Matt Pocock open-sources "dictionary-of-ai-coding": demystifying AI coding terms to help engineers onboard efficiently

In May 2026, TypeScript educator and AI Hero founder Matt Pocock published the open-source dictionary-of-ai-coding project on GitHub. This glossary explains many professional terms and jargon in AI coding in clear, accessible English, helping developers cross the “expert threshold” and avoid common pitfalls of "vibe coding". The repository sparked discussion in developer communities and is seen as an important educational supplement for programming in the AI era.

In the README, Pocock states plainly: "AI coding often feels like it's reserved for experts." Developers faced with new concepts such as RAG, Agentic Workflow, Evals, and Tracer Bullets often struggle to communicate and practice efficiently due to terminology barriers. This dictionary aims to address that pain point—translating AI-coding vocabulary into language that ordinary engineers can easily understand.

Key contents and design highlights

The project includes dozens of core terms commonly used in AI-coding practice, covering prompt engineering, building AI agents, evaluation systems, code-generation workflows, and more. Each entry seeks to be concise and practical, with real-world examples. Pocock emphasizes that the dictionary is more than a glossary; it functions as an "on-ramp" to AI coding, helping developers form a systematic understanding.

Typical entries include:

Vibe Coding: Letting AI generate code based on intuition or "feeling" without rigorous engineering review, often leading to maintenance problems later.
Evals: Systematic methods for evaluating AI-generated code or model outputs; a key practice for ensuring quality.
Agentic Harnesses: Frameworks or toolchains for building AI agents that can autonomously complete complex tasks.
Tracer Bullets: A rapid prototyping approach that builds the smallest viable version to test core assumptions.

The repository uses an open collaboration model, encouraging community contributions of new entries and improved explanations, reflecting open-source values. Pocock also designed the format to be suitable for direct inclusion in LLM contexts, so AI assistants can reference standard definitions during coding and reduce hallucinations.

Pocock is a well-known TypeScript instructor, a former developer advocate at Vercel, and founder of the AI Hero Newsletter (with over 62,000 subscribers). He says the dictionary grew out of observations from teaching and AI engineering practice: many developers can rapidly build prototypes with AI tools, but unclear terminology leads to issues in code quality and maintainability.

Impact on the AI-coding ecosystem and developers

1. Lowering the learning barrier for AI coding

With the spread of AI coding tools (such as Cursor, Claude Code, GitHub Copilot), more developers are entering the field. But the explosion of terminology has created a new "digital divide." The dictionary offers a standardized, easy-to-understand reference to help beginners and career-changers get up to speed, accelerating the shift from "using AI" to "mastering AI engineering." This is especially helpful for SMEs and individual developers, lowering the cost of entering the AI era.

2. Improving team collaboration and communication efficiency

In AI-driven development teams, differing interpretations of the same term can cause misunderstandings and rework. A unified dictionary builds a common language base and facilitates effective dialogue among product managers, engineers, and AI specialists. This helps teams iterate faster, reduce errors, and establish more standardized AI engineering processes.

3. Promoting engineering di