karpathy-system-prompt-learning

Skill 13: System Prompt Learning（系统提示学习范式）

Source: https://x.com/karpathy/status/1921368644069765486 "System prompt learning — missing LLM learning paradigm"

Core Principle

The system prompt is a textbook. Write it like one.

Karpathy's insight: there are three learning modes for LLMs (pretrain, finetune, RL) — but there's a fourth that's massively underused: system prompt learning. Explicitly encoding strategy into the system prompt is more sample-efficient than retraining and more reliable than hoping the model figures it out.

Think of it as writing a textbook for the LLM before each task. Not just instructions — strategy with reasoning.

The System Prompt Learning Framework

Good system prompts have three layers (mirroring how humans learn):

Layer 1: EXPOSITION (pretrain equivalent)
→ Background knowledge the model needs
→ Relevant concepts, terminology, context
→ "Here's what you need to know about [domain]"

Layer 2: WORKED EXAMPLES (SFT equivalent)  
→ Step-by-step demonstration of correct behavior
→ Explicit reasoning traces
→ "Here's exactly how to handle [situation]"

Layer 3: STRATEGY (RL equivalent)
→ Explicit decision rules
→ Error patterns to avoid
→ "When you see X, do Y because Z"

Building a System Prompt from Scratch

Build a system prompt that teaches an LLM to [TASK] reliably.

Task description: [WHAT THE LLM SHOULD DO]
Current failure mode: [HOW IT CURRENTLY GOES WRONG]
Desired behavior: [EXACTLY WHAT GOOD LOOKS LIKE]

Structure the system prompt with all three learning layers:

Layer 1 — Exposition:
Write 2-3 paragraphs of background knowledge the LLM needs. 
Include: key concepts, relevant context, domain vocabulary.

Layer 2 — Worked Examples:
Write 2 complete examples of the task done correctly.
Format: INPUT → [step-by-step reasoning] → OUTPUT
Show the thinking, not just the answer.

Layer 3 — Strategy:
Write explicit decision rules as IF/THEN/BECAUSE statements.
Include: common mistakes to avoid, edge cases to handle, quality checks.

End with: "Before responding, verify: [checklist of 3-5 quality checks]"