Sachin Rekhi on Building Functional AI Prototypes, Not ‘Slop’

In a recent LinkedIn post, Sachin Rekhi addresses what he terms the proliferation of “slop” in AI prototyping among product managers. He argues that the current approach often leads to superficial prototypes that are far from being shippable products, and outlines a path toward building more robust, functional prototypes that can genuinely inform product development.

Rekhi emphasizes a shift in methodology, suggesting that companies should focus on building functional prototypes for various problems and then iterating based on actual user interaction, a strategy he notes was once only feasible for giants like Apple but is now democratized by AI prototyping tools.

“AI prototyping makes it available to everyone.”

He identifies the core issue as the gap between generic, unshippingable AI outputs and production-grade prototypes. Rekhi proposes a structured approach, referencing a “15-skill ladder” that helps bridge this divide.

The Challenge of AI ‘Slop’

Rekhi elaborates on the nature of this “AI slop,” describing it as prototypes with generic features and basic scenarios that, while seemingly magical, would never be considered for release. He points out that simply asking an AI to “create a CRM,” for instance, yields a product that lacks the depth and polish of a real-world application.

“When you type ‘create a CRM’ you get generic styling, vanilla features, basic scenarios. Looks magical but you’d never ship it.”

To combat this, Rekhi champions the importance of design consistency. He suggests establishing a baseline design template by recreating and iterating on a product’s screenshot until it’s perfect. This foundational step, he argues, ensures that all subsequent prototypes automatically inherit the established design system, saving significant time and maintaining coherence.

Strategies for Production-Grade Prototypes

Rekhi further outlines several key strategies for moving beyond basic AI outputs to create truly functional prototypes. One critical technique he highlights is encouraging divergence in the design process.

Embracing Divergence and Functional Testing

Instead of generating a single design, Rekhi advocates for creating multiple variants. Tools like Magic Patterns are mentioned as facilitators of this process, allowing for the generation of four design variants, or even more when using multiple tools. This approach, he contends, provides broader brainstorming and exploration capabilities than traditional design methods.

Crucially, Rekhi stresses that the validation process must evolve. He advises integrating real-world elements into prototypes, such as the OpenAI API, analytics tools like PostHog, surveys, session recordings, and heatmaps. This allows for testing actual functionality rather than relying on static mockups.

“Now you’re testing real functionality, not mockups. Users interact with actual features.”

Tooling for Product Managers

To support these advanced prototyping efforts, Rekhi provides a breakdown of tools tailored for product managers. He categorizes them based on specific needs:

• Bolt: For speed.
• Magic Patterns: For product teams focused on divergence.
• Reforge Build: For context integration.
• Cursor: For technically inclined PMs seeking power.
• V0: For creating visually appealing UIs.

Rekhi concludes by underscoring that the necessary skills and tools are available. The ultimate determinant of success, in his view, lies in the willingness of product managers to learn and apply these methodologies.