Top 100 Words And Jargons Used In Vibe Coding

Good morning, let's start this week strong!

Today I'm releasing a glossary of the top 100 most commonly used words and jargons in vibe coding and app-development using AI.

This should help people get a better understanding of things whenever they read an article or guide online about vibecoding with a bunch of jargons.

General Basics (1-20)

These cover the foundations of app development, whether coding traditionally or with AI vibes, helping vibe coders grasp how apps are structured.

1. API (Application Programming Interface): A set of rules that allows different apps or services to communicate and share data, such as pulling real-time weather updates into a travel app.
   

2. Backend: The server-side component of an app that manages data storage, business logic, and security—like the kitchen in a restaurant where orders are processed before reaching the customer.
   

3. Frontend: The client-side part of an app that users directly interact with, including visual elements like buttons, layouts, and animations on websites or mobile screens.

4. Input: Any data or instructions provided to a system or AI, such as typing a user query, uploading images, or entering form details to trigger app actions.
   

5. Output: The result or response generated by a system or AI, like displaying search results, rendering a generated image, or executing code in an app.
   

6. Context: Additional background information supplied to AI or systems to improve understanding and relevance, such as including previous chat messages or document excerpts in a prompt.
   

7. Database: A structured system for storing, retrieving, and managing app data efficiently, like a digital filing cabinet holding user profiles, product inventories, or transaction records.
   

8. Server: A powerful computer or cloud service that hosts an app's backend, processes requests, and delivers content to users over the internet for reliable access.
   

9. Client: The end-user's device or software, such as a web browser or mobile app, that sends requests to the server and displays the received data.
   

10. Code: Human-readable instructions written in programming languages like Python or JavaScript that direct computers to perform tasks, forming the blueprint of any app.
    

11. Bug: An unintended flaw or error in code that causes unexpected behavior, such as a button not working or data not saving correctly in an app.
    

12. Debugging: The systematic process of identifying, analyzing, and resolving bugs in code, often using tools like print statements or AI assistants to trace issues.
    

13. Version Control: A system for recording changes to code over time, enabling collaboration and recovery from mistakes, much like an advanced "undo" history in a document editor.

14. Repository (Repo): A centralized storage location, often online via platforms like GitHub, where a project's code files, history, and assets are kept organized and shared.
    

15. Commit: The act of saving a specific set of code changes as a permanent snapshot in version control, including a message describing what was modified.
    

16. Branch: A parallel version of the codebase created for developing new features or fixes without affecting the main project, allowing safe experimentation.
    

17. Merge: The process of integrating changes from a branch back into the main codebase, resolving any conflicts to create a unified version.
    

18. IDE (Integrated Development Environment): A comprehensive software application that combines code editing, debugging, and testing tools in one interface, streamlining the development workflow.
    

19. Library: A collection of pre-built, reusable code modules that can be imported into an app to add functionality quickly, such as image processing without writing it from scratch.
    

20. Framework: A pre-structured set of libraries and guidelines that provides a foundation for app development, speeding up creation by handling common tasks like routing or UI components.

    AI Concepts (21-40)

    
    Core ideas in AI that vibe coders use to guide tools like Claude or Gemini, explaining how AI interprets and responds to creative inputs.
    

21. AI (Artificial Intelligence): Technology that enables machines to perform tasks requiring human-like intelligence, such as recognizing patterns, generating content, or making decisions in apps.
    

22. ML (Machine Learning): A subset of AI where systems improve performance on tasks by learning from data patterns, rather than following rigid, hand-coded rules.

23. LLM (Large Language Model): A powerful AI trained on vast text data to comprehend, generate, and manipulate human language, powering tools like chatbots for app ideation.
    

24. Prompt: The initial text or query entered into an AI to instruct it on a task, such as "Design a fitness tracker app with daily goal reminders," which shapes the output.
    

25. Prompt Engineering: The art of refining prompts with specifics, examples, or structure to elicit more precise and useful AI responses, improving efficiency in vibecoding.
    

26. Fine-Tuning: Adapting a pre-trained AI model by further training it on specialized datasets, tailoring it for niche app features like custom voice recognition.
    

27. Generative AI: AI systems designed to produce original content from user inputs, such as creating code snippets, artwork, or text narratives for app prototypes.
    

28. Model: The core algorithmic structure of an AI, trained on massive datasets to specialize in areas like text generation or image analysis for app building.
    

29. Training Data: The diverse collection of examples, texts, or images used to educate an AI model, influencing its accuracy and biases in real-world applications.
    

30. Inference: The phase where a trained AI applies its knowledge to new inputs for quick predictions or generations, essential for real-time app interactions.
    

31. Token: The basic unit of text (e.g., words or subwords) that LLMs process, with usage limits affecting how much content can be input or output in a single interaction.
    

32. Context Window: The maximum amount of information (measured in tokens) an AI can consider at once for coherent responses, impacting long-form app planning.
    

33. Hallucination: An AI error where it generates plausible but inaccurate information, often due to gaps in training data, requiring fact-checking in app development.
    

34. Chain of Thought: A prompting technique encouraging AI to break down reasoning into logical steps, leading to more reliable outputs for complex problem-solving in apps.
    

35. Agent: An AI entity capable of autonomous actions, such as querying databases or executing code, beyond simple response generation in interactive apps.
    

36. Agentic AI: Evolved agents that not only respond but also plan multi-step workflows, like automating app testing or integrating external APIs independently.
    

37. Embedding: A mathematical representation converting text or data into dense vectors, allowing AI to measure semantic similarity for features like search in apps.
    

38. Vector Database: A specialized storage solution for embeddings that enables rapid querying of similar items, powering recommendation engines or content search in AI apps.
    

39. RAG (Retrieval-Augmented Generation): A method combining AI generation with real-time data retrieval from external sources, reducing hallucinations by grounding responses in facts.
    

40. Zero-Shot: An AI's ability to perform unfamiliar tasks solely from descriptive instructions, without prior examples, useful for quick app prototyping.

    Tools & Models (41-70)

    Popular AI tools, models, and software for vibecoding apps—focus on no-code/low-code vibes, with distinctions from similar options to highlight unique strengths.
    

41. Gemini 3: Google's advanced 2025 multimodal AI family excelling in text, code, and image processing for creative app ideation; it outperforms earlier models in reasoning depth.
    

42. Gemini 3 Pro: The premium variant of Gemini 3 for intricate coding tasks with a massive 1M token context window; unlike the lighter Flash version, it prioritizes depth over speed for professional app architecture.
    

43. Gemini 3 Flash: A cost-effective, rapid-response edition of Gemini 3 ideal for iterative prototyping and casual chats; it trades some Pro-level complexity for lower latency, making it better for quick vibe-coding sessions than heavier models.
    

44. Claude 4.5: Anthropic's 2025 family of safety-focused AI models strong in ethical coding and reasoning; it emphasizes helpfulness over raw creativity, differing from GPT-5's more experimental style.
    

45. Claude 4.5 Haiku: The swift, budget-friendly Claude model for lightweight tasks like generating basic UI code; unlike the more capable Sonnet or Opus, it sacrifices depth for 4-5x faster speeds in simple app tweaks.
    

46. Claude 4.5 Sonnet: A versatile mid-tier Claude model balancing speed and intelligence for daily vibecoding, such as app logic design; it offers better nuanced reasoning than Haiku but is more efficient than the resource-intensive Opus.
    

47. Claude 4.5 Opus: Anthropic's flagship Claude for demanding challenges like multi-file debugging; it excels in accuracy and creativity over Sonnet or Haiku, though slower and pricier, ideal for high-stakes app features.
    

48. GPT-5: OpenAI's groundbreaking 2025 model family for versatile chat, code, and multimodal app creation with built-in reasoning; it leads in creative freedom but may hallucinate more than safety-tuned Claude models.
    

49. GPT-5.2: An enhanced iteration of GPT-5 with superior speed and instruction adherence for expert tasks; unlike base GPT-5, it sets new benchmarks in efficiency, making it preferable for fast-paced app iterations over slower predecessors.
    

50. GPT-5.2-Codex: A coding-specialized offshoot of GPT-5.2 focused on practical software engineering; it generates more reliable, real-world code than general GPT variants, differing from non-Codex versions by emphasizing production-ready outputs.
    

51. ChatGPT: OpenAI's accessible conversational interface built on GPT models for app brainstorming and queries; unlike specialized tools like Claude Code, it's a generalist chat for ideation rather than direct code editing.
    

52. Claude Code: Anthropic's command-line tool leveraging Claude models to interpret and refactor existing codebases via natural language; unlike Lovable's from-scratch app building, it specializes in augmenting and debugging pre-written code for iterative improvements.
    

53. Claude Skill: Modular extensions for Claude AI, consisting of instruction sets or scripts for tailored workflows like app-specific automation; they differ from base Claude Code by adding reusable, plug-and-play capabilities across non-coding tasks too.
    

54. Lovable: An AI-driven no-code builder that translates natural language descriptions into complete, editable apps with backend integration; unlike Claude Code's focus on code modification, Lovable generates full prototypes without requiring any coding knowledge.
    

55. Vibe-Coding: A relaxed approach to development where users describe app concepts in conversational "vibes" for AI to produce code; it contrasts with traditional coding by prioritizing intuition and iteration over syntax mastery.
    

56. GitHub Copilot: Microsoft's AI autocomplete tool that suggests inline code snippets while typing in editors; unlike the more conversational Cursor, it focuses on real-time completions rather than project-wide discussions or overhauls.
    

57. Cursor: An AI-enhanced IDE forked from VS Code, enabling natural language chats for code generation and multi-file edits; it stands out from GitHub Copilot by offering proactive, holistic project refactoring instead of just snippet suggestions.
    

58. Aider: An open-source terminal assistant that uses any LLM to make git-aware code changes via prompts; model-agnostic unlike Claude Code's Claude exclusivity, it automates commits for collaborative workflows in a lightweight CLI environment.
    

59. VS Code (Visual Studio Code): Microsoft's extensible, free code editor serving as a base for AI integrations like Copilot; it differs from full AI IDEs like Cursor by being a neutral platform that hosts plugins rather than embedding AI natively.
    

60. Replit: A cloud-based IDE running entirely in the browser for instant code execution and collaboration; unlike local tools like VS Code, it eliminates setup hassles, perfect for testing AI outputs on the fly without downloads.
    

61. Hugging Face: A community hub for downloading, sharing, and fine-tuning open-source AI models; it contrasts with proprietary platforms like OpenAI by offering free, customizable access for experimenting with non-commercial app AI.
    

62. Flutter: Google's UI toolkit for crafting natively compiled apps across platforms from a single codebase; unlike React Native's JavaScript roots, Flutter uses Dart for pixel-perfect, high-performance interfaces, often AI-generated for rapid prototyping.
    

63. React Native: Meta's framework for developing mobile apps using React web skills and JavaScript; it differs from Flutter by bridging web and native code more seamlessly but with potentially less consistent UI rendering across devices.
    

64. Node.js: A runtime environment for executing JavaScript on servers, enabling full-stack development; unlike browser-only JS, it powers scalable backends for AI apps, often paired with frameworks like Express for API handling.
    

65. Docker: A platform for containerizing apps into portable units that run consistently across environments; it simplifies deployment compared to bare servers, ensuring AI-built prototypes work identically from dev to production.
    

66. Firebase: Google's serverless backend service providing real-time databases, authentication, and hosting; no-code friendly unlike full servers like AWS, it's ideal for quick AI app prototypes needing instant scalability.
    

67. Postman: A collaborative tool for designing, testing, and documenting APIs through visual requests; it verifies AI-generated endpoints more intuitively than command-line alternatives, catching integration issues early.
    

68. Figma: A web-based collaborative design tool for wireframing and prototyping app interfaces; unlike simpler tools like Canva, it supports interactive prototypes and developer handoffs, often refined after AI descriptions.
    

69. TensorFlow: Google's open-source library for creating and deploying custom machine learning models in apps; more ecosystem-integrated than PyTorch for production, it's suited for embedding AI features like predictive analytics.
    

70. PyTorch: Meta's dynamic library favored for AI research and flexible model prototyping; it offers easier experimentation than TensorFlow's static graphs, appealing for vibe-coders tweaking models iteratively.

    Coding Essentials (71-85)

    Basic building blocks vibe coders need when reviewing or tweaking AI code, providing insight into how generated outputs function.

71. Variable: A symbolic name in code assigned to store and manipulate data values, such as a user's age or app settings, which can be updated dynamically.
    

72. Function: A self-contained block of code that encapsulates a task, like calculating totals, and can be invoked repeatedly with different inputs for modularity.
    

73. Loop: A control structure that executes code repeatedly based on a condition, such as iterating through a list of user inputs to process each one efficiently.
    

74. Conditional: Logic statements like if-else that branch code execution based on true/false evaluations, enabling decisions such as showing error messages for invalid data.
    

75. Class: A blueprint defining the structure and behavior for objects in object-oriented programming, grouping related data and methods like a user profile template.
    

76. Object: An instantiated version of a class with specific values, representing real entities in an app, such as a particular user's account derived from the class.
    

77. Inheritance: A mechanism allowing a new class to inherit properties and methods from an existing one, promoting code reuse like extending a base "Button" class for custom variants.
    

78. Polymorphism: The ability for different classes to implement the same method uniquely, allowing flexible code like various shapes responding differently to a "draw" command.
    

79. Algorithm: A precise sequence of steps to accomplish a goal, such as searching or sorting data, forming the logical core of efficient app features.
    

80. Data Structure: A format for organizing data to optimize access and manipulation, like stacks for last-in-first-out operations or queues for orderly processing.
    

81. Array: A fixed-size, indexed collection holding multiple values of the same type, useful for lists like app menus accessed by position numbers.
    

82. Dictionary: An unordered collection of key-value pairs for fast lookups, akin to a real dictionary where keys like "username" map to values like "john_doe."
    

83. Recursion: A technique where a function solves a problem by calling itself with simpler inputs, ideal for tasks like traversing tree-like app structures.
    

84. Exception Handling: Mechanisms like try-catch blocks to detect and respond to runtime errors gracefully, preventing app crashes from issues like network failures.
    

85. REST: An architectural style for web services using standard HTTP methods (GET, POST) for stateless communication, making APIs simple and scalable for app integrations.

    Benchmarks & Advanced Jargon (86-100)

    Tests measuring AI performance, plus niche terms for scaling vibecoded apps, helping evaluate tool reliability and app growth potential.
    

86. SWE-bench Verified: A rigorous benchmark assessing AI on resolving real GitHub software engineering issues, measuring practical coding prowess with verified fixes.
    

87. WebDev Arena: A competitive evaluation ranking AIs on end-to-end web app development, from UI design to deployment, highlighting strengths in full-stack tasks.
    

88. LMArena: A crowdsourced leaderboard comparing AI models on conversational quality, reasoning, and task completion across diverse scenarios.
    

89. GDPval: A challenging test probing AI on graduate-level domain problems in fields like math or science, gauging depth for specialized app intelligence.
    

90. Stack: The layered combination of technologies forming an app's architecture, such as a frontend React with Node.js backend, dictating development choices.
    

91. Full-Stack: Expertise or development covering both frontend user interfaces and backend data handling, enabling complete app creation without specialization gaps.
    

92. Microservices: An app architecture dividing functionality into independent, loosely coupled services for easier scaling and maintenance than monolithic designs.
    

93. Monolith: A unified app structure where all components are tightly integrated in one codebase, simpler for small projects but riskier for large-scale updates.
    

94. Cloud Computing: Delivering computing resources like storage and processing via internet-based services, allowing apps to scale dynamically without physical hardware.
    

95. Edge Computing: Processing data near the source (e.g., user devices) to minimize latency, contrasting cloud's centralized approach for real-time apps like gaming.
    

96. SEO (Search Engine Optimization): Strategies to enhance app or web visibility in search results, such as keyword optimization, boosting user discovery.
    

97. Accessibility (A11y): Design principles ensuring apps are usable by people with disabilities, like screen reader support, promoting inclusivity and legal compliance.
    

98. Encryption: Converting data into a secure, unreadable format using algorithms, protecting sensitive app information like passwords during transmission or storage.
    

99. Open Source: Software with publicly available source code under licenses allowing free use, modification, and distribution, fostering community-driven innovation.
    

100. Scalability: The capacity of an app or AI system to handle increased load, such as more users or data, through efficient design without performance degradation.

There you have your vibe coding dictionary!

Print it, bookmark it, or keep it open while you build.

If you're new like I was, start with a tool like Lovable since it’s super easy to use.

Keep vibing,

Brian Hanson