Learn AI · Reference
AI & terminal glossary, no jargon required.
Every term you will meet, one line each. Bookmark this page — it is the Rosetta stone for the whole site.
A–B
| Term | Meaning |
|---|
| Agent | An AI that can take actions — read files, run commands — not just answer. |
| API | A way for programs to talk to each other; how a terminal reaches a model. |
| Backend | The server side of a product; where auth and token pools live. |
| Benchmark | A standard test used to compare model performance. |
C–E
| CLI | Command-line interface: the terminal program you type into. |
| Context window | How much text a model can "see" at once, in tokens. |
| Deep research | A workflow that gathers sources and produces a cited brief. |
| Dependency | An extra program a tool needs. "Zero dependencies" = install-and-go. |
| Endpoint | A specific URL that accepts requests, e.g. the gateway. |
| Extension | A small add-on that gives the terminal new skills. |
F–L
| Fine-tuning | Extra training to specialize a model for a task. |
| Foundation model | A large general-purpose model (e.g. Claude, GPT, Gemini). |
| Free tier | A usage level with no cost (e.g. Community tune is ₹0). |
| Hallucination | Confidently wrong output. Great tools reduce it with citation. |
| Inference | Running a model to produce an output (vs training it). |
| LLM | Large language model — the engine behind the text. |
| Local / local-first | Running on your own machine rather than a cloud. |
M–R
| Model | A trained AI that turns prompts into outputs. |
| Multi-agent | Several AI roles cooperating on one task. |
| Multi-model | One tool that switches between many model families. |
| MCP | Model Context Protocol — a standard for connecting tools to models. |
| Obsidian | A notes app; the terminal can write and search your vault. |
| Prompt | The message you send to a model. |
| Provenance | Record of where a claim came from — a sidecar file shows your sources. |
| RAG | Retrieval-augmented generation: citing retrieved knowledge rather than guessing. |
S–Z
| Sidecar | An extra file beside an output recording how it was made. |
| Stack | The set of technologies a product runs on. |
| Token | The unit of text a model reads/writes; ~1–2 per word. |
| Token pool | Your monthly allowance of free/Pro tokens. |
| Workflow | A described multi-step task the AI runs (lit, audit, etc.). |
| Zero retention | Not storing your data after the session — privacy by design. |
Plus the recurring ones you will see across this site: Expert AI (the x CLI — this site's product), Expert AI Terminal Pro (separate $195 desktop/USB product, mentioned only for clarity), Unlimited plan (₹800/mo, 530M tokens), pass (6-hour token bundle), route (the lane a model uses, e.g. ssv.asia/free).
Products, cleanly separated: x = Expert AI (free-to-start CLI). Expert AI Terminal Pro = a paid desktop/USB terminal. Expert AEC Terminal = the AEC knowledge community. This site is about the first one.
Put the terms to work
