TypeScript educator Matt Pocock released a comprehensive glossary on May 1, 2026, explaining AI coding terminology in accessible language. The GitHub repository, which gained 372 stars shortly after release, addresses what Pocock describes as a deliberate knowledge barrier: "a whole VC-funded economy benefits from keeping it hard to understand."
70 Terms Organized Across Seven Practical Sections
The dictionary organizes approximately 70 terms across seven sections covering the AI coding landscape. Section 1 covers the model itself, including parameters, training, inference, tokens, next-token prediction, model providers, and prefix caching. Section 2 addresses sessions and context, explaining stateless versus stateful systems, context windows, agents, system prompts, sessions, and turns.
Section 3 focuses on tools and environment, covering filesystems, tool calls, tool results, permission requests, agent modes, and sandboxes. Section 4 details failure modes like hallucination, knowledge cutoffs, attention degradation, and the "smart zone." The remaining sections cover handoffs, memory systems, and work patterns including human-in-the-loop and "vibe coding."
Practical Definitions Bridge Traditional and AI-Assisted Development
Pocock defines "vibe coding" as a working pattern where users accept agent-generated code without human review, treating the diff as opaque and focusing on program behavior rather than code quality. He explains "training" as the one-time expensive process where model providers set parameters by exposing models to vast amounts of text, while "inference" is the relatively cheap process of running the trained model for each request.
The glossary also introduces "design concept," drawing on Fred Brooks' work in "The Design of Design", as the shared understanding between user and agent about what's being built. This bridges traditional software engineering principles with AI-assisted development.
Educator Background Shapes Accessible Approach
Matt Pocock, known for TypeScript education and developer advocacy, created the resource as part of his broader AI Hero initiative at aihero.dev, where he teaches AI coding courses. The repository shows 29 commits on the main branch, indicating active development and refinement of definitions based on community feedback.
Key Takeaways
- Matt Pocock released a comprehensive AI coding glossary on May 1, 2026, organizing approximately 70 terms across seven practical sections
- The repository gained 372 GitHub stars shortly after release, indicating strong developer interest in demystifying AI coding terminology
- The dictionary covers fundamental concepts from model training and inference to practical work patterns like "vibe coding"
- Pocock argues that AI coding complexity is partly maintained by industry interests benefiting from knowledge barriers
- The resource bridges traditional software engineering concepts with AI-assisted development practices