{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreih2varkajk3vvvpgr63cl3bx7hnwuxktsv2rftvsylty37hxxtbjq",
"uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mjm3e3ccldd2"
},
"path": "/t/prompting-vs-structure-a-boundary-test/1379094#post_1",
"publishedAt": "2026-04-16T08:55:09.000Z",
"site": "https://community.openai.com",
"tags": [
"(click for more details)"
],
"textContent": "I ran a simple test with an image model to see how far prompting can go when we ask for physical realism instead of aesthetics.\n\n**Setup (4 steps):**\n\n 1. Start with a standard prompt → result: visually pleasing image.\n\nAesthetic prompt (baseline) (click for more details)\n 2. Add physical requirements (airflow, light behavior) → result: still smooth, slightly improved.\n\nPhysical prompt (first correction) (click for more details)\n 3. Increase structural demands (irregularity, no symmetry, partial rainbow) → result: looks more complex, but not more accurate.\n\nStructural prompt (a deeper, systemic approach) (click for more details)\n 4. Add conflicting real-world constraints (turbulence over time, particle behavior, observer-dependent optics) → result: visual chaos, not physical consistency.\n\nStress test prompt (out-of-bounds) (click for more details)\n\n\n\n**Observation:**\nThe model consistently produces images that _look_ realistic, but it does not seem to consistently maintain physical relationships under more complex constraints.\n\n * Turbulence appears as texture, not as structured flow\n\n * The bird resembles a red kite, but lacks precise morphology\n\n * The rainbow is present, but not truly dependent on viewing geometry\n\n\n\n\n**Conclusion:**\nPrompting can steer appearance, but it does not create underlying structure.\n\nNo matter how precise the prompt becomes, the system prioritizes what looks coherent over what is physically consistent.\n\n* * *\n\nIn short:\n\n> It can generate convincing images,\n> but not fully consistent systems.\n\nThis is not a flaw of prompting — it’s a limitation of the model itself.\n\nCurious how others see this:\nAt what point do prompts stop improving results and start exposing system limits?",
"title": "Prompting vs Structure - A Boundary Test"
}