PurePerformance AI‑Native: Building Faster Than We Can Spec with Wolfgang Heider & Benedict Evert
Mar 16, 2026
Benedict Evert, an AI engineer who builds prototypes and tools with large models, and Wolfgang Heider, a product manager crafting AI-native prototypes and agent workflows, discuss building working apps in minutes. They talk about agent-driven roles, preserving engineering fundamentals like testing and governance, and why delivered value, not code generation, will set products apart.
AI Snips
Chapters
Transcript
Episode notes
TimeKiller Shows Prompts Can Produce Real Apps
- Wolfgang built and published a fully working iOS puzzle app called TimeKiller using AI, not just a mockup.
- He wrote prompts for minutes and ended with a Marketplace product his kids play, proving prototypes can become real products quickly.
Keep Classic Engineering Discipline With AI
- Do keep traditional software engineering practices: write story docs, epics, architecture, and run post-implementation checks.
- Benedict writes detailed specs so agents reference constraints, then reviews code, runs sanity checks, and aligns deployment with architecture.
Prompting Turns Engineers Into Agent Managers
- AI makes you a people manager of agents: you must learn prompt engineering and delegation to get consistent results.
- Wolfgang says artifacts (requirements, designs, runbooks) still matter to make outputs reproducible and fine-tunable.