A BIM-style tower is generated as a stack of floor slabs. A generative-design algorithm searches the trade-off between a compact, cheap cost-optimized envelope and a tapered, terraced sustainability-optimized envelope (daylight, greenery, solar shading) — you steer the search with the slider instead of running thousands of AI iterations yourself. Separately, IoT sensors embedded in the structure accumulate wear; a predictive-maintenance model estimates failure risk from that wear and, when AI is on, schedules a repair before the risk crosses the failure threshold.
footprint(f) = lerp(costShape(f), sustainShape(f), w)
fitness = 60 + 40 · sin(w · π) (w = sustainability weight, 0..1)
cost($) = floors · 85,000 · (1 + 0.25 · (1 − w))
risk(t) = risk(t−dt) + λ · dt (per-sensor wear hazard)
AI repair when risk(t) ≥ 0.6 → risk → ~0.03 (proactive fix, before failure)
no AI: risk(t) → 1 → sensor fails (stays red until Reset)
- Cost ⟷ Sustainability — the weight
w the generative-design search optimizes for; it reshapes every floor's footprint and re-tints the façade live.
- Floors — building height fed into the same generative search; the camera reframes automatically as the tower grows or shrinks.
- Maintenance AI — with predictive maintenance on, sensors are proactively repaired the moment risk crosses 60%; switch it off to watch untreated wear run to failure.
- Fitness peaks near a balanced
w ≈ 0.5 — the AI's own generative search found that a mixed envelope scores best, not either extreme.