dictionary-of-ai-coding

A dictionary that translates the vocabulary of AI coding into plain English, from what a model actually is to why the bill keeps climbing.

The README opens with the complaint that AI coding can feel like it is just for experts, with unexplained jargon, mysterious failures, and bills that do not match the work. Then it claims much of that confusion is manufactured, kept hard by an economy that benefits, and that the basic terms are learnable in an afternoon.

The AI entry calls AI a moving label rather than a technology, pointing at whatever computers can newly and impressively do. A table tracks what the word meant across eras, from symbolic reasoning in the 1950s to large language models in the 2020s, and the AI effect explains why solved techniques get renamed.

The Model entry defines a model as the parameters, stateless, doing next token prediction and nothing else. It cannot read files, run commands, or remember yesterday, and everything that feels like an agent working is the harness orchestrating many predictions in a row.

The README advises swapping the word AI for the precise term that is meant, the model, the harness, the agent, or the context it was given, since a claim like AI can't reason carries a hidden timestamp. Being strict about the word also sharpens diagnosis of disappointing output.

The dictionary points readers who want more than vocabulary to a newsletter, noting more than 62,000 developers. GitHub lists TypeScript as the primary language and no license identifier for the repository.

The dictionary's point is that the confusion is manufactured, not inevitable, and the terms are learnable. This review summarizes the README's framing without vouching for every entry.

• Source 1 · Project repository
• Source 2 · README