Reference

Web + AI glossary

49 definitions you will actually meet while building — from a11y and RSC to RAG and prompt injection.

A C D E F H I J L M O P R S T W Z

A

AbortController Web: Browser API to cancel fetch/requests so stale responses do not overwrite newer UI state.
Accessibility (a11y) Web: Designing interfaces so people with disabilities can perceive, operate, and understand them — keyboard, screen readers, contrast, and semantics.
App Router Web: Next.js routing model based on the app/ directory with Server Components, layouts, and nested routes.
ARIA Web: Accessible Rich Internet Applications — attributes that expose roles and states when native HTML is not enough. Prefer native elements first.

C

Cache-Control Web: HTTP header directing browsers/CDNs how long a response may be cached and when it must revalidate.
CDN Web: Content Delivery Network — edge servers that cache static assets (and sometimes HTML) closer to users.
CLS Performance: Cumulative Layout Shift — how much visible content jumps during load. Reserve space for images, fonts, and ads.
Core Web Vitals Performance: Google’s key UX metrics: LCP, INP, and CLS — used in ranking and as a quality bar for builders.
CORS Web: Cross-Origin Resource Sharing — browser rules for calling APIs from another origin. Configure deliberately on APIs.
CSP Security: Content Security Policy — HTTP header limiting script/style sources to reduce XSS impact.
CSR Web: Client-Side Rendering — HTML shell hydrates and fetches data in the browser. Flexible but can hurt first paint if overused.

D

Debouncing Web: Delaying a function until events pause (e.g. search-as-you-type) to cut request spam.
Design tokens Design: Named values for color, space, type, and shadow — the source of truth for a design system.
Discriminated union TypeScript: A union type with a shared literal field (e.g. status) that lets TypeScript narrow safely.
DX Culture: Developer Experience — how pleasant and fast it is to build, test, and ship in a codebase.

E

E-E-A-T SEO: Experience, Expertise, Authoritativeness, Trust — quality signals search engines associate with helpful content.
Edge Web: Compute or cache close to users (CDN/edge functions). Great for static and light dynamic work; careful with heavy DB calls.

F

Few-shot prompting AI: Including a few input/output examples in a prompt so the model mirrors the pattern.

H

Hallucination AI: When a model invents APIs, citations, or facts confidently. Mitigate with constraints, tools, and human review.
Hydration Web: Attaching React event handlers to server-rendered HTML. Excess hydration costs INP and bundle size.

I

Idempotency Backend: Repeating the same request yields the same effect — critical for payments and “submit” buttons.
INP Performance: Interaction to Next Paint — how quickly the page responds to clicks/keys. Heavy main-thread work hurts INP.
Island architecture Web: Static pages with small interactive “islands” instead of hydrating the entire document.

J

JWT Auth: JSON Web Token — compact claims often used for sessions/APIs. Still verify signatures server-side; treat payload as untrusted until verified.
JWT vs session cookie Auth: Tokens can be stored many ways; httpOnly secure cookies often beat localStorage for XSS resistance on classic web apps.

L

LCP Performance: Largest Contentful Paint — when the main content appears. Optimize hero media, fonts, and critical CSS.
LLM AI: Large Language Model — a model trained to predict text, used for chat, coding assistance, and generation.

M

Monorepo Tooling: Multiple packages in one repository (apps + libs) with shared tooling — powerful, needs discipline.

O

Observability Ops: Ability to understand system behavior via logs, metrics, and traces — especially under failure.
Optimistic UI Web: Updating the UI before the server confirms, then reconciling on success/failure.

P

Progressive enhancement Web: Core experience works without JS; enhancements layer on for capable browsers.
Prompt engineering AI: Crafting instructions (role, task, context, constraints, format) so models return reliable, usable outputs.
Prompt injection AI: Attack where untrusted text tries to override system instructions. Validate actions server-side; never grant tools blindly.

R

RAG AI: Retrieval-Augmented Generation — fetch relevant docs, then generate an answer grounded in those chunks.
Rate limiting Backend: Throttling requests per user/IP to protect APIs from abuse and accidental loops.
RLS Backend: Row Level Security — database policies that restrict which rows a role can read/write (common in Supabase/Postgres).
RSC Web: React Server Components — components that render on the server and send a serialized result, reducing client JS.

S

Schema.org SEO: Vocabulary for structured data (JSON-LD) that helps search engines understand organizations, articles, and FAQs.
Semantic HTML Web: Using elements for their meaning (nav, main, button, label) so assistive tech and crawlers understand structure.
Soft navigation Web: Client-side route changes without full document reload — faster, but you must manage focus and titles.
SSG Web: Static Site Generation — HTML built ahead of time and served from cache/CDN.
SSR Web: Server-Side Rendering — HTML generated per request (or on demand) before it reaches the browser.
System prompt AI: High-priority instructions that set role and rules for a model session. Keep secrets out of it.

T

Technical SEO SEO: Crawlability, indexation, sitemaps, canonicals, performance, and structured data — the engineering side of search.
Tree shaking Web: Bundler removal of unused exports to shrink JavaScript shipped to users.
Type narrowing TypeScript: Control-flow analysis that refines a type after checks (typeof, in, discriminators).

W

Web Vitals Performance: Family of metrics describing loading, interactivity, and visual stability — includes Core Web Vitals.

Z

Zero-shot prompting AI: Asking a model to perform a task with instructions only — no examples. Works for simple tasks; few-shot helps harder ones.
Zod TypeScript: Schema validation library often used to parse API/env input at runtime with inferred types.

Never miss a drop

Get new guides on web development and AI, Prompt Lab tips, videos, downloads and exclusive deals — straight to your inbox. No spam, unsubscribe anytime.