The ChatGPT image generation upgrade hints at a different headwind for software development regarding AI: less of a need for bespoke software.
Presently, LLMs can't write all of the code for Blender or Photoshop, but ChatGPT can produce higher fidelity photorealistic and 3D images than, probably, everyone but a few thousand people in the world. So, is it more impactful that it can't write the software it tries to replace, or that it rapidly replaces the need for that software?