GPT-6 Astra作者 Juy

洛伦兹吸引子演示:相差百万分之一的两条轨迹逐渐分离

Astra 编写了洛伦兹方程模拟:起点只差百万分之一的两条轨迹最终走向不同回路,带实时间距读数和时间拖动。

在线演示

Claude Code + Three.js(仅代码)Wandy 2
真实模型 · 360°两者都基于同一张照片:左边是 Claude Code(Opus 5.5)用 Three.js 代码建模的一次未经修改的运行结果;右边是 Wandy 2。

创建精细的 3D 模型

单靠代码只能做出简单的形状。Wandy 2 能用一张照片生成带纹理的 3D 模型,直接用在同一个场景里。

Wandy3D 制作说明

作者没有公开提示词,因此我们写下了复现这个作品的方法。

Build an interactive browser visualization of the Lorenz system that shows sensitive dependence on initial conditions. Simulate two trajectories with identical equations, one starting at (1, 1, 1) and one with the x value nudged by one millionth, integrated with a fourth-order Runge-Kutta method and a small time step. Draw the x-z projection on a dark navy background as a butterfly-shaped attractor: faint full paths and bright recent segments, one trail in cyan and the other in coral, with glowing head dots. Add a headline reading that the starts are one millionth apart, labels for both starting points, a simulation-time readout and a large live counter of the three-dimensional separation. Let the viewer change the starting difference and scrub through time. Accelerate time and include a short caption saying these are numerical trajectories, not a forecast. Done means the two trails overlap at first, then visibly split, with the counter growing from tiny values to double digits.

如何使用这条提示词

  1. 把构建说明贴给 GPT-6 Astra,要求生成使用 RK4 积分的单个浏览器页面。

  2. 修改初始偏差、颜色、时间速度或投影平面,观察不同表现。

  3. 拖动时间轴查看间距读数,再让它缩小步长来验证数值。

原帖你是作者吗? 请联系我们修改或删除此页面