Skip to content
Neural
Gist
AI news, explained
RSS
Compared
How-to
Trackers
Glossary
Glossary
Plain definitions for the words everyone uses and nobody explains.
19 articles
A
C
E
F
G
L
M
N
P
Q
R
T
A
AI agent
A model that can take actions in a loop — and the word the industry has worn out.
AI hallucination
When a model states something false with total confidence — and why that is the normal case, not a glitch.
C
chain-of-thought prompting
Asking for the working, and getting a better answer as a side effect.
context window
The model's working memory — and the reason a long chat starts forgetting things.
E
embedding
Turning meaning into coordinates — the trick that makes semantic search work.
F
fine-tuning
Teaching a model how to behave — not what is true.
G
GGUF file
The file format behind every local model you have downloaded — and how to read its name.
L
LLM benchmark
A standardised exam for models — and a number to treat with suspicion.
M
MCP (Model Context Protocol)
A standard plug for connecting AI models to your tools and data.
mixture of experts (MoE)
Why a "large" model can be fast: most of it sits out any given word.
N
NPU
The chip your laptop and phone now advertise — what it's for, and why its headline "TOPS" number tells you less than it looks.
P
prompt caching
Paying once for the part of your prompt that never changes — and the write fee nobody mentions.
prompt injection
The unsolved security problem at the bottom of every AI product that reads anything.
Q
quantization
Shrinking a model so it fits on hardware you own — and what you give up.
R
RAG (retrieval-augmented generation)
Giving the model the documents instead of hoping it memorised them.
RLHF
The training step that turned a text predictor into something worth talking to.
T
token
The unit models read, count and bill in — and it is not a word.
tool use
Letting the model call your code — the feature that turns a chatbot into software.
transformer
The 2017 architecture underneath essentially every model you have heard of.