The fundamentals of Agentic Coding (AKA Vibe Coding) - Theodor René Carlsen - NDC Copenhagen 2026

Summary

The talk demystifies 'agentic coding' by establishing a fundamental baseline for understanding AI-assisted development tools. The core concept is that these systems rely on three components: the models (the brain), the harnesses (the ability to act), and the tools (specific functions). Speakers emphasize that while the ecosystem moves rapidly, understanding this architecture—especially the feedback loop where the harness executes actions based on model intentions—is crucial for practitioners. A key recommendation is maintaining control by favoring open-source, customizable systems over locked-down, proprietary solutions.

Technical details

• Agent Architecture 310s

  An agent requires three components: the model, a harness (the execution layer), and tools. The process involves the model generating an intention, which the harness executes (e.g., reading files using the OpenAI JavaScript SDK). This output is then fed back into the model for continued processing.

• Context and Prompting 560s

  Effective use of agents requires providing rich context. Instead of relying solely on general prompts, developers should embed project-specific knowledge (e.g., domain language, setup commands) in files like `agents.md` or skills definitions to guide the model.

• Data Format for AI Interaction 480s

  Since models are trained on text and output text, any problem or data structure (like dependency graphs) that can be converted into a markup language format is highly readable and understandable by the agent.

Mentioned resources

• OpenAI JavaScript SDK (Tooling/Library)
• GitHub CLI (Command Line Tool)
• agents.md (Configuration File/Documentation)

Channel & topics

• NDC Conferences
• Agentic Coding
• AI Development Tools
• Model Architecture
• Open Source Software
• Context Engineering
• OpenAI JavaScript SDK
• GitHub CLI
• agents.md