Superblocks CEO Reveals AI System Prompts as Billion Dollar Startup Goldmine

Brad Menezes, CEO of Superblocks, an enterprise AI coding startup, believes the next wave of billion-dollar startup ideas is hidden in plain sight: the system prompts used by existing AI unicorns. These system prompts are extensive instructions—often over 5,000 words—that guide foundational AI models like OpenAI’s or Anthropic’s to perform specific application-level tasks.

Unlike typical prompts, system prompts serve as a master class in prompt engineering. Each AI startup crafts unique prompts tailored to their domain and task requirements, effectively teaching the AI how to behave like a specialized expert. While these prompts aren’t always public, Superblocks recently shared 19 system prompts from popular AI coding tools to foster learning and innovation.

Menezes emphasizes that the system prompt itself is only about 20% of the secret sauce. The remaining 80% lies in “prompt enrichment”—the infrastructure built around the AI calls, including additional instructions and response validation to ensure accuracy and relevance.

Decoding System Prompts: Role, Context, and Tools

Studying system prompts reveals three critical components:

• Role Prompting: Assigns the AI a consistent purpose and personality, instructing it to behave like a skilled human expert. For example, one prompt begins with “You are Devin, a software engineer who is a real code-wiz...”
• Contextual Prompting: Provides the AI with guardrails and task-specific context to improve clarity and reduce errors or unnecessary actions. For instance, instructing the AI to only call tools when necessary and avoid showing code unless requested.
• Tool Use: Enables the AI to perform agentic tasks beyond text generation, such as editing code, querying databases, or executing shell commands, enhancing its practical utility.

By analyzing these prompts, Menezes identified different focuses among AI coding tools: some prioritize fast iteration, while others emphasize creating full-stack applications with raw code output. Superblocks sees an opportunity to empower non-developers by handling complexities like security and enterprise data integration.

Superblocks’ own AI agent, Clark, enables business users to build applications without coding, leveraging integrations with enterprise systems like Salesforce. Internally, Superblocks uses AI agents to automate tasks such as lead identification, support tracking, and sales engineering assignments, reducing the need to buy external tools.

Menezes’ approach highlights a broader trend: the future of AI-driven innovation lies not just in the models themselves, but in the artful engineering of prompts and the surrounding infrastructure that turns raw AI capabilities into tailored, reliable enterprise solutions.