Brad Menezes, the CEO of Superblocks, recently highlighted an untapped potential in the world of artificial intelligence (AI) startups, specifically within system prompts. These prompts read like extensive rules for artificial intelligence generative models—foundational models already in widespread use from companies such as OpenAI and Anthropic. Or, they might be the less-apparent gems that inspire the next wave of promising billion-dollar startup concepts.
In a viral tweet that reached nearly 2 million viewers, Menezes outlined three critical parts of system prompts to study: role prompting, contextual prompting, and tool use. Yet his visions have captivated the imaginations of entrepreneurs and tech enthusiasts. He even goes so far as to say that knowing these key parts will help you discover exciting use cases in the AI world.
These system prompts, or instructions to the AI, are crafted in plain human language, but they are incredibly technical, sometimes well above 5,000 or 6,000 words. They guide AI models on the development of application-level products by walking through tasks and outlining expectations. According to Menezes, this level of specificity is necessary to truly direct the behavior of AI.
“You basically have to speak as if you would to a human co-worker,” Menezes explained. His approach is part of a broader trend among advanced AI startups to customize proprietary prompts to fit organizational operational requirements.
Menezes noted the extreme inconsistency in system prompts from company to company. “Every single company has a completely different system prompt for the same [foundational] model,” he noted. Its unique degree of customizability enables enterprises to tailor AI interactions for targeted domains and use cases.
In fact, Menezes has been busy “dogfooding” his own creation at Superblocks. Internally, he uses system prompts to generate customized solutions specifically for his team’s needs. To address these needs, his business associates through their entrepreneurial creativity supplied a wide variety of agents. One such agent uses Customer Relationship Management (CRM) data to find leads. This is just one example of how thoughtful use of system prompts can improve operational efficiency.
Just this week Superblocks raised $23 million in Series A funding, for a total funding haul of $60 million. This strategic investment is intended to enable the firm to confidently produce enterprise-grade applications in a world increasingly shaped by AI-derived application programming strategies. Menezes was clear to make the point that this strategy does not consist solely of procuring tools but rather developing them in-house. “This is basically a way for us to build the tools and not buy the tools,” he stated.
The use cases for system prompts seen at the best of breed companies show you just how big and creative this instruction can be. As an example, here’s how Devin’s system prompt starts off, writing to a real computer OS. You are a real code-wiz: few programmers are as talented as you at understanding codebases, writing functional and clean code, and iterating on your changes until they are correct. “This one prompt raises the bar on performance requirements and subject matter expertise the AI is supposed to achieve.
Similarly, Cursor’s system prompt instructs the AI to “only call tools when needed, and never mention tool names to the user — just describe what you’re doing.” This degree of accuracy is what gives users clear, understandable interactions, without the burdens of overengineering.
Replit’s full system prompt reveals remarkable functionality. It allows you to edit and search code, install new programming languages, create and query PostgreSQL databases, and run shell commands. Each piece is detailed down to the studs in order to squeeze every drop of potential out of the AI.
Menezes acknowledges that studying others’ system prompts has allowed him to discern what aspects resonate with developers and coders in the industry. “They’re trying to get the model to do exactly what’s required for a specific domain, specific tasks,” he explained, underscoring the importance of aligning AI capabilities with business needs.
Leave a Reply