GLOSSARY · A–Z

AI vocabulary, plainly.

Every term you’ll bump into using LADLE, defined without jargon. If it’s not here and you think it should be, email us.

A4 TERMS

AgentAn AI system that takes multiple steps toward a goal — typically by using tools, deciding what to do next, and iterating without human input at each step.AlignmentThe technical and philosophical problem of ensuring AI systems reliably do what humans want them to do — including cases where 'what humans want' is complex or contested.APIAn interface that lets one program talk to another — in AI, usually the way developers send prompts to a model and receive responses programmatically instead of through a chat UI.Answer engineA search product that returns direct answers with cited sources, rather than a list of links to click through — Perplexity, ChatGPT search, Claude's citations, Google's AI Overviews.

B1 TERM

BenchmarkA standardized test used to compare AI models on specific tasks — like MMLU for general knowledge, HumanEval for code, or GPQA for graduate-level reasoning.

C2 TERMS

Context windowThe amount of text a language model can consider in a single request — measured in tokens, roughly three-quarters of a word each.Chain of thoughtA prompting technique (or trained behavior) where the model produces intermediate reasoning steps before its final answer, which measurably improves accuracy on hard problems.

E1 TERM

EmbeddingA vector of numbers that represents a piece of text (or image, or audio) such that semantically similar items produce numerically similar vectors.

F2 TERMS

Fine-tuningAdjusting a pretrained model's parameters on additional data to specialize it for a particular task or style.Frontier modelA term for the most-capable AI models available at any given time — those pushing the leading edge of capability, typically trained with the most compute.

G1 TERM

GEOGenerative engine optimization — the practice of structuring content so AI answer engines can quote or summarize it accurately.

H1 TERM

HallucinationA confidently-stated but false claim produced by an AI model — a fabricated citation, an invented statistic, a made-up event.

I1 TERM

InferenceThe act of running a trained model to produce output — as opposed to training, which is the earlier one-time process of creating the model.

L2 TERMS

LLMA large language model — a neural network trained on huge amounts of text that predicts the next chunk of text given what came before.LatencyThe time between sending a request and getting a response — for AI chat, split into 'time to first token' (perceived latency) and 'total response time'.

M1 TERM

MultimodalAn AI system that processes more than one kind of input — text plus images, or text plus audio, or all four modes together.

O1 TERM

Open weightsA model whose trained parameters are publicly downloadable — as opposed to closed-weights models that are only accessible through the maker's API.

P1 TERM

Prompt engineeringThe practice of writing inputs to a language model in ways that reliably produce good outputs — a real skill, and a smaller one than the industry has commercialized it as.

Q1 TERM

QuantizationA technique that reduces the numerical precision of a model's parameters — trading small amounts of quality for large reductions in memory and compute.

R2 TERMS

RAGRetrieval-augmented generation — an AI pattern that searches a document store for relevant passages before generating an answer, so the answer is grounded in real content.Rate limitA cap on how many requests you can make to a service in a given time window — enforced to protect the service from overload and to give predictable capacity to all users.

S2 TERMS

StreamingSending model output to the user token-by-token as it's generated, rather than waiting for the full response before rendering.System promptThe instructions the model receives before any user messages — used by the product operator to set behavior, tone, and guardrails.

T2 TERMS

TokenThe unit a language model uses to read and write text — roughly three-quarters of a word on average, though common words are one token and rare ones may be several.TemperatureA parameter controlling how random the model's next-token choice is — higher temperature produces more varied output, lower produces more deterministic output.

V1 TERM

Vector databaseA database specialized for storing and searching vector embeddings — designed to answer 'find the N vectors closest to this query vector' very fast, at scale.

← Practical guides

0meals fundedROLLING TOTAL · UPDATED HOURLY

SEE THE RECEIPTS →