Jorge Coca talks with Peter Trost, a senior Flutter engineer at 1KOMMA5°, one of Germany’s fastest-growing clean-energy companies, where he works on software for home energy systems. Peter came to app development from an unusual direction, studying computer science with a focus on AI and machine learning, and that background shapes how he thinks about what today’s tools can and can’t do. He explains why the slow feedback of machine learning research pushed him toward user-facing work, and why Flutter’s hot reload, widget inspector, and the Dart language made it stick.
The heart of the conversation is the property management startup Peter is building largely on his own with AI. He walks through features like voice-first inspection reports that transcribe a manager’s spoken observations into structured work orders, and guided digital checklists that keep inspections consistent. He treats AI as a collaborator rather than an autocomplete, using agentic tools like Claude and LeanCode’s Marionette, an MCP server that lets an agent drive a running Flutter app to verify its own work.
Peter and Jorge unpack what minimal code really means: far less hand-written code, held to the same production bar through clear structure, tests, and runtime checks. Peter also reflects on introducing AI-assisted workflows inside a larger engineering organization at 1KOMMA5°, and the constant pull between focus and exploration that comes with running a startup on the side.
