Module 6.1 · Topic 2

System Prompts and Instruction Layers

Bottom Line Up Front: AI tools run on multiple instruction layers: system prompts (platform defaults), tool-level instructions, and custom instructions (your preferences). Understanding all three gives you control.

2.1 The Hidden Instructions That Shape AI Behavior

Every AI model runs on hidden system prompts that set safety boundaries, output format preferences, and operational rules.

The platform writes system prompts; you cannot change them directly. Tool refusals often stem from system prompts, not model incapacity.

See Module 2.1 for prompting techniques that work within these layers.

Your leverage is in upper layers—configurations and custom instructions.

2.2 How Platform Defaults Influence Every Interaction

Platform defaults shape response length, tone, and safety sensitivity. Most users accept defaults as-is.

Fast/Chat Tier Fast/Chat — Favors quick, conversational responses. Use for exploratory conversations.
Thinking/Reasoner Tier Thinking/Reasoner — Encourages structured, step-by-step analysis. Use for substantive work.
Deep Research Tier Deep Research — Prioritizes thoroughness and source citation. Use for authoritative research.

Unlock the Full AI Skill Building Experience

Visit LawQi for access to the full AI Skill Building experience.

Visit LawQi

40% discount using code REASONABLE for personal subscriptions.

Free 48-hour preview access when investigating for teams.