Decoder. plain-English AI glossary

Least-to-most prompting

▲ Rising

Breaking a hard problem into easier sub-problems, solving simplest first, then using solutions to tackle harder ones.

Think of it like

Learning to multiply: first practice single-digit facts, then two-digit numbers, then large multiplication.

Example

For "write a 1000-word essay," first write a 1-sentence outline, then a 1-paragraph draft, then expand to 1000 words.

How it actually works

Least-to-most is effective because it mitigates error accumulation. Solve the base case well, then leverage it. A model that struggles with a hard task may excel if broken into easy substeps. Structured prompting can automate this decomposition.

For product teams

Surprisingly effective for complex tasks. Minimal prompt overhead.

For engineers

Manually decompose tasks. Use a second model to auto-decompose if possible.

Decomposition prompting — Related to decomposition and step-by-step reasoning.
Prompt chaining — Related to prompt chaining.

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