Skip to content

4.8 — Vocabulary: The Complete Lexicon

Every term from every phase, organized by when it was introduced. Know these and you can hold your own in any technical conversation.

Previous vocabulary pages: Phase 0 | Phase 1 | Phase 2 | Phase 3


Terms introduced in Phase 0: See the Matrix.

TermDefinition
AISoftware that makes predictions and generates content by learning patterns from data
LLM (Large Language Model)The type of AI behind ChatGPT, Claude, etc. — trained on text, predicts the next word
ModelThe trained AI brain itself — Claude Opus, GPT-5, Gemini are different models
Generative AIAI that creates new content vs. AI that classifies or sorts existing data
Vibe codingBuilding software by describing what you want in natural language and letting AI generate the code
Terminal / CLIA text-based way to talk to your computer by typing instructions
IDEA specialized text editor for working with code — VS Code is the most common
GitA system that tracks every change you make to files so you can undo anything
GitHubA website where people store and share Git-tracked projects
Repository (Repo)A project folder tracked by Git
FrontendWhat the user sees and touches — the app, the website, the buttons
BackendWhat runs behind the scenes — logic, database, AI calls
APIHow two pieces of software talk to each other
DatabaseWhere data lives permanently — users, messages, settings
ServerA computer that’s always on, waiting to respond to requests
Local vs. RemoteLocal = on your computer. Remote = on someone else’s server.
Open sourceSoftware whose code is public — anyone can read, use, improve it
HarnessThe tool between you and the AI model — Claude Code, Cursor, ChatGPT
ContextEverything the AI can “see” when answering you
Context windowThe size limit on how much the AI can see at once
TokenThe unit AI uses to measure text — roughly 3/4 of a word
ProviderThe company that built and runs the model
BugSomething that doesn’t work as intended
DebugFinding and fixing the bug
DeployMaking your software live and available to users
DependencySoftware your project relies on that someone else built
ConfigSettings that control how software behaves
FlagAn option added to a command to modify behavior
PathThe address of a file on your computer
RootThe top-level folder of a project
ScriptA file containing commands that run automatically
RuntimeThe environment where code actually executes
PackageA bundle of code someone else wrote that you install and use
StackThe combination of technologies used to build something
BoilerplateStarter code or template that gives you a foundation to build on
LocalhostYour own computer acting as a server — only you can see it
MarkdownFormatted text format — documentation, notes, README files
JSONStructured data format for config, settings, data exchange
YAMLSame job as JSON, more human-readable — used in config files
.envFile containing secret settings — API keys, passwords. Never shared publicly.

Terms introduced in Phase 1: First Build.

TermDefinition
npmNode Package Manager — installs JavaScript packages and tools from the internet
PromptThe instruction you give to an AI
HallucinationWhen AI confidently generates something that’s wrong
IterationRefining results through multiple rounds of feedback
CommitA saved snapshot of your project at a specific moment
Push / PullSend to / download from a remote server
StagingMarking which changes should go into the next commit
Working directoryThe folder you’re currently in
READMEA file explaining what a project is and how to run it
Entry pointThe file where a program starts running
HostingA service that runs your project on a server so others can access it
PortA numbered channel on your computer for network traffic — localhost:3000 means “my computer, channel 3000”
DNSTranslates human-readable names (google.com) to computer addresses

Terms introduced in Phase 2: Builder’s Toolkit.

TermDefinition
Context engineeringDeliberately shaping what the AI can see to get better output
CLAUDE.mdProject-level instruction file Claude Code reads automatically every session
BranchA parallel version of your project for making changes safely
Pull Request (PR)A proposal to merge a branch, with description and review process
Merge conflictTwo branches changed the same line — you decide which to keep
DiffA view showing exactly what changed between two versions
Stack traceA list showing the chain of function calls that led to an error
RegressionA fix that breaks something that was previously working
Edge caseAn unusual or extreme input that exposes bugs normal inputs don’t
YAGNIYou Ain’t Gonna Need It — don’t build features you might need later
DRYDon’t Repeat Yourself — one source of truth for each piece of logic
ArchitectureThe high-level structure of how a system is organized
ComponentA self-contained, reusable piece of an interface
RouteA URL path that maps to a specific page or action
StateData that can change over time and affects what the user sees
Environment variableA setting stored outside your code so it can change between environments
RefactorRestructuring code without changing what it does
Technical debtShortcuts taken now that create more work later
Scope creepWhen a project gradually grows beyond its original boundaries
MVPMinimum Viable Product — the smallest version that delivers value
MonorepoOne repository containing multiple related projects
MicroserviceArchitecture where the system is split into small, independent services
MonolithArchitecture where everything lives in one application
ChangelogA record of what changed between versions
MiddlewareSoftware that sits between two systems and processes requests as they pass through

Terms introduced in Phase 3: Agent Architect.

TermDefinition
AgentAI system with tools, decision-making ability, and multi-step execution
Tool use / Function callingMechanism letting AI take actions beyond generating text
MCP (Model Context Protocol)Standard protocol for connecting AI to external tools and data sources
MCP ServerProgram that exposes tools from an external system to AI
HookScript that runs automatically on a specific event
RAG (Retrieval-Augmented Generation)Retrieving relevant info and injecting it into AI context
EmbeddingNumeric representation of text for measuring similarity between documents
Vector databaseDatabase optimized for storing and searching embeddings
Semantic searchSearch by meaning, not exact keywords
Knowledge baseAn organized collection of information an AI system can draw from
PartitionA section within a knowledge base — like folders within a filing cabinet
GuardrailA constraint placed on an agent to limit its actions
Human-in-the-loopAgent pauses for human approval at critical decision points
AutonomyThe degree to which an agent can act without human direction
OrchestrationCoordinating multiple agents or systems to accomplish a goal
PipelineA sequence of processing steps where output of one becomes input of the next
InferenceWhen an AI model processes input and generates output — every response = one inference call
LatencyTime between sending a request and getting a response
ThroughputHow many requests a system can handle in a given time period
Token budgetThe maximum tokens allocated for a task or session — controls cost
DeterministicSame input always produces same output — traditional code is deterministic
Non-deterministicSame input can produce different outputs each time — AI models are non-deterministic

Terms introduced in Phase 4: Orchestrator.

TermDefinition
SubagentAn agent spawned by another agent to handle a specific subtask
DelegationAssigning a task from one agent to another
Fan-out / Fan-inSplitting work across multiple agents (fan-out), then combining results (fan-in)
Graceful degradationSystem continues working when a component fails, rather than crashing entirely
Principle of least privilegeEach component gets only the minimum access it needs — nothing more
IdempotentAn operation that produces the same result whether run once or ten times
Race conditionTwo processes trying to change the same thing simultaneously, producing unpredictable results
ScalabilityA system’s ability to handle growth — more users, more data, more requests
Load balancerDistributes incoming requests across multiple servers
CacheStoring frequently accessed data in fast, temporary storage
CDN (Content Delivery Network)Servers distributed worldwide storing copies of your content close to users
Rate limitMaximum number of requests allowed in a time period
WebhookA URL that receives automatic notifications when something happens in another system
QueueA list of tasks waiting to be processed in order
ContainerPackaged unit of software that includes everything needed to run, anywhere
DockerThe most common tool for creating and running containers
CI/CDAutomated pipeline that tests and deploys code when you push changes
Production (prod)The live environment real users interact with
StagingA copy of production used for testing before deploying
SandboxAn isolated environment for safe experimentation
SSH (Secure Shell)A secure way to connect to and control a remote computer via terminal
VPN (Virtual Private Network)Encrypted tunnel between your computer and a private network
FirewallSecurity system controlling what network traffic is allowed in and out
SSL/TLSEncryption that makes web traffic secure — the “S” in HTTPS
IP addressA computer’s unique address on a network
Code reviewAnother developer examining your code before it’s merged
LintingAutomated checking of code style and common errors
Unit testTests a single function or component in isolation
Integration testTests how multiple components work together
End-to-end (E2E) testTests the entire application flow as a user would experience it
CoveragePercentage of code tested by automated tests
SemVer (Semantic Versioning)Version numbering: MAJOR.MINOR.PATCH — major = breaking, minor = new features, patch = fixes
Breaking changeA change that makes existing things stop working
Backward compatibleA change that works with existing code without requiring updates
DeprecationMarking something as “still works but will be removed in a future version”
API endpointA specific URL where an API accepts requests
PayloadThe data sent with an API request
Authentication (AuthN)Verifying who someone is — login
Authorization (AuthZ)Verifying what someone is allowed to do — permissions
JWT (JSON Web Token)A compact token used for authentication — carries user info in encoded form
OAuthStandard for giving one app limited access to your account on another service
UptimePercentage of time a system is operational
SLA (Service Level Agreement)A commitment to a specific level of reliability
IncidentAn unplanned event that disrupts service
RollbackReverting to a previous working version after a bad deploy
MonitoringContinuous automated checking of system health
AlertAn automated notification when something goes wrong
Audit trailA record of who did what and when
Spec (Specification)A document describing exactly what needs to be built and what “done” looks like
SprintA fixed time period (usually 1–2 weeks) during which a set of tasks is completed
BacklogA prioritized list of work that hasn’t been started yet
BlockerSomething preventing progress on a task
StandupA brief status update meeting or report
Acceptance criteriaSpecific conditions that must be true for a task to be considered complete

Total vocabulary: ~150 terms. This is the shared language between vibe coders and traditional engineers. Know these, and you can hold your own in any technical conversation.

Phase overview: Phase 4