Codegen AI is great for tab-complete. It does save typing, and it is pretty smart, but you still have to have an engineer in the seat that is paying attention and knows what they're doing.
Video AI on the other hand is already disrupting Hollywood. The costs are 10,000x cheaper, the delivery time is 100,000x faster, and it's accessible with 1,000x less personnel. These are insane numbers. Multiple orders of magnitude - full step function change - across multiple dimensions at once.
You still need to have an editor and VFX person to make video AI work, and you still need to incorporate photon-on-glass footage, but it's remarkable how usable these tools are.
A small studio with five people can use AI tools to do serious work that would ordinarily take 100 people or more and require huge budgets.
This isn't just hypothesizing - I work in this industry. This is not only myself, but this is dozens of the studios I talk to and work with.
Just to cite one example, there's a studio I know that used to bid $300k on projects. You've seen their work - they do Netflix show intro sequences, pharmaceutical ads, etc. They're now bidding just $50k and winning projects left and right.
I think the biggest area of AI investment and disruption is going to be video AI.