools, and inputs.

A code measure based on the number of independent paths through a program's control flow.

An early check that exits a function or block for a special case before the main logic runs.

A code smell where one conceptual change requires many small edits across different parts of a codebase.

A code smell where a method relies heavily on another object's data, suggesting its responsibility may be misplaced.

Examining code without executing it to discover possible errors or check properties of the program.

A small automated test that checks one function or class on its own, separate from the database, the network and the rest of the system.

One specific situation you check, with set inputs, steps and an expected result, for example 'logging in with a wrong password shows an error'.

An unusual situation at the limits of what a program expects, such as an empty list, a very long name or a date like 29 February.

Short for quality assurance: the work of checking that software meets its requirements and works well. It also names the people or team who do this work.

A bug where something that used to work stops working after a change, such as a new feature breaking an old one.

A test that checks how several parts of a system work together, for example your code with a real database or another service.

A measure, often a percentage, of how much of the code is run by automated tests. It shows which parts are tested and which are not.

A fake version of a real object or service used in tests, so you can control its behavior and check how your code uses it. Also a verb: to mock a service.

A test that runs a complete user journey through the real system, for example opening the site, logging in and placing an order, to check that everything works together.

Describing a test that sometimes passes and sometimes fails without any change to the code, often because of timing, the network or shared data.

A collection of tests that are run together, for example all the unit tests of a project or every test for the payment module.

Checking software by hand, with a person clicking through screens and trying actions, instead of using test scripts that run automatically.

Using code and tools to run tests automatically and compare the results with what is expected, so they can be repeated quickly after every change.

A line in a test that checks whether a result matches the expected value. If the check is false, the test fails.

A quick, basic test that checks the most important functions work at all, for example that the app starts and users can log in, before deeper testing.

The normal scenario in which everything goes as expected: the user enters valid data, nothing fails and the task completes successfully.

Short for test-driven development: writing a failing test first, then the code that makes it pass, then cleaning up the code, in short repeated cycles.

A separate copy of a system, with its own servers, data and settings, where software is tested without affecting real users.

Short for user acceptance testing: the last testing stage, in which real users or the client check that the software meets their needs before it goes live.

A record created when an app crashes, showing the error, the stack trace, the device, the OS version and other details that help developers find the cause.

Data about how people use an app, such as which screens they open and where they leave, collected as events to understand behavior and measure features.

The percentage of users or sessions that did not experience a crash in a given period; a key measure of an app's stability.

The wide variety of devices, screen sizes, hardware and operating system versions an app must work on, which makes testing harder, especially on Android.

An automated test that uses the app the way a person would, tapping buttons, typing text and checking what appears on the screen.

Faster than normal battery use caused by an app, for example through constant location tracking, frequent network calls or work that keeps running in the background.

How much space an app needs, both as a download from the store and after installation on the device, including its code, images and libraries.

Application Not Responding: an Android error shown when an app's main thread is blocked for too long, for example about five seconds without responding to a touch.

A quick, rough version of a game or feature made to test whether an idea is fun or works, before the team spends time on final art and polish.

A session in which people play an unfinished version of the game while the team watches and collects feedback on fun, difficulty and confusing moments.

An agreed point in development when specific work must be delivered and reviewed, such as a playable demo, alpha or beta; publishers often pay the studio per milestone.

The stage of development when all main features are in the game and it can be played from start to finish, but it still has many bugs, placeholders and no polish.

The stage after alpha when the game's content is complete and the team focuses on fixing bugs and polishing; also a public or closed test of that version.

A short but fully finished part of a game, such as one level, that shows final quality in art, sound and gameplay to prove what the whole game will be like.

How easily, quickly and successfully people can use a product to complete a task, and how satisfied they feel while doing it.

A research method in which real users try to complete tasks with a product or prototype while a researcher watches, to find out where they struggle.

A review in which usability experts check an interface against a list of accepted design principles, called heuristics, to find problems without testing with users.

A testing technique in which participants say what they are thinking, looking for and feeling while they use a product, so researchers can follow their reasoning.

The percentage of participants who complete a given task correctly in a usability test; one of the simplest and most common usability metrics.

Usability testing in which a researcher guides each session live, in person or on a video call, giving tasks, asking questions and observing the participant.

A simple test replacement for a real function, service, or module that returns preset data under controlled conditions.

Testing general rules about a program across many generated inputs instead of checking only individually chosen examples.

Making small deliberate changes to code and checking whether existing tests fail in response.

Trying many automatically generated or modified inputs to expose crashes, security weaknesses or other unexpected behavior.

A source or mechanism used to decide whether the observed result of a test is correct.

A test-design method that selects values at and near the limits of allowed or meaningful input ranges.

Dividing possible inputs into groups expected to behave alike, then testing representative values from those groups.

Testing whether communicating components follow agreed expectations about requests, responses or messages.

Prepared data, objects or resources that put a test into a known starting state.

A mistake in a program that makes it behave in a way nobody intended, such as showing wrong results, freezing or closing suddenly.

To change code so that a problem no longer happens. As a noun, 'a fix' is the change that solves the problem.

To search for the cause of a problem in a program and remove it, often by running the code step by step and checking values.

To stop working suddenly because of an error, so that the program closes or stops responding. Also a noun: 'a crash'.

To make a bug happen again on purpose by following the same steps, so that you can see it yourself and find its cause.

A record of events, messages and errors that a program writes while it runs, used later to understand what happened. Also a verb: to log an error.

A report printed when an error happens, listing the chain of function calls that led to it, with file names and line numbers.

The deepest, original reason a problem happened, as opposed to the symptoms that people noticed first.

A temporary way to avoid a problem without really fixing it, so that work can continue until a proper solution is ready.

An error event that interrupts the normal flow of a program. The code can catch and handle it, or it can make the program crash.

A marker you put on a line of code so that the program pauses there while it runs, letting you inspect variables and step through the code.

Text that a program shows when something goes wrong, saying what failed and sometimes why, for example 'File not found' or 'Connection refused'.

To look for the cause of a problem step by step and fix it, especially in a running system, a setup or a user's environment.

A written description of a problem in software, usually saying what happened, what should have happened and how to make it happen again.

A mistake in how code is written, such as a missing bracket or a misspelled keyword, which stops the program from compiling or running.

A tool that lets you pause a running program, go through it line by line and look at the values of variables to find out why it misbehaves.

To run code one line or instruction at a time in a debugger, checking what happens after each step.

Code that deals with things going wrong, such as a failed request or bad input, so the program can recover, retry or show a clear message instead of crashing.

A bug where a program keeps memory it no longer needs, so its memory use grows over time until it slows down or crashes.

A loop that never stops because its exit condition is never met, which can freeze a program or use 100% of the CPU.

A common bug where a loop or calculation goes one step too far or stops one step too early, often because counting starts at zero.

The exact list of actions someone must take to make a bug appear again, written in a bug report so a developer can see the problem for themselves.

Continuously collecting and checking data about systems, such as CPU use, error rates and response times, so that problems are noticed quickly.

An automatic notification sent to engineers when a monitored value crosses a limit or something looks wrong, for example when the error rate is too high.

An unplanned event that disrupts a service or makes it worse for users, such as an outage, a big slowdown or a data problem, and that needs a quick response.

A period when a service, system or part of it is not working or cannot be reached by users.

The amount of time a system or service is not available to users, either because of a failure or because of planned maintenance.

The time or percentage of time that a system is working and available, often written as a figure such as 99.9 percent over a month or a year.

Short for service level agreement: a contract that promises customers a certain level of service, such as 99.9 percent uptime, and says what happens if it is not met.

Available outside normal hours to respond quickly if an alert or incident happens. Engineers take turns being on call for their services.

A number measured regularly over time to show how a system is behaving, such as CPU use, memory, request count or error rate.

A screen that shows the most important metrics and charts about a system in one place, so engineers can see its health at a glance.

How well you can understand what is happening inside a system from the data it produces, especially logs, metrics and traces, even for problems nobody predicted.

Following a single request as it passes through many services, recording how long each step takes, so you can see where time is lost or errors happen.

A regular automatic test, often a request to a special URL such as /health, that checks whether a server or service is working and able to take traffic.

A written review after an incident that explains what happened, the impact, the timeline, the root cause and the actions that will stop it happening again.

Having extra copies of servers, network links, disks or data, so that the system keeps working if one of them fails.

The plans, tools and processes for bringing systems and data back after a major event, such as losing a whole data center or region.

Short for service level objective: an internal reliability target for a service, such as 99.9 percent of requests succeeding over 30 days.

Short for site reliability engineering: using software engineering methods to run production systems reliably. It also names the engineer who does this work.

A written step-by-step guide for handling a specific task or alert, such as restarting a service or clearing a full disk, so anyone on call can follow it.

A file or record saved when a program crashes, containing memory and state data for later debugging.

In a debugger, to continue execution until the current function finishes and control goes back to the code that called it.

Very detailed logging that records small execution steps, used to follow program flow during debugging or hard-to-find failures.

A debugging method where you pause execution, inspect state, and test commands step by step inside a running program.

A debugger setting that pauses execution when a selected memory location is read or changed, depending on its configuration.

Accessing a memory region after it has been released, when it is no longer valid for the original object.

A write beyond the allocated boundary of a memory buffer, which may damage other data or cause a crash.

A failure where tasks keep reacting or retrying but prevent one another from making useful progress.

A scheduling problem where a high-priority task must wait for a resource held by a lower-priority task.

An informal name for a bug that changes or disappears when developers try to observe or investigate it.

A ready-made structure for building applications that gives you rules, tools and common parts, so you only write the code that is specific to your product.

A library, service or other piece of software that your code needs in order to work, so a change or failure in it can affect your project.

The set of programming languages, frameworks, databases and tools that a team uses to build and run a product.

The high-level structure of a software system: its main parts, how they communicate with each other and the key technical decisions behind them.

A proven, reusable solution to a common design problem in code, such as Singleton, Factory or Observer, described as a general approach rather than finished code.

A situation where getting more of one good thing means accepting less of another, such as speed versus accuracy or simplicity versus flexibility.

Able to handle more users, data or traffic by adding resources, without a complete redesign and without becoming much slower.

An architecture in which an application is built as many small, independent services, each with one job, that talk to each other over the network.

An application built and deployed as one single unit, where all features share the same codebase and usually the same database.

Built from separate, independent parts with clear boundaries, so each part can be understood, tested, changed or replaced without touching the others.

Hiding complex details behind a simpler interface, so you can use something without knowing how it works inside, like calling save() without knowing the database query.

A design principle saying that each part of a program should handle one clear job, such as showing data, applying business rules or storing data, without mixing them.

A design in which the parts of a system depend on each other as little as possible, so a change in one part rarely forces changes in the others.

A technique where an object receives the things it needs, such as a database client, from outside instead of creating them itself, which makes code easier to test.

A document written before building something big that explains the problem, the proposed solution, the other options considered and their trade-offs, so others can review it.

Making a solution more complex than the problem needs, for example adding layers, options or flexibility for future needs that may never come.

A small, quick version of an idea, built only to show that it can work technically before the team spends time building it properly.

A software framework that gives developers ready-made systems for graphics, physics, sound, input and scripting, so they can build a game without writing everything from scratch.

Any piece of content used in a game, such as a 3D model, texture, sound effect, animation or music track, usually stored as a file and imported into the engine.

The number of images a game draws on the screen each second, measured in frames per second (fps). A higher, steadier frame rate makes the game look and feel smoother.

A document that describes the planned game in detail, including its concept, mechanics, story, levels, characters and controls, so the whole team builds the same game.

Short for non-player character: any character in the game that is controlled by the computer, not by a player, such as a shopkeeper, a quest giver or an enemy guard.

Short for heads-up display: the information shown on top of the game view while you play, such as health, ammo, a mini-map and the score.

A 2D image used as an object in a game, such as a character, an enemy, a coin or an explosion. Several sprites shown one after another create an animation.

The part of a game that checks whether two objects touch or overlap, for example a bullet hitting an enemy or the player landing on a platform.

A small program that runs on the graphics card and decides how surfaces look, for example their color, lighting, shine and transparency, or effects like water and fire.

A piece of code that controls how something in the game behaves, such as a door that opens, an enemy that patrols or a quest that starts, usually attached to objects in the engine.

The code cycle at the heart of every game that runs many times per second: read the player's input, update the game world, then draw the new frame.

One part of a game project in the engine, such as a menu, a level or a test area, that contains the objects, lights and cameras placed in it.

A saved, reusable game object with all its parts and settings, such as an enemy or a door, that you can place many times; editing the prefab updates every copy.

The system that simulates real-world physics in a game, such as gravity, collisions, friction and forces, so objects fall, bounce and push each other believably.

An invisible shape around a character, or part of one, that the game uses to decide whether an attack or bullet has hit. It is often simpler than the visible model.

A 2D image wrapped around a 3D model to give it color and surface detail, such as wood grain, skin or rust. Extra textures can control shine, bumps and roughness.

The number of polygons, usually triangles, that make up a 3D model or a whole scene. Fewer polygons are faster to draw but show less detail.

Work that makes a game run faster and use less memory, battery or bandwidth, for example by simplifying models, reducing draw calls or rewriting slow code.

To make a character, enemy or item appear in the game world, or to appear in it. A spawn point is the place where this happens.

A point in a level where the game saves the player's progress, so after failing or dying they start again from there instead of from the beginning.

A large game world that players can explore freely and in almost any order, instead of moving through a fixed series of levels.

An early version of a level built from simple grey shapes with no final art, used to test layout, scale, routes and timing before artists spend time on it.

The shortest route through a level or game that the player must follow to finish it, as opposed to optional side areas, secrets and side quests.

Creating game content such as maps, dungeons, terrain or loot automatically with algorithms and rules, instead of building every piece by hand.

Short for navigation mesh: a simplified map of the walkable surfaces in a level that game AI uses to find paths and move characters around obstacles.

An object in the physics engine that keeps its shape and is moved by forces such as gravity and collisions, for example a falling crate or a thrown grenade.

A check in which the game sends an invisible straight line from a point in a direction to find what it hits first. It is used for shooting, line of sight and ground detection.

A physics setup in which a character's body becomes a set of connected parts that fall and react to forces realistically, usually when the character dies or is knocked down.

A frame in an animation where the animator sets an important pose or value; the software then fills in the movement between keyframes automatically.

The hidden skeleton of bones and controls inside a 3D model that animators move to pose and animate it, like the strings of a puppet.

A command the CPU sends to the graphics card to draw a group of polygons with one material. Too many draw calls per frame can slow a game down.

Short for graphics processing unit: the chip that renders a game's images by running shaders and drawing millions of pixels every frame. The CPU handles most of the game logic.

Visual effects applied to the whole image after the scene has been rendered, such as bloom, motion blur, color grading and depth of field.

A tool that creates effects such as fire, smoke, sparks, rain and magic by producing and moving many small images or shapes called particles.

A technique that swaps a model for simpler versions as it gets farther from the camera, so distant objects use fewer polygons and the game runs faster.

To calculate something expensive, such as lighting, shadows or navigation data, before the game runs and save the result, so it does not need to be computed in real time.

A tool that measures where a game spends its time and memory, frame by frame, so developers can find what makes it slow instead of guessing.

A system design approach that separates responsibilities into layers, such as presentation, business logic, and data access.

An architectural style where system capabilities are organized as separate services that communicate through defined interfaces.

A software design style where components communicate by producing and reacting to events instead of direct, tightly coupled calls.

A design that isolates resource pools or components so a failure in one part has less effect on other parts.

Keeping a sequence of domain events as the authoritative record and deriving application state from those events.

A design principle that separates operations which change state from operations which return information without changing state.

A design that keeps domain logic independent of external systems by connecting them through defined ports and adapters.

Replacing a legacy system gradually by moving capabilities to a new system and routing their traffic to it.

A way for a receiving component to slow or limit incoming work when it cannot process that work fast enough.

The property that repeating an operation has the same intended effect as performing it once.

A later business operation that offsets the effects of an earlier completed operation when the whole workflow cannot finish.

A piece of functionality that users can see or use, such as a search bar, dark mode or the option to export a report.

A record in a tracking tool such as Jira that describes one piece of work, like a bug, task or feature, and shows who owns it and its status.

A fixed period, usually one or two weeks, in which a team commits to finishing a chosen set of work items.

A short daily team meeting, often 15 minutes, where each person says what they did, what they will do next and whether anything is blocking them.

Something the software must do, or a condition it must meet, as agreed with users, clients or the product team.

The ordered list of work a team has not started yet, such as features, bugs and improvements, with the most important items at the top.

To judge roughly how much time or effort a task will take before you start it. As a noun, 'an estimate' is that rough figure.

A problem that stops someone from continuing their work until it is solved, such as missing access, an unanswered question or a broken environment.

A short description of a feature from the user's point of view, often written as 'As a [type of user], I want [goal] so that [reason]'.

Describing a way of working in short cycles, delivering small pieces often and changing plans based on feedback, instead of following one long fixed plan.

To have a short meeting or chat to share updates and make sure everyone has the same information and plan. Also a noun: 'a quick sync'.

Not part of the work that was agreed for a project, sprint or ticket, so it will not be done now, even if it is a good idea.

A popular agile framework in which a small team works in short, fixed periods called sprints, with set roles and regular meetings to plan and review the work.

A meeting at the end of a sprint or project where the team talks about what went well, what went badly and what to change next time.

To decide which tasks are most important or urgent and should be done first, especially when there is not enough time for everything.

A list of specific conditions that a feature or user story must meet before the team and the product owner agree that it is done.

A unit that agile teams use to estimate how much effort a task needs, based on its size, complexity and uncertainty rather than on hours.

To send someone a short message, usually in team chat, to ask a question, remind them or get their attention. Technically, also to test whether a server can be reached.

Short for 'estimated time of arrival'. At work it usually means when you expect a task, fix or delivery to be finished.

A way of working in which two developers write code together at one computer or shared screen: one types, while the other reviews and thinks ahead.

The slow, uncontrolled growth of a project's requirements after work has started, as small extra requests keep being added without changing the deadline or budget.

Any person or group that is affected by a project or has a say in it, such as customers, managers, other teams, the legal team or support.

The person who decides what a product should do and why, by understanding users and the business, and who guides the team that builds it.

A high-level plan that shows what a product team intends to work on over the coming months or quarters, and why, without detailing every task.

Short for minimum viable product: the simplest version of a product that can be released to real users to learn whether the idea is worth building further.

Short for product requirements document: a document that explains what a feature or product should do, for whom and why, so the whole team works from the same plan.

The moment a new product or feature is made available to users, together with the marketing, sales and support activities planned around it.

In Scrum, the person accountable for the product backlog: deciding its order, making items clear and making sure the team works on the most valuable things.

A specific situation in which someone uses a product to reach a goal, described step by step or as a short scenario.

Short for objectives and key results: a goal-setting method in which a team sets an inspiring objective and two to five measurable results that show progress toward it.

A large piece of work that is too big for one sprint and is broken down into several smaller user stories, such as “Multi-currency support”.

A prioritization method that gives each idea a score based on Reach, Impact, Confidence and Effort, so that very different ideas can be compared.

A prioritization method that sorts requirements into four groups: Must have, Should have, Could have and Won't have (this time).

The planned order and dates of the main steps of a piece of work, from the start to the launch.

To check with evidence, such as user tests, data or experiments, that an idea, assumption or solution is actually right before investing more in it.

Short for specification: a document that describes in detail how a feature should look and behave, so that designers and engineers can build it correctly.

A suggestion from a customer, user or colleague asking for a new feature or for a change to an existing one.

A shared checklist of quality conditions that every piece of work must meet before the team can call it done, such as tested, reviewed and documented.

Regular work in which the product owner and the team review upcoming backlog items, clarify them, split large ones and estimate them so they are ready for a sprint.

To disagree with or resist a request, decision or deadline, usually politely and with reasons; also, to move a date to later.

A team made up of people with different skills, such as product, design, engineering and data, who work together toward the same product goal.

To give formal approval that a document, design or piece of work is acceptable, so that the next step can begin.

A short, regular report that tells stakeholders what has been done, what is coming next and what risks or blockers exist.

A formal decision, usually made in a meeting shortly before launch, about whether a product or release is ready to go ahead or should be delayed.

To gradually retire a product or feature and stop supporting it, usually after warning users and helping them move to an alternative.

A metric that changes early and helps predict a future result, such as weekly logins predicting renewals, so a team can act before it is too late.

A metric a team watches during an experiment or launch to make sure the change does not harm something important, such as speed, revenue or complaints.

To review incoming bugs, requests or issues quickly and sort them by urgency and importance, deciding which to handle now, later or not at all.

The practice of a company using its own product internally, often before release, to find problems and understand the user experience firsthand.

A team estimation technique where participants reveal estimates together and discuss differences before estimating again.

A minimal working implementation that connects the main architectural parts through a small end-to-end path.

A collaborative practice that brings business, development and testing perspectives together to clarify expected behavior.

Short for application programming interface: a set of rules that lets one program ask another program for data or actions in a fixed, documented way.

A specific URL of an API that accepts requests for one kind of action, such as GET /users to list users or POST /orders to create an order.

Short for representational state transfer: a common style for web APIs in which resources have URLs and are handled with HTTP methods such as GET, POST, PUT and DELETE.

Short for JavaScript Object Notation: a simple text format for data, made of key-value pairs and lists, used by most web APIs.

A three-digit number in an HTTP response that tells the client what happened, such as 200 for success, 404 for not found or 500 for a server error.

The actual data carried by a request, response or message, as opposed to the headers and other information around it.

The process of checking who a user or system is, for example with a password, a token or a fingerprint, before giving access.

Made or provided by an outside company or developer, not by your own team, such as a payment service, library or analytics tool.

Short for software development kit: a package of libraries, tools and documentation that helps developers build on a platform or use a service.

A change to an API, library or file format that makes existing code that uses it stop working unless that code is updated too.

A message that a client sends to a server or API, asking it to do something such as return data or save a new record.

The message a server or API sends back after receiving a request, usually containing a status code, headers and data such as JSON.

A unique secret code that a program sends with its requests so that the API can identify who is calling and decide what they are allowed to do.

A string of characters that proves a user or app has already logged in or has permission. It is sent with each request instead of a password.

The process of checking what an identified user or app is allowed to do, for example whether they can read a file or delete an account.

A way for one system to automatically send an HTTP request to another system's URL when an event happens, such as a successful payment, instead of waiting to be asked.

A cap on how many requests a client may send to an API in a given time, for example 100 requests per minute, to protect the service from overload or abuse.

Still working with older versions, data or clients after a change, so existing users do not have to update their code or settings.

Hypertext Transfer Protocol: the set of rules that browsers and servers use to send requests and responses across the web.

The secure version of HTTP, which encrypts the connection between the browser and the server with TLS so that others cannot read or change the data.

To request and receive data from a server or API, for example the latest messages or a list of products, often while the page is already open.

The word at the start of an HTTP request that says what the client wants to do, such as GET to read, POST to create or DELETE to remove.

A name and value pair sent with an HTTP request or response that carries extra information, such as Content-Type, Authorization or Cache-Control.

An attack in which someone gets harmful JavaScript into a page that other users open, for example through a comment, so that the script runs in their browsers.

An attack in which someone types SQL code into a form or URL so that the database runs it, which can expose or delete data.

Cross-Origin Resource Sharing: browser rules that let a server say which other websites (origins) may read its responses through JavaScript.

Cross-Site Request Forgery: an attack that tricks a logged-in user's browser into sending an unwanted request, such as changing an email address, to a site the user trusts.

To turn data such as a password into a fixed-length string with a one-way function, so that the original value cannot be read back from it.

Short for identity and access management: the system that controls who or what can sign in to a cloud account and what they are allowed to do there.

The security principle that every user, service or program should get only the permissions it really needs for its job, and nothing more.

A sensitive value that a system needs to work, such as a password, API key, token or private key, which must be stored safely and never shown in code or logs.

A virtual firewall in the cloud with rules that say which network traffic, by port and source, may reach or leave a server or other resource.

Turning data into an unreadable form that can only be turned back with the right key, so that people without the key cannot read it.

The idea that security in the cloud is split: the provider protects the data centers, hardware and core services, and the customer protects their own data, settings and access.

A short message sent from a server to a user's device that appears on the screen even when the app is not open, for example about a new message or a delivery.

The user's approval that an app needs before it can use sensitive features or data, such as the camera, location, contacts or notifications.

A link that opens a specific screen inside an app, such as a product page or an order, instead of just starting the app on its home screen.

A notification that the app schedules on the device itself, for example a reminder at 9 a.m., without needing a server to send it.

Apple Push Notification service: Apple's system that delivers push notifications from a server to iPhones, iPads and other Apple devices.

Firebase Cloud Messaging: Google's free service for sending push notifications and data messages to Android, iOS and web apps.

A unique string that identifies one installation of an app on one device; the server needs it to send push notifications to that device.

A small number or dot shown on an app icon or a tab, usually in red, that tells users how many unread items or new notifications are waiting.

To choose actively to take part in something, such as receiving notifications or sharing data; the opposite is to opt out.

Apple's encrypted storage for small secrets such as passwords, tokens and keys, protected by the system and kept separate from normal app data.

Confirming a user's identity with a physical feature, such as a fingerprint or a face scan, instead of or in addition to a password or PIN.

Anything that could cause harm to a system or to data, such as an attacker, a piece of malware, a careless employee or even a flood in the server room.

A weakness in software, hardware, a configuration or a process that an attacker could use to cause harm or to gain access they should not have.

Short for 'malicious software': any program written to damage systems, steal data, spy on users or take control of a device without permission.

A trick in which attackers send fake emails, messages or websites that look trustworthy, hoping people will click a link, open a file or reveal their passwords.

Malware that encrypts a victim's files or systems and demands a payment, usually in cryptocurrency, in exchange for the key to unlock them.

A secret string of characters that a person types to prove their identity and log in to an account, device or system.

A login method that asks for two or more different kinds of proof, such as a password plus a code from a phone, before granting access.

A hardware device or piece of software that checks network traffic and allows or blocks it according to a set of security rules.

A security incident in which personal or confidential data is accessed, stolen, disclosed or lost without authorisation.

The chance that a threat will use a weakness and cause harm, combined with how serious that harm would be for the organisation.

A piece of code or a technique that takes advantage of a vulnerability to make a system do something it was not designed to do.

To get unauthorised access to or control of a system, account or network, so that it can no longer be trusted.

Virtual private network: a technology that creates an encrypted connection over the internet so that remote users or offices can reach a private network safely.

An authorised, simulated attack on a system, application or network, carried out by security specialists to find weaknesses before real attackers do.

The state of following the laws, regulations, standards and internal policies that apply to an organisation, and being able to prove it.

The rules and technical measures that decide who or what can use a system, view data or enter a place, and what they are allowed to do there.

Manipulating people, rather than breaking technology, to make them reveal information, give access or take an action that helps an attacker.

Distributed denial of service: an attack that floods a website or network with traffic from many machines at once so that real users cannot reach it.

Malware disguised as a useful or harmless program, such as a free tool or a document, that secretly performs harmful actions once it is run.

Malware that copies itself and spreads from one computer to another across a network without any action from the user.

Malware that secretly watches what a user does and collects information such as browsing habits, messages, passwords or location.

To turn encrypted data back into its original, readable form, normally by using the correct key.

A secret value, usually a long string of bits, that an encryption algorithm uses to lock and unlock data. Without it, the encrypted data cannot be read.

An electronic document that links a public key to an identity, such as a website or a person, and is signed by a trusted certificate authority.

Transport Layer Security: the standard protocol that encrypts data travelling between two systems, such as a browser and a website, and checks the server's identity.

A tool that monitors network traffic or system activity for signs of attacks or policy violations and raises an alert when it finds them.

A vulnerability that is unknown to the vendor or has no patch yet, so defenders have had 'zero days' to fix it when it is first exploited.

Common Vulnerabilities and Exposures: a public catalogue that gives each publicly known vulnerability a unique ID, such as CVE-2021-44228.

An automated check of systems, networks or applications that looks for known weaknesses, missing patches and insecure settings and lists them in a report.

Security information and event management: a platform that collects logs from many systems, correlates them and raises alerts about possible attacks.

An alert or test result that wrongly reports a threat or vulnerability when nothing harmful is actually there.

The stage of incident response in which the team stops an attack from spreading or causing more damage, for example by isolating infected machines.

The work of fixing a security problem at its source, such as applying a patch, changing a configuration or rewriting vulnerable code.

To reduce the likelihood or impact of a risk or attack when it cannot be removed completely, for example with a workaround or an extra control.

A formal document, approved by management, that sets the rules an organisation follows to protect its information, systems and people.

All the points where an attacker could try to get into a system or organisation or take data out, such as open ports, web apps, user accounts and suppliers.

An attack that tries a huge number of passwords or keys one after another, usually automatically, until the right one is found.

The details that prove who a user is and let them log in, most often a username and password, but also keys, tokens and certificates.

A login system that lets a user sign in once with one set of credentials and then use many different applications without logging in again.

A way of managing permissions in which rights are given to roles, such as 'accountant' or 'admin', and users get access by being assigned to a role.

A type of cryptography that uses a pair of related keys: a public key that can be shared with anyone and a private key that its owner keeps secret.

Encryption in which only the sender and the intended recipient can read the messages, so even the service provider in the middle cannot see their content.

Dividing a network into separate zones, with controlled traffic between them, so that a problem in one part cannot easily spread to the others.

A security model that trusts no user or device by default, even inside the company network, and checks identity and context for every request.

Endpoint detection and response: security software on laptops and servers that records activity, detects suspicious behaviour and lets analysts investigate and respond remotely.

Collected and analysed information about attackers, their methods and their tools that helps an organisation prepare for and detect attacks.

A piece of technical evidence that suggests a system may have been attacked, such as a malicious IP address, file hash or domain name.

The collection and analysis of digital evidence, such as disk images, memory and logs, to find out what happened during an incident in a way that can stand up in court.

A written, step-by-step guide that tells a security team how to handle a specific type of alert or incident, such as phishing or ransomware.

To pass an alert, incident or problem to a more senior person or team because it is serious or beyond your own authority or skills.

The ongoing process of finding, testing, approving and installing software updates across all systems, and tracking which ones are still missing.

Common Vulnerability Scoring System: a standard method for rating how severe a vulnerability is, with a score from 0.0 to 10.0.

A security weakness caused by settings that are wrong, insecure or left at their defaults, rather than by a bug in the software itself.

A formal, independent review that checks whether an organisation's security controls and processes meet a standard, a law or its own policies.

A structured process for identifying risks, estimating how likely and how serious they are, and deciding which ones need action first.

The agreed conditions of a security test: what may be tested, when, with which methods and whom to contact if something goes wrong.

A security problem or weakness that is identified and documented during an audit, assessment or penetration test, usually with a severity rating.

The first stage of an attack or security test, in which information about the target is gathered, such as domain names, staff, technologies and exposed services.

Gaining higher access rights than one was given, for example moving from a normal user account to administrator, usually by abusing a flaw or misconfiguration.

The way attackers move from the first compromised machine to other systems inside the same network, looking for valuable data or higher privileges.

An attack in which someone secretly places themselves between two communicating parties so that they can read or change the traffic passing between them.

Software or a small hardware device that secretly records every key a user presses, often to capture passwords and private messages.

A network of many infected computers or devices, called bots, that an attacker controls remotely and uses together, for example to send spam or run DDoS attacks.

A hidden way into a system that bypasses normal authentication, either left by developers or installed by attackers so they can return later.

An isolated, controlled environment where suspicious files or programs can be run safely to see what they do, without any risk to real systems.

A firewall that sits in front of web applications and filters HTTP traffic to block common web attacks such as SQL injection and cross-site scripting.

A network port on a device that is accepting connections, which means a service is listening there and can be reached by others.

A cryptographic value attached to a message, document or file that proves who created it and shows that it has not been changed since it was signed.

Data after it has been encrypted, which looks like random characters and cannot be understood without the correct key. The readable original is called plaintext.

Actively searching through systems and logs for signs of attackers who have not yet been detected by automated alerts, often starting from a hypothesis.

Any safeguard, whether technical, administrative or physical, that reduces a security risk, such as a firewall, a policy, a training course or a locked door.

General Data Protection Regulation: the European Union law, applied since May 2018, that sets rules for how organisations collect, use and protect personal data.

The process of making a system more secure by removing unnecessary software and services, changing default settings and applying secure configurations.

Reporting a vulnerability privately to the vendor or owner first and giving them reasonable time to fix it before any details are made public.

A programme in which an organisation invites outside researchers to find and report vulnerabilities in its systems and pays rewards for valid reports.

A person or group responsible for a cyberattack or threat, such as a criminal gang, a state-backed group, a hacktivist or a malicious insider.

Faking an identity or source, such as an email sender, a phone number, a website or an IP address, so that a message or connection seems to come from someone trusted.

A group of computers, phones and other devices that are connected by cables or radio signals so that they can exchange data and share resources.

A device that connects different networks and decides, packet by packet, the best path for data to reach its destination, usually based on IP addresses.

A device with many ports that connects computers, phones, printers and other equipment inside the same local network and forwards each frame only to the port where its destination is.

A unique numerical label given to every device on an IP network so that data can be sent to it, for example 192.168.1.20 in IPv4.

The maximum amount of data a connection can carry in a given time, usually measured in megabits or gigabits per second (Mbps, Gbps).

A set of agreed rules that defines how devices format, send, receive and check data, so that equipment from different vendors can talk to each other.

Short for local area network: a network that connects devices in a limited area such as a home, an office floor or a single building.

Short for wide area network: a network that links sites across cities or countries, usually over lines leased from a telecom operator or over the internet.

The common name for wireless local networks based on the IEEE 802.11 standards, which connect devices to a network by radio, mainly in the 2.4, 5 and 6 GHz bands.

Short for Dynamic Host Configuration Protocol: a protocol that automatically gives devices an IP address, subnet mask, default gateway and DNS servers when they join a network.

The router address that a device sends traffic to when the destination is outside its own subnet; without it, a device can only talk to its local network.

A small unit of data sent across a network, made of a header with addressing and control information and a payload that carries part of the actual content.

Either a physical socket on a device where a network cable is plugged in, or a number from 0 to 65535 that identifies a particular service on a host, such as 443 for HTTPS.

The family of protocols that the internet and almost all modern networks are built on, named after its two best-known members, TCP and IP.

The standard technology (IEEE 802.3) for wired local networks, which defines how devices send frames over copper or fibre cables at speeds from 10 Mbps up to hundreds of gigabits per second.

To set up a device or service by entering the settings it needs, such as addresses, passwords, VLANs and routing, so that it works the way the network requires.

A hardware address, usually set in a network card at the factory, that identifies a device on a local network; it has 48 bits and is written as six pairs of hexadecimal digits.

The process of choosing the path that traffic takes from one network to another, carried out by routers using routes that are either entered by hand or learned from other routers.

A list stored in a router or host that shows which networks it knows about and where to send packets for each one, such as the next router or an outgoing interface.

Short for virtual LAN: a separate logical network created on a switch, so that devices plugged into the same hardware, for example staff PCs, guests and cameras, can be kept apart.

A device, usually mounted on a ceiling or wall, that sends and receives Wi-Fi signals and connects wireless devices to the wired network.

A situation in which some packets sent across a network never reach their destination, usually shown as a percentage; it causes slow downloads, choppy calls and frozen video.

A reference model that splits network communication into seven layers, from the physical cable (layer 1) up to the application (layer 7), so that problems and technologies can be described clearly.

Short for Transmission Control Protocol: a protocol that sets up a connection and makes sure data arrives complete and in the right order, resending anything that is lost.

Short for User Datagram Protocol: a simple, fast protocol that sends data without setting up a connection and without checking whether it arrived; used for voice, video, DNS and gaming.

Version 4 of the Internet Protocol, which uses 32-bit addresses written as four numbers from 0 to 255 separated by dots, such as 192.168.10.5; it allows about 4.3 billion addresses.

Version 6 of the Internet Protocol, which uses 128-bit addresses written as groups of hexadecimal digits separated by colons, such as 2001:db8::1, giving an almost unlimited number of addresses.

A number such as 255.255.255.0 that shows which part of an IPv4 address identifies the network and which part identifies the device; it can also be written as a prefix such as /24.

Short for network address translation: a method in which a router or firewall rewrites the addresses in packets, typically so that many devices with private addresses can share one public IP address.

An IP address that is set by hand or permanently reserved for one device, so that it does not change, unlike a dynamic address handed out by DHCP.

The layout of a network: how its devices and links are arranged and connected, for example in a star, ring or mesh shape.

Any device on a network that has an IP address and sends or receives data, such as a computer, server, phone or printer.

Short for service set identifier: the name of a Wi-Fi network, up to 32 characters long, that devices see in their list of available networks.

A diagnostic tool that lists every router a packet passes through on its way to a destination, with the delay to each one, so you can see where traffic slows down or stops.

Variation in the delay of packets arriving over a network; when packets arrive at uneven intervals, voice and video calls become choppy or robotic.

Short for quality of service: a set of techniques that give important traffic, such as voice and video, priority over less urgent traffic when a link is busy.

Short for Border Gateway Protocol: the routing protocol that networks on the internet use to exchange routes with each other, so that traffic can find its way between different operators and companies.

Short for Open Shortest Path First: a routing protocol used inside one organisation's network, where routers share information about their links and each one calculates the best path itself.

A route that an engineer enters into a router by hand, telling it exactly where to send traffic for a particular network, instead of letting a routing protocol learn it.

A switch port that carries traffic for several VLANs over one link, adding a tag to each frame so the device at the other end knows which VLAN it belongs to.

A protocol that stops switches from forming loops by blocking extra paths between them, and opens a blocked path again if an active link fails.

An agreed period, usually at night or at the weekend, when engineers are allowed to make changes that may interrupt service, such as software upgrades and hardware swaps.

Using a mobile network other than your own operator's, usually abroad; in Wi-Fi, moving from one access point to another without losing the connection.

A recording of the actual packets passing through a network interface, saved to a file so that engineers can analyse exactly what was sent and received.

Short for Simple Network Management Protocol: a protocol that lets monitoring systems collect information such as traffic, errors and temperature from network devices, and lets devices send alerts.

Continuously checking the health and performance of network devices and links, such as availability, traffic, errors and latency, and alerting engineers when something goes wrong.

A situation in which more traffic is trying to cross a link or device than it can handle, so packets are queued or dropped and everything slows down.

The time it takes for a packet to travel from a source to a destination and for the reply to come back, usually shown in milliseconds; ping reports this value.

Short for access control list: an ordered list of rules on a router, switch or firewall that permits or denies traffic based on things like source, destination, protocol and port.

A NAT rule that sends traffic arriving on a public IP address and port to a specific device and port inside a private network, so that an internal service can be reached from outside.

Short for demilitarised zone: a separate network segment between the internet and the internal network, where public-facing servers such as web or mail servers are placed.

Short for Multiprotocol Label Switching: a technique used in operator networks that forwards traffic using short labels instead of full IP lookups; it is widely sold to companies as a private WAN service.

Short for software-defined wide area network: an approach that manages a company's WAN links centrally through software and sends each application over the best available link, such as internet, 4G or MPLS.

One step in the path of a packet, from one router to the next; the number of hops shows how many routers a packet passes through to reach its destination.

Short for Address Resolution Protocol: the protocol a device uses on a local network to find the MAC address that belongs to a known IPv4 address.

Short for Internet Control Message Protocol: a protocol that network devices use to send error and status messages, such as 'destination unreachable'; ping and traceroute rely on it.

A message sent to every device on the local network segment at once, instead of to a single address; routers do not forward broadcasts to other networks.

Short for Classless Inter-Domain Routing: the system of writing an IP network as an address plus a prefix length, such as 192.168.0.0/22, which replaced the old fixed address classes.

An IP address from the ranges reserved for internal networks, such as 10.0.0.0/8, 172.16.0.0/12 and 192.168.0.0/16, which cannot be routed on the public internet.

Short for network operations centre: the team and room, often staffed 24/7, that monitors a network, responds to alarms and coordinates the repair of faults and outages.

Describes a game or mode in which several people play together or against each other, online or on the same device. As a noun, it means the multiplayer part of a game.

A computer, usually in a data center, that runs only the game server and the shared game world, while players connect to it instead of one player hosting the match.

The networking code of an online game that sends game data between players and servers and keeps everyone's view of the match in sync despite delays.

The system that puts online players into matches with or against each other, based on things like skill level, region and connection quality.

A noticeable delay between a player's action and the game's reaction, usually caused by a slow or unstable network connection, which makes the game feel jumpy.

How many times per second a game server updates the game state and sends it to players, measured in hertz. A higher tick rate makes online play more precise.

A virtual waiting area where players gather before an online match starts, to form teams, choose characters or settings, and wait for others to join.

A netcode technique in which the player's own device shows the result of their input immediately, before the server confirms it, and corrects it later if the server disagrees.

A server that has the final say on what happens in an online game: players send their inputs, and the server decides the real results, which makes cheating harder.

An infrastructure layer that manages service-to-service communication, often adding routing, security, and monitoring features.

A component between clients and services that routes requests and may handle security, limits, logging, or protocol conversion.

The opening TLS exchange that negotiates security settings and session keys, usually authenticating the server and sometimes the client.

An unpredictable value included in a request and checked by the server to help reject cross-site request forgery.

A server or software layer that receives a request and passes it to another service, often for routing, filtering, or security.

The process of asking the Domain Name System to translate a domain name into an IP address or related record.

Selecting a representation of a resource, such as its format or language, using information exchanged between client and server.

A retry strategy that increases the waiting interval by a multiplying factor after successive failures, often up to a limit.

A separate queue that holds messages which could not be processed normally, for later inspection, correction or retry.

An identifier carried across related operations or messages so their records can be connected during investigation.

An attack that resends a previously valid message to make a system accept an action that is no longer authorized or intended.

A value intended to be used only once within a defined protocol or cryptographic context.

Restricting a client's accepted server certificates or public keys beyond the usual certificate trust checks.

A TLS connection in which both client and server authenticate themselves using certificates.

To put a new version of software onto servers or devices so that it starts running there, for example in a test environment or for real users.

A version of a product that is made available to users, or the act of making it available. Also a verb: to release a new version.

The live environment where real users use the software, as opposed to development or test environments.

The process of turning source code into a program or package that can run, or the result of that process. Also a verb: to build the project.

To release a product or feature so that real users can use it. The word is borrowed from the idea of sending goods to customers.

Written material that explains how software works and how to use, set up or change it, such as guides, API references and README files.

Short for continuous integration and continuous delivery (or deployment): automatically building and testing every change, then releasing it quickly and safely.

An urgent fix for a serious problem in production, released quickly outside the normal release schedule.

To return a system to an earlier working version after a release or change causes problems.

An environment set up to be as close as possible to production, where a team checks a release one last time before real users get it.

Old code that is still in use, often written by people who have left, with little documentation or few tests, so it feels risky to change.

To officially mark a feature, function or API as outdated and not recommended, usually as a warning that it will be removed in a later version.

A short document published with a new version of software that lists new features, bug fixes, known issues and other changes users should know about.

A small update that fixes bugs or security problems in software without adding big new features. It is also a verb: to patch a system.

To move to a newer, usually better version of software, a library or a tool, for example from Python 3.10 to Python 3.12.

To become available to real users for the first time, for example when a new system, feature or website is launched in production.

A file or page that lists the notable changes in each version of a project, usually with the newest version first. It is often named CHANGELOG.md.

A version-numbering rule with three parts, MAJOR.MINOR.PATCH: the first number changes for breaking changes, the second for new features and the third for bug fixes.

A period before a release, or during busy times such as holidays, when no new code changes may be merged except critical fixes.

To raise a version number or update a dependency to a newer version, usually as a small change. Informally, also to raise a task's priority or remind people about a message.

The readable address of a website, such as example.com, which people type instead of a numeric IP address.

A service that keeps a website's files and code on servers connected to the internet, so that people can visit the site at any time.

A digital file installed on a server that proves the site's identity and lets the browser set up an encrypted HTTPS connection.

Domain Name System: the internet's directory that translates domain names like example.com into the IP addresses of servers.

Software, such as Nginx or Apache, that accepts HTTP requests and serves web pages and files; the term is also used for the machine that runs it.

Renting servers, storage, databases and other computing resources over the internet from a provider, instead of buying and running your own hardware.

A way of working in which developers and operations engineers share responsibility for building, releasing and running software, using automation to deliver changes quickly and safely.

An automated series of steps, such as building, testing and deploying, that runs every time code changes and stops if one step fails.

A lightweight, isolated package that holds an application together with everything it needs to run, so it behaves the same on a laptop, a test server or in the cloud.

A widely used platform for building container images and running containers. It made containers popular after its release in 2013.

An open-source system that runs and manages containers across many machines: it starts them, restarts failed ones, spreads traffic and scales them up or down.

A complete computer, with its own operating system, that runs as software on a physical machine. One physical server can host many virtual machines at once.

One running copy of a virtual server or service in the cloud. You choose its size, such as how many CPUs and how much memory it has, and pay while it runs.

A company that owns large data centers and rents computing, storage, networking and other services to customers over the internet.

A geographic area where a cloud provider has a group of data centers, such as Frankfurt or Ireland. You choose the region where your resources run.

A separate copy of the whole system, with its own servers, settings and data, used for one purpose such as development, testing or serving real users.

Managing servers, networks and other cloud resources by writing configuration files that tools read and apply, instead of clicking through a console by hand.

Short for infrastructure as a service: renting basic building blocks such as virtual machines, disks and networks, while you manage the operating system and everything above it.

Short for platform as a service: a ready-made platform where you only upload your code, and the provider runs the servers, operating system and runtime for you.

Short for software as a service: complete software that customers use through a browser or app and pay for by subscription, while the vendor runs everything behind it.

Describes a way of running code in which the provider starts and scales servers automatically for each request, and you pay only for the time your code actually runs.

Describes servers and software that a company runs in its own building or data center, rather than renting them from a cloud provider.

A setup that combines a company's own on-premises systems or private cloud with a public cloud provider, with the two connected and working together.

One of several separate data center locations inside a cloud region, each with its own power and network, so a failure in one should not affect the others.

A read-only package with an application, its libraries and settings, used as the template from which containers are started.

A group of connected machines that work together as one system, for example to run containers in Kubernetes or to store data in a database.

One machine, physical or virtual, that is part of a cluster. In Kubernetes, worker nodes are the machines where pods actually run.

The smallest unit that Kubernetes runs: one or more containers that are always placed together on the same node and share a network address and storage.

The automatic management of many containers or services: deciding where they run, starting and restarting them, connecting them and scaling them.

The practice of merging small code changes into the shared main branch often, with each change automatically built and tested so problems are found early.

The practice of keeping code always ready to release, so that any tested version can go to production at the push of a button.

A file or package produced by a build, such as a JAR file, a binary or a container image, which is stored and later deployed.

A service that receives incoming traffic and spreads it across several servers, sending requests only to the ones that are healthy.

Automatically adding servers or containers when demand rises and removing them when it falls, based on rules such as CPU use or number of requests.

The time a request or piece of data takes to travel through a system and get a response, usually measured in milliseconds.

To create and set up computing resources, such as servers, databases or networks, so they are ready to use.

A widely used infrastructure-as-code tool from HashiCorp. You describe resources in configuration files, and it creates, changes or deletes them to match.

Using tools to keep the software, settings and files on many servers in a known, consistent state, instead of changing each server by hand.

To quickly create and start a new server, container or environment, usually with a command or a few clicks.

Software that creates and runs virtual machines, sharing one physical computer's CPU, memory and storage between them and keeping them separate.

An application, service or job that runs on cloud resources and uses computing power, for example a website, a batch job or a database.

A cloud service where the provider handles setup, patching, backups and scaling, so you only use it and do not manage the servers underneath.

A type of cloud storage that keeps files as objects, each with a unique key and metadata, and lets you read and write them over an HTTP API.

A named container in object storage, such as Amazon S3 or Google Cloud Storage, that holds files and has its own access and storage settings.

Short for virtual private cloud: your own isolated private network inside a public cloud, where you control IP ranges, subnets, routing and access.

A smaller range of IP addresses inside a larger network such as a VPC, used to group resources and control which of them can reach the internet.

Describes a pricing model in which you pay only for the resources you actually use, for example by the second, hour or gigabyte, with no long-term contract.

Reducing cloud spending without hurting performance or reliability, for example by removing unused resources, choosing the right sizes and using discounts.

A copy of a disk, volume or database exactly as it was at one moment in time, which can be used to restore it or create a new copy.

To handle more load by adding more servers, containers or instances that share the work, rather than making one machine bigger.

To handle more load by giving one machine more power, such as more CPU or memory, instead of adding more machines.

The amount of work a system completes in a given time, for example requests per second, messages per minute or megabytes per second.

The one slowest or most limited part of a system, such as the database or the network, that holds back the performance of everything else.

A release method with two identical production environments: the new version is put on the idle one, then all traffic is switched to it at once.

Releasing a new version to a small share of users or servers first, watching its metrics, and only then rolling it out to everyone.

A setting that turns a feature on or off without deploying new code, so it can be shown to some users, tested in production or switched off quickly.

To release a new version, feature or change to users or servers, often step by step over a period of time rather than all at once.

A text file with step-by-step instructions, such as FROM, COPY and RUN, that Docker follows to build a container image.

A storage service for container images, where pipelines push new images and servers or clusters pull them when they need to run them.

A human-readable text format for configuration files that uses indentation and simple 'key: value' lines. Most DevOps tools are configured with it.

Cloud capacity you commit to use for one or three years in exchange for a large discount compared with normal on-demand prices.

Spare cloud capacity sold at a much lower price, which the provider can take back at short notice when it needs the capacity again.

Data leaving a network, for example from a cloud provider to the internet or to another region. Cloud providers usually charge for egress but not for incoming data.

To delete a server, environment or set of resources completely when it is no longer needed, usually with a command or script.

Testing a nearly finished version of an app with a group of real users before the public release, to find bugs and collect feedback.

The small image that represents an app on the home screen, in the store and in the settings; it is often the first thing users notice.

To refuse to accept an app or update into the store because it breaks a store rule or does not work properly during review.

Apple's service for sending beta builds of iOS apps to testers, who install them with the TestFlight app before the version is released on the App Store.

The page for an app in an app store, with its name, icon, description, screenshots, ratings and other information that helps users decide to download it.

The number that identifies a release of an app for users, such as 3.2.1, shown in the store and usually in the app's settings.

Adding a digital signature to an app with a certificate, which proves who made it and shows that it has not been changed since it was built.

An Apple file that connects an app ID, signing certificates and, for testing, specific devices, so that a signed iOS app is allowed to run or be distributed.

A unique identifier for an app, usually written in reverse domain style such as com.company.appname, which the system and the store use to recognize it.

Releasing an update to a small percentage of users first and increasing it step by step, so problems can be found before everyone gets the new version.

A screen or rule that blocks an old version of an app and makes users install the latest one before they can continue using it.

App Store Optimization: improving an app's store page, such as its title, keywords, screenshots and ratings, so it ranks higher in search and gets more downloads.

Short for downloadable content: extra content released for a game after it comes out, such as new levels, characters, story chapters or costumes, often sold separately.

The hardware or store that a game runs on and is sold through, such as PC, PlayStation, Xbox, Nintendo Switch, iOS or Android, or a store like Steam.

Adapting a game for players who speak another language and live in another culture: translating text and voice, and adjusting things like fonts, dates, jokes and images.

A model in which a game is updated continuously for years after launch with new content, events and seasons, to keep players playing and spending.

Short for daily active users: the number of different people who play a game on a given day. MAU is the same figure for a month.

A software release focused on fixes, minor improvements, and stability rather than major new features.

A prepared set of steps for restoring the previous stable version when a deployment or release causes serious problems.

A nearly final software build that should become the public release unless testing finds an important remaining problem.

To move a fix or feature from a newer code version into an older maintained branch that still receives updates.

The amount of unreliability allowed within a measurement period by a service-level objective.

A difference that develops between a system's actual configuration and its intended or recorded configuration.

An approach where deployed infrastructure instances are replaced with new versions rather than changed in place.

An update that replaces or upgrades groups of running instances in sequence while other instances continue serving.

In Kubernetes, a check used to determine whether a container is ready to receive service traffic.

In Kubernetes, a check used to determine whether a running container needs to be restarted.

Stopping a process in a controlled way that stops new work and finishes or safely releases existing work before exit.

An inventory of the components included in a software product, often recording versions and dependency relationships.

Structured Query Language: the standard language used to create tables and to read, add, change and delete data in relational databases.

A structure in a relational database that stores data about one kind of thing in rows and columns, for example all customers or all orders.

A vertical part of a table with a name and a data type, holding one kind of information for every row, such as email or created_at.

A request sent to a database to read or change data, usually written in SQL. As a verb, to query means to ask a database for data.

A type of database that stores data in tables made of rows and columns and links tables to each other through keys. Data is usually managed with SQL.

Database management system: the software that stores the data, runs queries, controls access and keeps data safe, such as PostgreSQL or Oracle.

A column, or a set of columns, whose value uniquely identifies each row in a table. It cannot be empty and no two rows can share it.

A column in one table that points to the primary key of another table, linking the two and stopping you from storing references to rows that do not exist.

The SQL statement that reads data from one or more tables and returns the matching rows without changing anything in the database.

The part of a SQL statement that sets a condition, so only the rows that meet it are read, updated or deleted.

A SQL operation that combines rows from two or more tables using a related column, such as a customer ID. As a verb, to join tables means to combine them this way.

The structure of a database: its tables, columns, data types, keys and relationships. In some products it is also a named group of tables inside a database.

A group of database operations that are treated as one unit: either all of them succeed together or none of them is applied.

The kind of value a column can hold, such as whole numbers, text, dates or true/false values. It controls what can be stored and how much space it takes.

A special marker meaning that a value is missing or unknown. It is not the same as zero or an empty string.

The SQL statement that adds one or more new rows to a table, using INSERT INTO followed by the column names and their values.

The SQL statement that changes values in rows that already exist, usually limited by a WHERE clause so that only the right rows are changed.

The SQL statement that removes rows from a table. With a WHERE clause it removes only matching rows; without one it removes every row.

A copy of a database saved at a certain moment so that the data can be brought back after a failure, a mistake or an attack.

A broad group of databases that do not use the classic table-and-SQL model, storing data as documents, key-value pairs, wide columns or graphs instead.

A rule set on a table or column that the database enforces automatically, rejecting any data that would break it.

The process of organizing tables so that each fact is stored in only one place, which reduces repeated data and prevents update mistakes.

The SQL clause that sorts the rows of a result by one or more columns, in ascending (ASC) or descending (DESC) order.

The SQL clause that puts rows with the same values into groups so that a total, count or average can be calculated for each group.

A SQL function that takes many rows and returns a single value from them, such as COUNT, SUM, AVG, MIN or MAX.

A join that returns only the rows that have a match in both tables; rows without a partner on the other side are left out.

A join that returns all rows from the first (left) table and the matching rows from the second; where there is no match, the second table's columns are NULL.

The SQL statement that makes a new table, defining its name, its columns with their data types and its constraints such as the primary key.

The SQL statement that changes the structure of an existing table, for example to add, rename or drop a column or to add a constraint.

A saved SELECT query that you can use like a table. It usually stores no data of its own and shows the current result of the query each time.

A named block of SQL code saved inside the database that can take parameters and be run again and again with a single call.

Atomicity, Consistency, Isolation, Durability: the four guarantees that make database transactions reliable even with errors, crashes and many users at once.

The step-by-step strategy the database chooses to run a query, showing which indexes it uses, how tables are joined and how many rows it expects.

The part of a database engine that looks at a query and the table statistics and picks the cheapest way to run it, producing the execution plan.

Keeping copies of the same database on several servers by sending every change from one server to the others, for safety and to share the load.

To bring a database back from a backup, either to fix data loss or to create a copy of it on another server.

Database administrator: the person responsible for installing, securing, backing up, tuning and keeping a company's databases running.

A drawing of a database design that shows the entities (usually tables), their attributes and the relationships between them, such as one-to-many.

A link between two tables where one row in the first table can be connected to many rows in the second, but each of those belongs to only one, like a customer and their orders.

A link where many rows in one table can relate to many rows in another, like students and courses. It needs a third table in between to store the pairs.

A rule that stops two rows from having the same value in a column or combination of columns, for example two users with the same email address.

Describes a numeric column whose value the database fills in automatically with the next number in sequence each time a new row is added.

A NoSQL database that stores each item as a self-contained document, usually in a JSON-like format, so related data can be kept together without fixed tables.

The simplest kind of NoSQL database, which saves each value under a unique key and finds it again very quickly by that key.

A query written inside another query, in brackets, whose result is used by the outer query, for example as a list of values to filter by.

The part of a SQL query that filters groups after GROUP BY has run, usually using the result of an aggregate function such as COUNT or SUM.

A form of the SELECT statement that removes duplicate rows from the result, so each combination of returned values appears only once.

A temporary short name given to a table or column inside a query, usually with the AS keyword, to make the query shorter or the output clearer.

A key made of two or more columns that identify a row only when used together, such as order ID plus product ID in an order items table.

An artificial key with no business meaning, such as an auto-increment number or a UUID, created only to identify each row.

The rule that every reference between tables must point to a row that really exists, so there are no orders for missing customers.

Of a delete or update: to pass automatically from a parent row to all the child rows that point to it through a foreign key.

The value a column gets automatically when a new row is inserted without a value for that column, such as 0, 'pending' or the current time.

Deliberately storing some data in more than one place, or combining tables, to make reads faster, while accepting more work to keep the copies correct.

One of a series of rules for table design (1NF, 2NF, 3NF and higher); each level removes a particular kind of repeated or badly placed data.

Either the type of relationship between tables (one-to-one, one-to-many, many-to-many) or the number of distinct values in a column.

An index that decides the physical order in which a table's rows are stored, so a table can have only one; the other indexes are non-clustered.

Reading every row of a table to find the ones a query needs, instead of jumping straight to them through an index.

A query that takes longer than an agreed limit to finish, slowing down the application and using up database resources.

A setting that controls how much one transaction can see of other transactions' changes, trading data safety against speed and concurrency.

A mark the database places on a row, page or table so that other transactions cannot change it, or sometimes read it, until the current transaction ends.

A situation where two or more transactions each wait for a lock the other holds, so none of them can continue until the database cancels one.

A piece of code stored in the database that runs automatically when a certain event, such as an insert, update or delete, happens on a table.

A versioned script that changes a database schema in a controlled way, or, more broadly, a project to move data from one database or system to another.

Restoring a database to its exact state at a chosen moment, such as one minute before a bad DELETE, by replaying saved change logs on top of a backup.

A copy of a database that receives changes from the main server and is used only for read queries, taking load off the primary.

Switching work from a failed primary database server to a standby or replica so that the application can keep running.

A permission that allows a database user or role to do a specific action, such as reading a table, inserting rows or creating new objects.

To give a database user or role permission to do something, using the SQL GRANT statement; REVOKE takes the permission away again.

A set of open database connections that an application keeps ready and reuses, instead of opening a new, slow connection for every request.

Object-relational mapping: a library that lets developers work with database rows as objects in their programming language and generates the SQL for them.

Create, Read, Update, Delete: the four basic operations on stored data, matching the SQL statements INSERT, SELECT, UPDATE and DELETE.

Splitting one large dataset across several database servers, each holding a different part of the rows, so the system can grow beyond one machine.

Dividing one large table into smaller pieces, called partitions, by a rule such as date range or region, while it still looks like one table to queries.

Designing a database system so that it keeps working with very little downtime, even if a server, disk or data center fails.

A file that contains a database's structure and data written out as SQL statements or in a special export format, which can be loaded into another database.

A guarantee that, if no new changes are made, all copies of the data will become the same after a short time, although reads may briefly return old values.

The idea that when a network failure splits a distributed database, it must choose between consistency (same data everywhere) and availability (always answering).

A NoSQL database that stores data as nodes connected by relationships, making it fast to follow links such as 'friends of friends'.

In a document database such as MongoDB, a group of documents stored together, similar to a table in a relational database.

A database that keeps its data mainly in RAM rather than on disk, so reads and writes are extremely fast.

A central database built for reporting and analysis that collects historical data from many systems and is designed for large read queries rather than small updates.

Online transaction processing: database workloads made of many small, fast reads and writes from users, such as placing orders or updating profiles.

To insert a row if it does not exist yet, or update it if it already does, in a single operation. The word combines 'update' and 'insert'.

Common table expression: a named temporary result defined with WITH at the start of a query, which the main query can then use like a table.

A SQL function that calculates a value across a set of related rows, such as a running total or a rank, while still returning every individual row.

A SQL operator that stacks the results of two or more SELECT statements into one result, one under the other, removing duplicate rows.

To remove all rows from a table at once with TRUNCATE TABLE, keeping the table and its structure; it is much faster than deleting rows one by one.

Data definition language: the SQL statements that create, change and remove database objects, such as CREATE, ALTER and DROP.

Data manipulation language: the SQL statements that work with the data inside tables, mainly INSERT, UPDATE, DELETE and often SELECT.

The work of making slow queries faster by studying their execution plans and then adding indexes, rewriting SQL or changing database settings.

A named set of database privileges that can be given to many users at once, such as read_only or app_writer, to make permission management simpler.

A conceptual model that describes data as entities, their attributes, and the relationships between them in a database design.

A log in which a database records changes before writing the corresponding changes to durable data pages.

Stored results of a query that can be read directly and refreshed when underlying data changes.

Checking whether data changed after it was read before accepting an update, often using a version value.

Reading a change made by another transaction before that transaction commits it.

A transaction anomaly where repeating a query finds a changed set of matching rows because another transaction changed the data.

A concurrency error where a write based on older data overwrites another update without preserving its effect.

An index that contains all the column values a particular query needs, potentially avoiding separate table-row reads.

A pattern where one query loads a list and a separate extra query is issued for each of its N items.

The part of a website or web app that users see and interact with in the browser, built mainly with HTML, CSS and JavaScript.

The server side of a website or app that users do not see: it stores data, runs business logic and sends the front end what it needs.

Covering both the front end and the back end of a web application, from the user interface to the server and the database.

A program such as Chrome, Firefox, Safari or Edge that downloads web pages and turns their code into the screens people read and use.

HyperText Markup Language: the standard language for describing the structure and content of a web page, such as headings, paragraphs, links and images.

Cascading Style Sheets: the language that controls how HTML looks, including colors, fonts, spacing, layout and animations.

The programming language that runs in every web browser and makes pages interactive; with Node.js it also runs on servers.

Uniform Resource Locator: the full address of a page or file on the web, such as https://example.com/blog?page=2.

A computer or program that waits for requests from browsers and other clients and sends back pages, files or data.

The side of a connection that asks for data, usually the browser or a mobile app, as opposed to the server that answers.

A single building block of an HTML page, such as a paragraph, image or button, usually written as an opening tag, content and a closing tag.

A piece of HTML code in angle brackets, such as <h1> or </p>, that marks where an element starts or ends.

Extra information written inside an opening HTML tag as a name and a value, such as href in a link or alt in an image.

The part of a CSS rule that says which elements the styles apply to, for example p, .button or #header.

Document Object Model: the tree of objects a browser builds from a page's HTML, which JavaScript can read and change while the page is open.

An approach to building pages so that the same layout adapts to any screen size, from phones and tablets to large monitors.

The practice of making websites usable by everyone, including people who are blind or deaf, have limited movement or use assistive technology.

An application that runs in a web browser instead of being installed on the device, such as online email, maps or a project management tool.

Tools built into browsers that let developers inspect HTML and CSS, debug JavaScript, watch network requests and measure performance.

In CSS, the name of a style feature you set, such as color, margin or display, which is always paired with a value.

A file or block of CSS rules that describes how a page should look; most sites link to one or more external .css files.

The way elements are arranged on a page: where columns, menus, images and text sit and how they move when the screen size changes.

A CSS layout method for arranging items in a single row or column and controlling how they grow, shrink, align and share space.

A CSS layout system for placing elements in rows and columns at the same time, useful for full page layouts and card galleries.

A CSS rule that applies styles only when a condition is true, such as a minimum screen width, dark mode or printing.

A reusable, self-contained piece of a user interface, such as a button, modal or product card, with its own markup, styles and logic.

Something that happens in the browser and that code can react to, such as a click, a key press, a form submission or the page finishing loading.

An organized store of data, such as users, orders or posts, that a back end can save to, search and update quickly.

A small piece of data that a website stores in the browser and that is sent back with later requests, used for logins, settings and tracking.

How long a page takes to download and appear on screen in a usable state, usually measured in seconds or milliseconds.

A place where copies of files or data are kept for a while so they can be served again faster, for example in the browser, on a CDN or on the server.

Search Engine Optimization: work that helps a site's pages to be found, understood and shown higher in search engines such as Google.

A short written description of an image, added with the alt attribute, that screen readers read aloud and browsers show if the image fails to load.

A title for a page or section, marked up with the elements h1 to h6, where h1 is the main title and the others show lower levels.

Clickable text or an image that takes the user to another page or place, created in HTML with the a element and its href attribute.

A part of a page where users enter information, such as a login, search or contact form, which is then sent to a server.

A box or control in a form where a user types or selects a value, such as a name, an email address, a password or a date.

HTML that uses elements according to their meaning, such as nav, main, article and button, instead of generic div and span elements for everything.

The CSS idea that every element is a rectangular box made of content, padding, border and margin, which together decide its size and spacing.

The space outside an element's border, which pushes other elements away from it.

The space inside an element, between its content and its border, which keeps the content away from the edges.

The visible area of a web page inside the browser window or phone screen, which changes size when the window is resized or the phone is rotated.

A function that waits for a specific event on an element, such as a click or a key press, and runs when that event happens.

Describes code that starts a task, such as loading data, and goes on running other code instead of waiting for that task to finish.

A JavaScript object that stands for a result that is not ready yet, such as data from a server, and that later succeeds or fails.

The data an app or component holds at a given moment, such as whether a menu is open or what a user has typed, which controls what is shown.

A web app that loads one HTML page and then updates the content with JavaScript as the user moves around, instead of loading a new page each time.

The period during which a server remembers a logged-in user, usually through an ID stored in a cookie, until the user logs out or the session expires.

Happening on the server rather than in the user's browser, for example server-side code, validation or rendering.

A named setting, such as a database URL or an API key, that is given to an app from outside the code so that it can differ between environments.

Content Delivery Network: a network of servers around the world that stores copies of a site's files and delivers them from the location closest to each user.

Working the same way in different browsers, such as Chrome, Firefox, Safari and Edge, and in their different versions and on different devices.

The rule CSS uses to decide which style wins when several rules target the same element: ID selectors beat class selectors, which beat element selectors.

The standard package manager for JavaScript and the online registry it installs packages from, used through commands such as npm install.

A programming language built on JavaScript that adds types, so that many errors are found while you write the code instead of when it runs.

A tool that combines many JavaScript, CSS and other files and their dependencies into a few optimized files that browsers can load quickly.

A technique that waits to load images, videos or code until they are actually needed, for example when the user scrolls near them.

To make code files smaller by removing spaces, comments and line breaks and by shortening names, without changing how the code works.

Making image files as small as possible while they still look good, for example by compressing and resizing them and by using modern formats.

Three metrics from Google that measure real user experience on a page: loading (LCP), responsiveness (INP) and visual stability (CLS).

Accessible Rich Internet Applications: a set of HTML attributes, such as role and aria-label, that give assistive technology extra information about elements.

Using a site with the keyboard only, moving between links, buttons and fields with Tab and activating them with Enter or Space.

A number that compares the brightness of text with the brightness of its background; high ratios such as 7:1 are easier to read than low ones such as 2:1.

An HTML tag in the head of a page that gives browsers and search engines information about the page, such as its description or viewport settings.

A file, usually sitemap.xml, that lists the pages of a website so that search engines can find and crawl them more easily.

For a search engine bot to visit pages automatically, follow their links and download their content so that they can be indexed.

A program designed to run on a smartphone or tablet, usually downloaded from an app store such as the App Store or Google Play.

Apple's operating system for the iPhone. Apps for it are built with Apple's own tools, mainly Xcode and the Swift programming language.

Google's mobile operating system, used by many phone makers such as Samsung, Xiaomi and Google itself. Its apps are mainly written in Kotlin or Java.

An app built for one specific platform with that platform's own languages and tools, such as Swift for iOS or Kotlin for Android.

Describes an app or framework that runs on more than one platform, usually both iOS and Android, from one shared codebase.

A program that copies the hardware and software of a mobile device on a computer, so you can run and test an app without a real phone.

Apple's tool, included with Xcode, that runs iOS apps on a Mac in a window that looks like an iPhone or iPad, without imitating the real device hardware.

One full view of an app that the user sees at a time, such as the login screen or the settings screen; apps are planned as a set of screens.

To touch the screen quickly with one finger to select something, for example a button, a link or an item in a list.

To move the content of a screen up, down or sideways to see more of it, usually by sliding a finger across a list or page.

The series of states an app moves through from launch to being closed, such as active, in the background or suspended, and the events the system sends at each change.

The state of an app that is still running or kept in memory but is not visible on the screen, for example after the user switches to another app.

The state of an app that is open and visible on the screen, so the user can interact with it directly.

A way of working in which an app stays usable without an internet connection, showing saved data and sending changes later when the connection returns.

Space on the device itself where an app saves data, such as settings, files or a database, so it is available without contacting a server.

Something a user buys inside an app, such as extra features, game coins or a subscription, paid through the App Store or Google Play.

Apple's development environment for writing, building, testing and publishing apps for iPhone, iPad, Mac and other Apple platforms. It runs only on macOS.

Google's official development environment for Android apps, based on IntelliJ IDEA, with a code editor, the Android Emulator, profilers and build tools.

The programming language Apple introduced in 2014 for building apps on iOS and its other platforms; it has largely replaced the older Objective-C.

A modern programming language created by JetBrains that Google recommends as the main language for Android development; it works alongside Java code.

A cross-platform framework created by Meta that lets developers build iOS and Android apps with JavaScript or TypeScript and React, using real native UI components.

Google's open-source framework for building mobile, web and desktop apps from one codebase; it uses the Dart language and draws its own UI on the screen.

On iOS, the bar at the top of a screen that shows the title, a back button and action buttons; on Android, the name also refers to the system bar with the back and home controls.

A bar, usually at the bottom of the screen, with icons that let users switch between the main sections of an app, such as Home, Search and Profile.

To move a finger quickly across the screen in one direction, for example to see the next photo, delete an item in a list or go back.

The first screen people see while an app starts, usually showing the logo on a plain background until the main content is ready.

The first screens a new user goes through after installing an app, which explain the main features, ask for permissions or help set up an account.

A display setting that shows light text on a dark background across the system and in apps that support it, often chosen for use at night.

To make the data on the device and on the server the same by sending local changes and downloading new ones; short for synchronize.

An app built with web technologies such as HTML, CSS and JavaScript and wrapped in a native shell, so it can be installed from the stores and use some device features.

A component that shows web pages inside an app without opening a separate browser, so developers can display web content within a native screen.

A development feature that shows code changes in the running app within a second or two, usually without losing the current screen or its state.

The thread on which an app draws its interface and handles user input; if slow work runs there, the app freezes and stops responding.

Launching an app when it is not already in memory, so the system has to create everything from the beginning; it is the slowest kind of app launch.

In Android, a core app component that represents one screen with a user interface and has its own lifecycle, from being created to being destroyed.

In iOS UIKit, an object that manages one screen or part of a screen: it loads the views, responds to user actions and handles events such as the screen appearing.

Apple's declarative framework, introduced in 2019, for building user interfaces in Swift by describing what the screen should show for each state.

Google's modern declarative toolkit for building Android user interfaces in Kotlin, using composable functions instead of XML layouts.

A small view from an app that shows live information on the home screen or lock screen, such as the weather or the next meeting, without opening the app.

A panel that slides up from the bottom of the screen to show extra options or content, while part of the screen behind it often stays visible.

A screen or window shown on top of the current one that users must deal with, by finishing a task or closing it, before they can return to what they were doing.

A gesture in which the user drags the top of a list downward and releases it to load the newest content, while a small spinner shows that it is loading.

The part of the screen that is not covered by the notch, camera cutout, status bar or home indicator, where important content and buttons can be placed safely.

The way a device is held and the screen is shown: portrait (upright) or landscape (sideways); an app can support one or both.

How many pixels a screen has in a given physical area, measured in dots per inch (dpi); high-density screens need larger image files to look sharp.

A small, fast database engine stored in a single file, built into both iOS and Android and widely used to save structured data on the device.

A message an app sends over the internet to a server, for example to load a list or save a form, together with the response it waits for.

To save data so that it stays on the device after the app is closed or the phone restarts, for example in a database, a file or the settings.

Checking with Apple or Google that a purchase made in the app is real and still valid before giving the user the paid content or features.

How easily people reach their goals and how they feel when they use a product, service or website, from the first visit to getting help later.

A simple, usually black-and-white drawing of a screen that shows where content, buttons and navigation will go, without final colours, images or styling.

A static, realistic image of a screen or product that shows the final colours, fonts, images and spacing but cannot be clicked.

The study of people who use or might use a product, through interviews, observation, surveys and tests, to understand their needs, behaviour and problems.

A fictional but realistic profile of a typical user, built from research, with a name, goals, needs and frustrations, that helps a team design for real people.

The way content and features in a website or app are organised, grouped and labelled so that people can find what they need and understand where they are.

A diagram or sequence that shows the screens, steps and decisions a user goes through to complete one task in a product, such as resetting a password.

A shared collection of reusable components, design rules, tokens and documentation that keeps a company's products consistent and helps teams build faster.

The use of size, colour, contrast, spacing and position on a screen so that people notice the most important elements first and read the rest in a clear order.

The choice and arrangement of typefaces, font sizes, weights, line height and letter spacing to make text readable and to set the tone of a design.

The part of design that decides how a product behaves when people use it: what happens when they tap, type, swipe or wait, and how the system responds.

An approach that puts users' needs, goals and limits at the centre of every stage of design, using research and testing with real people to guide decisions.

A specific problem, frustration or obstacle that users experience, such as a confusing step, a long wait or a missing feature.

A clickable or tappable element that lets users perform an action, such as submitting a form, saving changes or opening a menu.

A button, link or short message that asks users to take the next step, such as 'Sign up', 'Add to cart' or 'Start free trial'.

The menus, links, tabs and other elements that let people move around a website or app and understand where they are.

The stage when designers pass finished designs, specs and assets to developers so they can build them, ideally with a walkthrough and clear notes.

A one-to-one conversation with a user or potential user, usually 30 to 60 minutes long, to learn about their goals, habits, problems and opinions.

Research that collects words, observations and stories from a small number of people to understand why they think and behave the way they do.

Research that collects numbers from many users, such as survey ratings, click rates or task times, to measure behaviour and spot patterns.

A research method in which participants group topics written on cards into categories that make sense to them, showing how users expect content to be organised.

Describes an early design, wireframe or prototype that is rough and simple, without real colours, images or detailed content, so it is quick to make and change.

Describes a design or prototype that looks and behaves almost like the final product, with real content, colours, images and interactions.

A prototype in which screens are linked together, so clicking or tapping buttons and links takes the user to the next screen as in a real product.

The signal a system gives in response to a user's action, such as a button changing colour, a loading spinner or a success message, so people know what happened.

A meeting where designers show work in progress and colleagues give structured, honest feedback on whether the design meets its goals.

A quality of an object or interface element that suggests how it can be used, for example a raised button that looks pressable or a handle that invites pulling.

A menu that stays closed and shows a list of options only when the user clicks or taps it, letting them pick one without taking up space.

A document that sets the visual and writing rules for a brand or product, such as colours, typography, logo use, icon style and tone of voice.

The difference in brightness between text or an icon and its background, which decides how easy it is to see, especially for people with low vision.

Short for Web Content Accessibility Guidelines: the international standard, published by the W3C, that explains how to make websites and apps accessible.

The empty space between and around elements on a page, which gives content room to breathe, groups related items and makes the design easier to read.

A set of invisible columns, rows and gaps that designers use to place and align elements consistently across a page or screen.

The fixed set of colours chosen for a product or brand, usually with primary, secondary and neutral colours plus status colours such as red for errors and green for success.

A small picture or symbol that stands for an action, object or idea, such as a magnifying glass for search or a bin for delete.

A person's own idea of how something works, based on past experience, which shapes what they expect a product to do.

The amount of mental effort a person needs to understand and use something; too much of it makes people slow, confused or likely to give up.

Anything in an experience that slows users down or makes a task harder, such as extra steps, long forms, unclear labels or slow loading.

What a screen shows when there is no content yet, such as an empty inbox, a search with no results or a new account with no projects.

A small text box that appears when users hover over, focus on or tap an element, giving a short explanation or the name of an icon.

A small square that users click or tap to select or clear an option; several checkboxes in a group can be selected at the same time.

A switch that turns a setting on or off with one tap and usually takes effect immediately, like turning on dark mode or notifications.

A row of links, usually near the top of a page, that shows where the current page sits in the site structure, such as Home > Shoes > Running.

An icon made of three horizontal lines that opens a hidden navigation menu, often used on mobile screens to save space.

A named value that stores a design decision, such as a colour, font size or spacing amount, so designers and developers use the same value everywhere.

A ready-made collection of interface elements, such as buttons, forms, icons and navigation bars, that designers drag into their files to build screens quickly.

The amount of space between and inside elements, such as the gap between cards or the padding inside a button, chosen to create rhythm and grouping.

Matching the design exactly, down to every pixel of size, spacing and colour, especially when a built product is compared with the original design file.

A small, single-purpose moment of interaction with its own feedback, such as liking a post, switching a toggle or a field shaking after a wrong password.

The way an element looks when the mouse pointer is over it, for example a button getting darker, which shows that it can be clicked.

A movement of the fingers on a touchscreen, such as a tap, swipe, pinch or long press, that tells a device to do something.

The short pieces of text in an interface, such as button labels, hints, error messages and confirmations, that guide users through a task.

A rectangular container that groups related content about one item, such as a product photo, title and price, and often links to more details.

The visible style, usually an outline or ring, that shows which element is currently selected when someone moves through a page with a keyboard.

An approach that designs for the full range of human diversity, including ability, age, language, culture and situation, so that more people can use a product.

A research method that checks whether people can find items in a site's structure by showing them only a text menu tree, without any visual design.

A principle stating that the time needed to reach a target depends on its distance and size: bigger and closer targets are faster and easier to hit.

A software framework that gives structure, tools, and reusable parts for building the user interface of a website or app.

A software framework that gives structure, tools, and conventions for building server-side apps, APIs, and database-connected features.

A collection of reusable UI components that teams use to build screens with consistent behavior, styling, and accessibility.

Handling events from several child elements through a listener on an ancestor, using event propagation.

Building a usable basic experience first, then adding richer behavior where the user's browser or device supports it.

A build optimization that removes unused exported code from the output when the tool can determine it is unnecessary.

Dividing application code into separate bundles that can be loaded independently rather than as one large file.

A browser-managed script that runs separately from a page and can handle events such as network requests for controlled pages.

An interface behavior that keeps keyboard focus within an active region, such as a modal dialog, until that region is dismissed.

To delay an action until events have stopped arriving for a chosen interval, usually combining rapid triggers into one action.

The field of computer science that builds systems able to do tasks that normally need human intelligence, such as understanding language, recognizing images or making decisions.

A branch of artificial intelligence in which computers learn patterns from data and use them to make predictions or decisions, instead of following rules written by hand.

The result of training: a program with learned parameters that takes an input, such as an image or a sentence, and produces an output, such as a label, a number or a text.

A collection of examples, such as images, texts or table rows, gathered and organized for training, testing or analyzing a model.

The examples a model learns from during training. Their quality, size and variety largely decide how well the model works on new data.

The correct answer attached to a training example, such as 'spam' for an email or 'cat' for a photo, which a supervised model learns to predict.

To teach a model by showing it many examples and repeatedly adjusting its parameters so that its outputs get closer to the correct answers.

A type of machine learning in which the model learns from examples that come with correct answers (labels), so that it can predict the answer for new examples.

A model made of many connected units arranged in layers, loosely inspired by the brain. Each connection has a weight that is adjusted during training.

The part of machine learning that uses neural networks with many layers. It powers modern speech recognition, image recognition and large language models.

A formula that measures how far a model's predictions are from the correct answers. Training tries to make this value, called the loss, as small as possible.

A problem in which a model learns its training data too closely, including noise and accidents, so it scores well on that data but poorly on new data.

A part of the data kept aside and used only at the end, to measure how well a finished model works on examples it has never seen.

The share of a model's predictions that are correct, out of all its predictions. A model with 90% accuracy is right on nine of every ten examples.

The stage in which a trained model is used to produce outputs for new inputs, for example answering a user's question or classifying a new photo.

Short for large language model: a very large neural network trained on huge amounts of text to predict the next token, which lets it answer questions, write, summarize and code.

The text or other input given to a generative model to tell it what to do, such as a question, an instruction or a few examples followed by a task.

Extra training of an already pretrained model on a smaller, task-specific dataset, so that it does better on that task or follows a particular style.

Systematic unfairness in a model's outputs, for example working worse for some groups of people, usually caused by skewed data or design choices.

A task in which a model puts each input into one of a fixed set of categories, for example deciding whether an email is spam or which product category a photo shows.

The area of AI that deals with human language, building systems that read, understand, translate, summarize or generate text and speech.

The area of AI that teaches computers to understand images and video, for example recognizing objects, reading text in photos or tracking movement.

A type of machine learning that finds structure in data without correct answers (labels), for example grouping similar customers or spotting unusual events.

A part of the data kept out of training and used during development to compare models, tune settings and decide when to stop training.

One complete pass of the training process through the whole training dataset. Models are usually trained for several epochs.

A hyperparameter that controls how big a step the model takes each time it updates its parameters. Too high makes training unstable; too low makes it very slow.

A setting chosen by engineers before training rather than learned from data, such as the learning rate, the batch size or the number of layers.

The number of training examples a model processes together before it updates its parameters once.

The basic optimization method behind training: it repeatedly changes each parameter a little in the direction that reduces the loss the most.

A problem in which a model is too simple or trained too little to capture the patterns in the data, so it performs poorly even on the training data.

Of all the cases a model marked as positive, the share that really were positive. High precision means few false alarms.

Of all the cases that really were positive, the share that the model managed to find. High recall means few missed cases.

A single number that combines precision and recall by taking their harmonic mean. It is high only when both precision and recall are high.

A standard set of tasks and data used to measure models in the same way, so that results from different teams can be compared fairly.

A neural network architecture, introduced in 2017, that uses attention to process all parts of a sequence in parallel. It is the basis of today's LLMs.

One level of a neural network: a group of units that takes the output of the previous level, transforms it and passes the result on to the next one.

A confident but false or made-up output from a generative model, such as an invented fact, quote, source or piece of code.

The maximum amount of text, measured in tokens, that a language model can take into account at once, including the prompt, any documents and its own answer.

AI systems that create new content, such as text, images, audio, video or code, instead of only sorting inputs into classes or predicting values.

A list of numbers (a vector) that represents a word, sentence, image or other item in such a way that items with similar meanings end up close to each other.

Instructions given to a chat model before the conversation starts, which set its role, rules, tone and limits for all the messages that follow.

Short for retrieval-augmented generation: a method in which a system first searches a document collection for relevant passages and then gives them to an LLM to write the answer.

The work of adding labels or notes to raw data, such as drawing boxes around objects in images or marking the topic of each sentence, so that it can be used for training or evaluation.

The steps that clean and prepare raw data before a model uses it, such as removing duplicates, filling in missing values, resizing images or normalizing text.

A type of machine learning in which an agent learns by trial and error, taking actions in an environment and receiving rewards or penalties for the results.

An unsupervised task that groups similar items together without predefined labels, for example sorting customers into segments by how they shop.

The learned numbers on the connections of a neural network, which decide how strongly each input affects the next layer. Together they hold what the model has learned.

A simple mathematical function applied to the output of each unit in a neural network. It adds non-linearity, which lets the network learn complex patterns.

The method that calculates how much each weight in a neural network contributed to the error, by passing the error backward from the output layer toward the input.

A mechanism that lets a model decide which other parts of its input are most relevant to each part, and give those parts more weight when it builds its output.

The correct, verified answers that a model's outputs are compared with, usually created by careful human labeling or by real measurements.

Creating extra training examples by making small changes to existing ones, such as flipping or cropping images or rephrasing sentences, so that a model sees more variety.

A table that shows, for each true class, how many examples the model assigned to each class, so you can see at a glance which classes it mixes up.

An evaluation method that splits the data into several parts, trains on all but one and tests on the remaining part, repeating until every part has been used for testing.

Any technique that stops a model from becoming too complex during training, in order to reduce overfitting and help it work better on new data.

The first and largest training stage, in which a model learns general patterns from a huge amount of mostly unlabeled data before it is adapted to specific tasks.

A setting that controls how random a generative model's output is. Low values give more predictable, focused answers; high values give more varied and creative ones.

The practice of designing, testing and improving prompts so that a model gives accurate, useful and consistent outputs for a task.

Giving a model a few worked examples of a task inside the prompt, so that it can follow the same pattern for a new input without any extra training.

Reusing what a model learned on one task or dataset as the starting point for a different but related task, instead of training a new model from scratch.

A model that has only been pretrained, before any fine-tuning or instruction training. It continues text well but does not reliably follow instructions or chat.

Short for convolutional neural network: a network that slides small learned filters over an image to detect patterns such as edges, textures and shapes.

A computer vision task that finds objects in an image or video and gives both their class, such as 'car' or 'person', and their position, usually as a box.

The component that splits text into tokens and turns them into numbers a model can read, and that turns the model's output numbers back into text.

An NLP task that detects the opinion or emotion in a text, for example whether a product review is positive, negative or neutral.

The goal that a model treats different groups of people equitably, for example by giving similar error rates or approval rates across genders, ages or regions.

To reach a stable state during training, where the loss stops going down in a meaningful way and more training brings little benefit.

A situation in which some classes in a dataset have far more examples than others, for example 990 normal transactions for every 10 fraudulent ones.

A mistake in which information that won't be available at prediction time, such as test examples or future data, gets into training and makes results look better than they are.

Describes the best result or method known at a given time, usually measured on a well-known benchmark.

Able to work with more than one type of data, such as text, images, audio and video, like a model that can answer questions about a photo.

A system built around an LLM that can plan steps and use tools, such as search, code or company software, to complete a task with little human help.

Step-by-step reasoning that a model writes out before giving its final answer, which often improves results on math, logic and multi-step problems.

A database designed to store embeddings and quickly find the ones most similar to a query, which makes it a common part of semantic search and RAG systems.

Short for low-rank adaptation: a fine-tuning method that keeps the original weights frozen and trains only small added matrices, which saves a lot of memory and storage.

Short for reinforcement learning from human feedback: a training stage in which people rank model answers, a reward model learns their preferences, and the LLM is then tuned to score higher.

Training a smaller 'student' model to copy the outputs of a larger 'teacher' model, so that it keeps much of the quality while being faster and cheaper to run.

A computer vision task that labels every pixel of an image with a class or an object, giving exact shapes instead of rough boxes.

A rectangle drawn around an object in an image and stored as coordinates, showing where the object is. It is the standard output of object detection.

A generative model that learns to remove noise step by step, so it can create new images, audio or video by starting from pure noise and gradually refining it.

An NLP task that finds the names of people, organizations and places, as well as dates, amounts and similar items, in a text and labels each one with its type.

A large, organized collection of texts or speech recordings used to train language models or to study how a language is used.

How well people can understand why a model made a particular prediction or decision, often with the help of tools that show which inputs mattered most.

A rule or check placed around a model to stop harmful, off-topic or unsafe inputs and outputs, for example a filter that blocks personal data in answers.

Testing a model or AI system by deliberately trying to make it fail or behave harmfully, in order to find weaknesses before real users or attackers do.

Making an AI system's behavior match human intentions and values, so that it is helpful and honest and avoids causing harm. Also the research field that studies this.

The practices and tools for running machine learning reliably in production: versioning data and models, automating training and deployment, and monitoring models over time.

A change over time in the data a deployed model receives, compared with the data it was trained on, which can slowly make its predictions worse.

Storing a model's weights, and sometimes its activations, with fewer bits, for example 8-bit or 4-bit numbers instead of 16-bit, to make it smaller and faster.

The process of finding and fixing problems in data, such as wrong formats, typos, duplicates and missing values, so that it can be trusted for analysis.

Empty cells or fields in a dataset where no value was recorded, often shown as NULL, NaN or a blank, which need to be handled before analysis.

A value that is very far from most of the other values in a dataset, for example one order of 50,000 dollars among orders of around 40 dollars.

The first stage of an analysis, in which you summarise and plot a dataset to understand its structure, spot patterns and problems, and form questions to test.

The practice of showing data as charts, graphs and maps so that people can see patterns, comparisons and changes more quickly than in a table.

Short for key performance indicator: one of a small number of metrics that a company or team has chosen to measure progress towards its main goals.

An experiment in which users are randomly split into two groups that see different versions of something, so you can measure which version performs better.

A sign that a result is unlikely to be explained by random chance alone, usually judged by whether the p-value falls below a threshold chosen in advance, such as 0.05.

A series of automated steps that moves data from its sources, transforms it and delivers it to a destination such as a warehouse, dashboard or model.

Short for extract, transform, load: the process of taking data out of source systems, cleaning and reshaping it, and loading it into a warehouse.

The value you get by adding all the numbers in a group and dividing the total by how many numbers there are; the usual meaning of 'average'.

The middle value when all the numbers in a group are sorted from smallest to largest; half the values are below it and half are above it.

The way the values of a variable are spread out, showing which values are common, which are rare and how far apart they are; often viewed as a histogram.

A relationship between two variables in which they tend to change together, often measured by a coefficient from -1 to 1, where 0 means no linear relationship.

Data in the form in which it was first collected or exported, before anyone has cleaned, combined or summarised it.

Any place that data comes from, such as a database, an app, a spreadsheet, a survey, an external API or a file sent by a partner.

A record that appears more than once in a dataset, for example the same customer or order stored twice, which can make counts and totals too high.

How accurate, complete, consistent and up to date data is, and therefore how far it can be trusted when people make decisions.

A measure of how spread out values are around the mean: a small value means most values are close to the average, a large one means they vary a lot.

A smaller group selected from a larger population or dataset and studied in order to draw conclusions about the whole.

The general direction in which a value moves over a period of time, such as a steady rise in sales, ignoring short-term ups and downs.

A set of measurements of the same thing recorded in order at regular time intervals, such as daily revenue or hourly website visits.

A group of customers or users who share certain characteristics, such as country, device or buying behaviour; as a verb, to divide data into such groups.

A specific, testable statement about what you expect to happen and why, written down before an experiment or analysis starts.

The group in an experiment that does not get the change being tested, so its results show what would have happened without that change.

The probability of seeing a result at least as extreme as the one observed if there were really no effect, that is, if the null hypothesis were true.

The number of observations, people or units included in a sample or an experiment; larger samples usually give more reliable estimates.

The percentage of users who complete a desired action, such as buying or signing up, out of all the users who had the chance to do it.

The loss of customers or users who stop using a product or cancel a subscription during a given period; also the rate at which this happens.

A chart that uses rectangular bars of different lengths to compare values across categories, such as sales by region or users by country.

A chart that shows the relationship between two numeric variables by drawing one dot for each observation, with one variable on each axis.

To combine many rows of data into summary values, such as totals, counts or averages, usually grouped by a category such as day, product or region.

A one-off analysis done quickly to answer a specific question that has just come up, rather than a regular report or a planned project.

To fill in missing values with estimated ones, for example with the column's median, the most common category or a value predicted from other columns.

The work of turning messy raw data into a clean, well-structured form for analysis, including reshaping, merging, filtering and fixing formats.

Describes a distribution that is not symmetrical and has a long tail of values on one side, such as income data where a few people earn far more than most.

A small set of numbers that describe a dataset at a glance, such as the count, mean, median, minimum, maximum and standard deviation of each column.

A quick, simple test to make sure a result is reasonable before you trust it, for example checking that a total matches a figure you already know.

The value below which a given percentage of observations fall; for example, the 90th percentile of load time is the time within which 90% of page loads finish.

A relationship in which a change in one thing directly produces a change in another, rather than the two simply moving together.

A range of values, calculated from sample data, that is likely to contain the true value; a 95% interval comes from a method that captures the true value 95% of the time.

One of the versions shown to users in an experiment, usually the new version that is compared with the control.

The default assumption in a statistical test that there is no effect or no difference, for example that a new design does not change the conversion rate.

A group of users who share a starting point or experience in the same period, such as everyone who signed up in January, and who are followed over time.

The series of steps users go through to reach a goal, such as visit, sign-up and purchase; fewer people remain at each step, like the shape of a funnel.

A chart that shows the distribution of one numeric variable by dividing its range into intervals, called bins, and drawing a bar for how many values fall in each.

A chart that connects data points with a line to show how a value changes over time, such as monthly revenue or daily sign-ups.

To move from a summary view to more detailed data, for example from yearly sales to monthly sales and then to single stores, in order to find the cause of a pattern.

To keep only the rows of data that meet certain conditions, such as orders from 2024 or users in Turkey, and leave out the rest.

A large storage system that holds huge amounts of raw data in its original format, structured or not, until someone needs to process and analyse it.

Processing data in large groups at scheduled times, for example once every night, rather than handling each record the moment it arrives.

A simple reference point, such as last year's figure, the current version or a very basic model, against which new results are compared.

The work of creating new input variables from raw data to help a model learn, for example turning a date of birth into an age or counting a customer's purchases.

The value a model is trained to predict, such as whether a customer will churn or how much they will spend; also called the label or dependent variable.

To rescale numeric values to a common range, such as 0 to 1, or to adjust figures so that groups of different sizes can be compared fairly.

A document or table that describes every field in a dataset: its name, meaning, data type, allowed values, source and owner.

Recording specific user actions in an app or website, such as clicks, page views or purchases, as events with details like time, user and device.

Collecting data automatically from websites with a program that downloads pages and extracts the parts you need, such as prices or product names.

A data point or pattern that clearly differs from what is expected, such as a sudden spike in sign-ups or a day with no orders at all.

A regular pattern in data that repeats over a fixed period, such as higher sales every December, more traffic on weekdays or a dip every summer.

The probability that a test will detect an effect of a given size if that effect really exists; tests are commonly designed to have 80% power.

A chart that shows values in a grid using colours, where darker or warmer colours usually mean higher values, such as orders by day of the week and hour.

A table that summarises a larger dataset by grouping rows and columns by categories and calculating totals, counts or averages for each combination.

The level of detail in data, meaning what one row represents, for example one row per order, per customer per day or per store per month.

A record of where data comes from, which steps and systems it passes through and how it changes along the way, right up to the reports that use it.

A measure showing how much each input feature influences a model's prediction or decision compared with other features.

The process of measuring how well a machine learning model performs on chosen data and metrics after training.

Ending model training when a monitored validation measure stops improving according to a chosen rule.

A change over time in the relationship between model inputs and the outcome the model is meant to predict.

A system that manages reusable input features for machine learning, helping make them available for training and prediction.

A mismatch between the data or feature processing used in model training and the data or processing used for live predictions.

A substantial loss of previously learned capabilities when a model is trained on new tasks or data.

A sequence-training method that supplies the correct previous output as input for predicting the next step.