---
id: "concept-context-sandwich"
type: "concept"
source_timestamps: ["01:30:00"]
tags: ["prompt-engineering", "best-practices", "workflow-design"]
related: ["framework-context-sandwich", "action-build-context-sandwich", "prereq-basic-prompting", "entity-igor-pogany"]
definition: "A prompting framework that surrounds a core task with deep personal context and explicit criteria for a successful output."
---
# The Context Sandwich (Concept)

## Definition
A prompting framework that surrounds a core task with deep personal context and explicit criteria for a successful output.

## Why Most Users Fail With AI
Most users fail with AI because they only provide the *meat* — the raw task itself (e.g., "write an email"). Generic prompts produce generic outputs.

## The Three Layers
- **Top Bun — Who I am.** Deep context about the user's business, audience, tone, and background.
- **Meat — What I need.** The specific task to perform.
- **Bottom Bun — What good looks like.** Exact criteria for success: format, length, constraints, emotional/strategic outcome.

For the structural template, see [[framework-context-sandwich]]; for an action plan, see [[action-build-context-sandwich]].

## Effectiveness
Mirrors industry-standard "system + user + few-shot example" structures. Anthropic and OpenAI documentation report 2–3x accuracy improvements over raw prompts. Conceptually adjacent to Chain-of-Thought (Wei et al., 2022) and Tree-of-Thoughts prompting.

## Prerequisites
Requires [[prereq-basic-prompting]] and [[prereq-llm-access]].

## Connected Notes
- Framework form: [[framework-context-sandwich]]
- Action: [[action-build-context-sandwich]]
- Synergy with branding: [[concept-brand-dna]]