Plain-English field lexicon
12 terms · 3 clusters
Familiar analogy, precise definition, and a real example for the language behind agentic systems—without sanding off the technical meaning.
What the model is, underneath — before any talk of "agents".
Sophisticated autocomplete — the same idea as your phone guessing the next word, trained on far more text and far better at it.
A "large language model": software that predicts the most likely next chunk of text from patterns it learned in training. It generates language; it does not look things up.
Ask it for a quote and it produces something that sounds right — because it is predicting plausible text, not retrieving a stored record.
A chunk of text — roughly three-quarters of a word. The unit the model reads and bills in.
The fragments a model breaks text into. Around 100 tokens ≈ 75 English words. Limits and pricing are counted in tokens, not words or characters.
"strawberry" can be three tokens — which is part of why a model can miscount the letters in it.
The model's working memory — a desk. Pile on too many papers and the earliest ones slide off the edge.
Everything the model can see at once: your prompt, the conversation so far, attached files, and its own reply. Past the limit, earlier content is dropped — and nothing persists once the session ends.
In a long chat it "forgets" what you said at the start — that text fell off the desk.
The brief. Closer to writing instructions for a new contractor than to programming.
The text you give the model — question, instructions, context. Its quality shapes the quality of the output.
"Summarise this" versus "Summarise this in three bullets for a CFO" pull sharply different results from the same model.
What turns a chatbot that answers into something that does things.
A worker you delegate to: brief it, and it takes steps on its own. A chatbot answers; an agent acts.
An LLM that runs tools in a loop toward a goal — it acts, checks the result, decides the next step, and repeats until done.
"Fix the failing test": it reads the test, edits the code, re-runs the suite, and stops once it passes.
The adjective for "acts like an agent" — multi-step and self-directed, not one question and one answer.
Describes AI that plans and takes a sequence of actions with some autonomy, adapting as new information comes back.
An "agentic workflow" books the whole trip end to end; a chatbot only tells you how to book it.
Letting the model pick up the phone. On its own it can only write text; a tool lets it do something in the real world.
Also called "function calling": the model requests an action from outside software — search the web, read a file, hit an API — and the result comes back into its context.
Asked for today's weather, it calls a weather tool instead of guessing.
USB-C for AI — one standard plug so any agent can connect to any data source or tool.
The Model Context Protocol: an open standard (Anthropic, 2024) for connecting AI systems to external tools and data without building a custom integration for each one.
An MCP server for your docs lets an agent read them without bespoke glue code.
The scaffolding, the guardrails, and why it sometimes makes things up.
The cockpit around the engine. The model is the engine; the harness is everything that makes it useful and safe to fly.
The runtime around the model — the loop, tool access, memory, prompts, and guardrails. The model reasons; the harness does everything else.
Two products built on the same model can behave nothing alike because their harnesses differ.
Letting it check the notes before answering, instead of going from memory.
Retrieval-Augmented Generation: fetch relevant documents at question time and have the model answer from them. It reduces made-up answers — it does not eliminate them.
A support bot that quotes your actual help docs rather than inventing a policy.
A dial, not a switch — from "asks before every move" to "acts and reports back".
How much the agent does without you: in-the-loop (you approve each action), on-the-loop (it acts, you monitor), or out-of-the-loop (fully autonomous).
An agent that asks before deleting files is in-the-loop; one that opens a pull request for your review is on-the-loop.
A confident bluff. It is generating plausible text, so "I don't know" rarely comes naturally.
When a model states something false as if it were true — because it predicts likely text rather than retrieving verified facts.
It cites a court case that does not exist: fluent, well-formatted, and entirely invented.