Building AI-Native Apps and Teams with Founder William Imoh
How a non-technical founder leveraged AI agents to build mobile apps and manage operations.
1Motivation for Mural
William Imoh, founder of Hackmamba, created the AI language tutor Mural after a 1,000-day streak on Duolingo failed to provide conversational fluency in Norwegian. He designed the platform to prioritize complete immersion, forcing users to interact entirely in the target language rather than relying on English translations or repetitive gamified exercises.
2Coding Without Experience
Despite having no background in Swift or Kotlin, Imoh successfully built native iOS and Android applications using Codex. He discovered that when AI agents handle the actual coding, the founder's specific programming knowledge becomes secondary to having a clear product vision and the ability to articulate technical requirements through precise user stories.
3Design and Prompting
Imoh emphasizes that product management experience is critical for effective prompting. By structuring requests as epics and user stories, he provides agents with clear goals. He also applies design judgment to AI outputs, iteratively refining visual styles to ensure the final application meets high-quality aesthetic standards rather than accepting default dark modes.
4Real-time Voice Economics
The platform relies on GPT-Live-1 to facilitate real-time voice interactions, which Imoh identifies as essential for immersion. While a 30-minute session costs between $3 and $5, he considers this a viable expense. It remains significantly cheaper than hiring an on-demand human tutor while providing the necessary consistency for effective language learning.
5Lord of the Rings Agents
Imoh manages his 25-person company using a team of AI agents assigned specific roles based on Lord of the Rings characters. Gandalf serves as chief of staff, while Legolas handles marketing and Aragorn manages sales. These distinct personas simplify the process of delegating tasks and clarifying responsibilities across the entire organization.
6Context Through Data Access
To make agents truly effective, Imoh grants them broad access to internal tools like Google Workspace, Linear, Notion, and Pipedrive. He notes that providing access to relevant data is more impactful than writing longer prompts. He maintains security by keeping sensitive banking and customer data protected while allowing agents to process information.
7Custom Systems of Record
Imoh developed Three Pipes, an internal HR tool that integrates email, calendar, and meeting notes to track candidate rankings. Instead of forcing his workflow into generic software, he built a custom database that matches his company's actual needs. This approach highlights the potential for AI to support bespoke internal operational systems.
8Future of AI Expertise
As AI models improve, the need for broad technical skill libraries diminishes. Imoh suggests that durable expertise, such as growth strategy and content planning, remains vital for founders. When agents stall, he recommends restarting sessions with clearer instructions and implementing a secondary review process to ensure the quality of the final output.
Original article: Lenny's Newsletter — Building AI-Native Apps and Teams with Founder William Imoh