{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreiem3csmrfyuqrgolm2zmk27bplguswbkpqh7fotb46dlpwdkaokgq",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mk2lb5svyqo2"
  },
  "path": "/t/from-prompting-to-system-design-a-10-stage-model-of-llm-users/1379453#post_2",
  "publishedAt": "2026-04-22T02:11:33.000Z",
  "site": "https://community.openai.com",
  "textContent": "Phase 1: Message Control Stage\n\nThis stage focuses on refining the content and format of a question to obtain better answers.\n\n1 → 2 (Reactive → Request-Oriented)\n\nRequired capability: Awareness of output format\n\nTransition condition:\nMoving beyond simply asking questions and beginning to specify the desired output format, such as tables, summaries, or comparisons.\n\n2 → 3 (Request-Oriented → Constraint-Oriented)\n\nRequired capability: Parameter control and constraint design\n\nTransition condition:\nDefining length, level, restrictions, and scope to constrain the response.\n\nNote:\nEarly stages (1–3) often overlap in practice.\nA single prompt may include elements from multiple stages simultaneously.\n\n3 → 4 (Constraint-Oriented → Structured)\n\nRequired capability: Question decomposition (Modular Input)\n\nTransition condition:\nBreaking the question into structured components such as [Goal / Scope / Constraints / Output].\n\nPhase 2: Reasoning Management Stage\n\nThis stage treats the model as a reasoning engine, managing errors and injecting structured thinking patterns.\n\n4 → 5 (Structured → Verification-Oriented)\n\nRequired capability: Critical validation and hallucination control\n\nTransition condition:\nNo longer fully trusting the model and requiring separation of facts, assumptions, and uncertainty.\n\n5 → 6 (Verification-Oriented → Framing)\n\nRequired capability: Application of analytical frameworks\n\nTransition condition:\nInstead of asking isolated questions, applying a structured framework to guide how the problem is analyzed.\n\nExamples:\n\n“Analyze this in terms of cause / effect / impact”\n“Organize by advantages / risks / trade-offs”\n“Break this down into structure / flow / importance”\n6 → 7 (Framing → Loop Design)\n\nRequired capability: Iterative process and feedback design\n\nTransition condition:\nMoving beyond one-off prompts to building repeatable workflows:\nInput → Process → Validation → Feedback → Iteration\n\nPhase 3: System and Governance Stage\n\nThis stage moves beyond individual usage toward system design and operational principles.\n\n7 → 8 (Loop Design → Meta Design)\n\nRequired capability: System architecture thinking\n\nTransition condition:\nMoving beyond writing prompts to designing external layers that coordinate the model’s reasoning process.\n\n8 → 9 (Meta Design → Engine Orchestration)\n\nRequired capability: Quantitative system operation and optimization\n\nTransition condition:\nTreating the model not as a conversational partner,\nbut as a reasoning engine that generates, compares, and selects among multiple outputs.\n\nOperating a structured system:\nCandidate generation → Evaluation → Optimal selection\n\nNote:\nThis transition occurs when the goal shifts from producing better answers\nto designing systems that consistently produce good outcomes.\n\n9 → 10 (Orchestration → Interaction and Collaboration Design)\n\nRequired capability: Interaction design and collaborative system definition\n\nTransition condition:\nMoving beyond obtaining better answers to designing how humans and AI collaborate,\nincluding role distribution and continuously improving workflows for problem-solving.",
  "title": "From Prompting to System Design: A 10-Stage Model of LLM Users"
}