A Copilot is an AI assistant embedded within an existing application to support the user—without taking control away from them. The most well-known example is Microsoft Copilot, integrated into Word, Excel, Teams, and Outlook.

The concept: The AI works side-by-side with you. It drafts emails, summarizes meetings, creates presentations from notes, or analyzes spreadsheets—but you decide what gets adopted. A Copilot is a productivity tool, not a replacement for human judgment.

A Custom GPT is a tailored version of ChatGPT configured for a specific purpose—with its own system prompt, instructions, knowledge base, and tools.

Example: An ‘HR-GPT’ that knows a company’s personnel policies and answers employee questions about holidays, expenses, or labor law. Custom GPTs are available via OpenAI’s GPT Store or can be created internally for teams. They are a simple entry point into tailored AI but have limitations with complex integrations.

An AI Wrapper is an application that uses an existing AI model (e.g., GPT-4 or Claude) as its foundation and layers its own interface, additional logic, or specialized functions on top.

Many AI startups are essentially wrappers: they use the API of a large model and add a sleek user interface, industry-specific knowledge, or workflow automation. This is not negative—the added value often lies in the intelligent integration, not the model itself. At nuwai, we also build upon existing models and orchestrate them for specific business needs.

Multimodality means that an AI model works not only with text but also with other data types—so-called ‘modalities’. These include images, audio, video, and code.

Modern models like GPT-4o or Gemini are multimodal: they can describe images, analyze diagrams, understand speech, and even interpret video content. For businesses, this opens up possibilities such as automated document processing (scanning invoices), visual quality control, or analyzing customer videos.

Vibecoding is an informal term for writing software using natural language instead of traditional code. You describe in words what the software should do, and the AI generates the functional code.

The term was coined in early 2025 by Andrej Karpathy (ex-OpenAI, ex-Tesla). The idea behind it: With AI tools like Cursor, Claude Code, or GitHub Copilot, even non-programmers can create functional prototypes, websites, or automations. It’s less about technical perfection and more about fast, pragmatic implementation. For SMEs, this means many simple software problems can be solved faster and more affordably today than ever before.

Cursor and Claude Code are AI-assisted development tools that help programmers (and increasingly non-programmers) write software faster.

• Cursor is a code editor with deeply integrated AI. You can describe in natural language what you want to build, and the editor generates, explains, and refactors code in real-time.
• Claude Code is a command-line tool from Anthropic that can autonomously handle entire coding tasks—including file structure, debugging, and refactoring.

Both tools are core instruments in the modern AI-assisted software development process.