Large language model (LLM)
A model trained on huge amounts of text that can read and write natural language. Claude, GPT and Gemini are LLMs. They power chatbots, copilots and agents.
Learn more about Large language model (LLM)Glossary
AI & app terms, explained simply
Forty words you'll hear when building an AI product or app, in plain English. Bookmark it, share it with your team.
AI
AI agent
An LLM-powered system that can plan steps, call tools or APIs, and act on the results to finish a task, rather than only replying with text.
Learn more about AI agentCopilot
An AI assistant embedded inside a product that helps a user do their work, for example drafting replies or summarising data, while the user stays in control.
Retrieval-augmented generation (RAG)
A technique where relevant documents are retrieved from your own data and given to the model at question time, so answers are grounded and can cite sources.
Learn more about Retrieval-augmented generation (RAG)Fine-tuning
Further training a model on your examples to change its behaviour, format or tone. Best for consistent outputs; rarely the best way to add knowledge.
Learn more about Fine-tuningEmbedding
A list of numbers that represents the meaning of a piece of text or an image, so similar meanings sit close together. Used for semantic search and RAG.
Vector database
A database optimised to store embeddings and quickly find the most similar ones. Examples include pgvector, Pinecone and Weaviate.
Prompt engineering
Designing the instructions, examples and context given to a model so it produces reliable, useful outputs.
Context window
The maximum amount of text (measured in tokens) a model can consider at once, including instructions, documents and conversation history.
Token
The unit models read and write, roughly three-quarters of an English word. Model pricing and limits are measured in tokens.
Learn more about TokenHallucination
When a model states something confidently that is not true. Reduced with grounding (RAG), evaluation and clear instructions.
Evals (evaluation suite)
A set of realistic test cases with expected outcomes, run automatically against every prompt or model change to measure quality.
Guardrails
Checks around an AI system that block unsafe, off-topic or malformed inputs and outputs before they reach users or other systems.
Tool calling
A model capability to request a function or API call with structured arguments, which is how agents take actions like booking or searching.
Human-in-the-loop
A design where a person reviews or approves an AI's work at key moments, especially for decisions with real-world consequences.
Multimodal model
A model that understands more than one type of input, such as text plus images, audio or documents.
On-device AI
Running a model directly on a phone or laptop instead of in the cloud, for privacy, offline use and lower latency.
Intelligent automation
Combining AI (to read, classify and decide) with workflow tools (to route and act) to replace repetitive manual processes.
Learn more about Intelligent automationMobile
Native app
An app built with each platform's own language and tools, Kotlin for Android and Swift for iOS. Best for deep device features and peak performance.
Learn more about Native appCross-platform app
An app built once with a framework such as Flutter or React Native and shipped to both Android and iOS.
Learn more about Cross-platform appCold start
The time it takes an app to open from scratch. Fast cold starts strongly influence first impressions and retention.
Crash-free rate
The percentage of user sessions that end without a crash. Healthy consumer apps target above 99.5%.
App Store Optimisation (ASO)
Improving an app's store listing, from title and keywords to screenshots and ratings, so more people find and install it.
Learn more about App Store Optimisation (ASO)Push notification
A message an app sends to a user's device outside the app. Powerful for re-engagement when timely and relevant.
In-app purchase (IAP)
Buying digital goods or subscriptions inside a mobile app through Google Play or the App Store billing systems.
Deep link
A link that opens a specific screen inside an app rather than its home screen, used in campaigns, emails and sharing.
Product & SaaS
MVP (minimum viable product)
The smallest version of a product that delivers the core value to real users, built to learn fast before investing more.
Learn more about MVP (minimum viable product)SaaS
Software as a Service: software delivered over the internet on a subscription, with no installation for the customer.
Learn more about SaaSMulti-tenancy
An architecture where one application serves many customer organisations while keeping each one's data separate and secure.
Product discovery
The research and prototyping phase that decides what to build and why, before committing engineering effort.
Learn more about Product discoveryDesign system
A shared library of components, styles and rules that keeps a product consistent and makes new screens faster to build.
Learn more about Design systemTechnical debt
The future cost of shortcuts taken today in code or architecture. Some is healthy; too much slows every release.
CI/CD
Continuous integration and delivery: automated testing and deployment so changes reach users quickly and safely.
Core Web Vitals
Google's metrics for real-world page experience: loading (LCP), responsiveness (INP) and visual stability (CLS). They influence search ranking.
Learn more about Core Web VitalsUsage-based pricing
Charging customers by how much they use, such as per seat, per API call or per document processed, instead of a flat fee.
Growth
Retention
The share of users who keep coming back over time, usually measured on day 1, day 7 and day 30. The best predictor of a product's long-term success.
Learn more about RetentionChurn
The rate at which customers or users stop using or paying for a product in a given period.
Activation
The moment a new user first experiences the core value of a product, the "aha moment". Shortening the path to it lifts retention.
Paywall
The screen that asks a user to subscribe or pay. Good paywalls show value first and are tested continuously.
A/B testing
Showing two versions of a screen or feature to different users and measuring which performs better on a chosen metric.
LTV (lifetime value)
The total revenue a customer is expected to generate over their relationship with a product. Compared with acquisition cost to judge growth.
[ 07 ] — new project
reply < 24h
Still sounds like jargon?
Tell us what you want to build in your own words. We'll translate.
no spam. no fluff. just shipping ✶