The Shift from Ideation to Execution

Arthur Sroken, co-founder of Google's AI Futures Fund, interviewed founders from two portfolio companies, Emergent and Whering, to discuss the realities of building AI-powered products. A recurring theme is that AI has fundamentally changed the role of the "idea person." Previously dismissed for lacking technical execution, founders can now use AI agents to bridge the gap between concept and a production-ready application.

Engineering Rigor and Evaluation

Both companies emphasize that success in the current AI landscape requires more than just calling an API.

  • Emergent's Approach: The team built their own sandbox container technology rather than relying on third-party sandboxes. This allows their coding agents to receive real-time feedback on system failures, enabling rapid, autonomous iteration. They prioritize "experiment velocity," using automated evals to test how models perform against business metrics like retention and conversion.
  • Whering's Approach: Whering balances AI-forward features (like conversational chat) with "behind-the-scenes" AI (like computer vision for wardrobe digitization). They stress the importance of maintaining a deep backlog of user research to prioritize features that align with user needs rather than just chasing the latest model release.

Strategic Model Selection and Efficiency

Both founders argue against a simplistic focus on "price per token." Instead, they advocate for a holistic view of efficiency. A more intelligent model might cost more per token, but if it completes the task with fewer errors or fewer steps, the overall cost and quality improve.

  • Orchestration: Emergent uses a sophisticated orchestration layer that routes tasks to the most appropriate model (e.g., Gemini for UI/frontend tasks, Flash for speed-sensitive interactions).
  • Human-in-the-loop: Both teams emphasize that manual evaluation remains a critical gatekeeper. Before automating, they recommend having humans review model outputs to develop a "vibe check" for quality.

The Future of AI-Enabled Apps

Both founders see a future where AI is not just a feature but the core of the user experience. For Whering, this means moving toward voice-driven styling and reducing the friction of onboarding. For Emergent, it means enabling non-technical domain experts to build production-grade software, effectively democratizing the ability to ship complex applications.