Decoder. plain-English AI glossary

Constraint

● Core

A restriction placed on the model — length, style, content rules — to shape output.

Think of it like

Like a painting competition: "Portrait only, max 500x500 pixels, pastels only."

Example

"Summarize in 2 sentences." "Do not use technical jargon." "Answer only with yes or no."

How it actually works

Constraints narrow the output space, making it easier for the model to satisfy requirements. Soft constraints (suggested) are easier to ignore than hard constraints (enforced). Some models support hard constraints via structured output or token limiting.

For product teams

Ensures outputs fit requirements (length, format, content rules).

For engineers

Add to instructions or enforce via output validation.

Instruction — Paired with instructions and format specs.

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open Decoder Add to Chrome