Concise, plain-English definitions of the technical terms we use in client conversations and project documentation.
The smallest version of a product released to learn from real users.
An MVP enforces scope discipline to validate the idea via fast market entry. It ships the core value proposition only, not the full feature set. Typical delivery: 4–8 weeks.
Google's open-source cross-platform UI framework.
Flutter ships high-performance apps for iOS, Android, web and desktop from a single Dart codebase. Its widget-based render engine delivers near-native 60fps performance.
Meta's React-based cross-platform mobile framework.
React Native bridges JavaScript/TypeScript code to native UI components. Mature ecosystem, broad library support and hot reload enable rapid iteration.
A React meta-framework with server-side rendering and React Server Components.
Next.js standardizes the SSR/SSG output critical for SEO. Its App Router lets you pick per-page rendering strategy (static, server, client); ISR enables caching of dynamic content.
Software delivered over the cloud on a subscription model.
SaaS products require no user installation; they run on multi-tenant architecture with continuous updates and monthly/annual billing. Examples: Notion, Slack, Linear.
A large AI model trained on massive text data that generates natural language.
LLMs (GPT-4, Claude, Gemini, Llama) can be customized via prompt engineering, RAG (retrieval-augmented generation), agent tooling and fine-tuning. Used for chatbots, content generation, coding, analysis.
A method to ground an LLM's answers in your own document corpus.
RAG searches a user query against a vector database, then feeds matched documents as context to the LLM. Reduces hallucination and integrates fresh, private knowledge. Components: embedding model, vector DB (Pinecone/Weaviate/pgvector), retrieval logic.
Generating hundreds-to-thousands of long-tail pages via data + template + schema architecture.
Unlike classic SEO, programmatic SEO auto-generates pages from data sets (like [city] × [service]) instead of manually writing each. Done well it 10x's traffic; to avoid thin-content penalties, each page needs 200–400 unique words plus local signals.
An open-source, self-hostable workflow automation platform.
n8n is an alternative to Zapier and Make. 400+ nodes connect APIs, databases, messaging, AI and file operations in a node-based UI. Self-hostable for data privacy and cost benefits.
A structured-data vocabulary that explains page content to search engines.
Schema.org markup embedded via JSON-LD enables Google, Bing and AI engines to surface rich results and extract answers. Types: Article, Product, FAQPage, BreadcrumbList, Service, Person, Organization.
Google's 3-metric page-experience ranking signal.
LCP (Largest Contentful Paint < 2.5s), INP (Interaction to Next Paint < 200ms) and CLS (Cumulative Layout Shift < 0.1). Pass at the 75th percentile across all page types.
Disciplines for optimizing content for AI engines.
AEO (Answer Engine Optimization), GEO (Generative Engine Optimization), LLMO (LLM Optimization) — optimizing entity signals, answer-first content, llms.txt and Speakable schema to be cited as a source by ChatGPT, Perplexity and AI Overviews.
Language/region pairing signal for multilingual sites.
The tag tells Google which language/region variants of the same content exist. Self-referencing is required, return tags must be reciprocal, x-default fallback is recommended. Misconfiguration (e.g. en-UK instead of en-GB) invalidates the entire cluster.
