With the advent of Claude Flow (an orchestration service for Claude Code instances) it becomes feasible to let an entire swarm of AI agents deal with complex software engineering problems, provided that the “human-in-the-loop” takes particular care to analyze all preconditions, desired features and constraints thoroughly, and to make that analysis available to the AI. This may happen in the form of a formal requirements analysis, as would otherwise only be common in regulated markets, such as medical devices. A test-driven development goes a long way towards ensuring a sufficient quality, as does the V&V process once the code is finished. Unlike in today’s cartoon, humans should stay in the driver seat and should not trust the AI blindly. If you are unsure about what is possible today and how to deal with the substantial changes ahead, do contact us. W can assist with both technical and economic questions.
Have fun, Rüdiger
