Engineer: AI Workflows Rediscover SE Patterns

- The 'discuss-plan-execute-verify' workflow mirrors the "plan-and-execute" agent architecture in AI, which separates reasoning from action. These agents first create a structured plan, then execute each step using available tools, and verify the outcome before proceeding, a departure from simpler "ReAct" models that reason and act at every step. - The "verify" step in systems engineering is a formal process that answers the question, "Did you build the system correctly?" This is accomplished through a combination of analysis, inspection, demonstration, and testing to provide objective evidence that all requirements have been fulfilled. - For safety-critical AI, verification is evolving to include formal methods, which are mathematically-grounded techniques used to prove a system's correctness and security. This provides a much higher degree of assurance than empirical testing alone, a standard practice in fields like cryptography that is now being applied to AI safety. - A primary formal methodology in aerospace is Model-Based Systems Engineering (MBSE), which uses digital models as a central source of truth for a system's requirements, design, and analysis, replacing traditional document-based approaches. The International Council on Systems Engineering (INCOSE) champions MBSE as critical for managing modern system complexity. - The integration of these fields works in two directions: "Systems Engineering for AI" uses SE principles to manage the development of complex AI, while "AI for Systems Engineering" uses AI to improve the engineering process itself. For example, AI tools can accelerate development by automatically generating or checking parts of MBSE models. - The U.S. Department of Defense has made MBSE a key part of its digital engineering