We're building structured prompt engineering. Instead of text strings, developers compose AI from decks of cards with examples and specs that make it reliable.
Our deck system brings structured engineering to LLM applications:
- Decks: Composable collections of cards that define specific AI behaviors
- Cards: Hierarchical specifications organized by category (persona, behavior, etc.)
- Specs: Clear, testable requirements that define precise AI capabilities
- Context: Structured information that guides AI understanding and responses
- Samples: Rated examples (-3 to +3) showing good and bad behaviors
Assistant cards turn AI instructions into testable specifications:
- Persona cards: Define who the AI should be (traits, voice, constraints)
- Behavior cards: Define what the AI should do (goals, steps, formats)
- Examples: Show good and bad behaviors with -3 to +3 ratings
- Specifications: Break down requirements into clear, testable units
- Company vision: Making LLMs 99% reliable through structured assistant cards
- Product plan: How we're building the card system
- Inference philosophy: Why inference-time control matters
- Team story: Who we are and why we're doing this
Contact: Questions? Email us at contact@boltfoundry.com