If you use the target positions directly, all points go straight to that place making these straight lines.

So I used a method that takes the alignment between the sphere normals and the point positions, the more aligned, the more it displaces.

Also, the displacements are not towards the points, but just out of the sphere (this requires the power and the float curve to make a good transition)

Nodes

• Distribute the points
• For every vertex of the sphere, capture the index of the neares point
  • We will sample the position of this point
  • We will also use a random scaling based on this index
• As for the offset:
  • We sample the normal of the nearest point
  • Use the dot product between the sphere normal and the point normal to get how much alignment there is
  • Do the math (power makes it more spiky, color ramp leaves part of the sphere untouched, float curve tunes the shape)
  • Use this offset to displace the points along their normals
  • And a scale based on the id of the nearest point (so some spikes can get bigger than others)
  • Blurring hels eliminate hard transitions when two spikes meet