System Prompt
A system prompt is the set of instructions given to a language model before the user's input—defining the model's persona, constraints, output format, and behavioral rules. In the OpenAI and Anthropic APIs, it occupies the "system" role in the message array. The user never sees it, but it shapes everything the model does. System prompts are where enterprises encode their business logic: "You are a customer support agent for Acme Corp. You may only reference information from the provided knowledge base. Never discuss competitor pricing. Always respond in the customer's language." A well-written system prompt is the cheapest, fastest way to control model behavior. A poorly written one is why your AI chatbot told a customer they could get a full refund on a non-refundable ticket. System prompt engineering is a core skill that most teams underinvest in.
Related terms:
Private Tokens
Proprietary organizational data and institutional knowledge that generic AI can’t access—encompassing conversational transcripts, internal documentation,...
Pace Layers
Pace layers separate complex systems into fast-changing parts that learn and slow-moving parts that stabilize, helping teams choose experiments vs. strategy.
Inference
Inference is the process of running a trained model on new input to generate a prediction or output—such as sending a prompt to GPT-4 and receiving a...