What's next gen is something we've spent a lot of time thinking about.
A few hints about those discussions, is that the general feeling that one app that attempts to do everything for everyone and save all in one proprietary scene isn't really what people need anymore. The data you're dealing with may not even fit in RAM or be local. Also, there is a great canyon between game tech and the tech that the DCCs are built on. Procedural is nice, but having everything procedural as it is today is not necessarily futuristic or optimal, for example if it's all designed to run on the CPU - or saved when you don't even need to edit that construction history. Thing is, a lot of these workflows were built to run on the CPU, often on limited SGI hardware, and overly generalized to the point that the software is doing things in the background that people don't care about and don't scale with complexity. Anyways, those were a few of those thoughts, there were many more.

Little information ... Shows where AD apps fit and interact with each other in a modern pipeline scenario.

http://usa.autodesk.com/adsk/servlet/index?id=16116742&siteID=123112

For me as a freelancer a next gen ap would look like this. I know this would not likely be the best for large studios where everyone is specialized but here goes:

The interface is just skins that you can change easily. Buttons that you can just drag where you want them etc. It could be made to act and look just like Maya, Max, SI or whatever with 1 click.

Complete GPU acceleration throughout wherever practical, especially simulation and volume rendering.

Basic categories just like SI Model, Animate etc. but with a few more:

Modeling expanded: (three or more tabs)

Basic Polys/SubD's (much like current).
Solids and Nurbs with excellent import of CAD data like Solidworks etc.
Achitectural (great archvis tools like LWCAD).
Mudbox completely integrated. No sending to, getting from. Just do. Ptex supported throughout.

All modules highly inter-operable.

Complete Nature Tools: Something like Vue but with a real interface.

I'd like to be able to plant a few trees and build a building all in the same package.

Simulation: Real smoke and fire like FumeFX but GPU accelerated. I wan't to blow stuff up easily and have it look great. So do a lot of people...

Rendering: As easy a quick as possible for 3rd party companies to integrate their rendering engine. Then, you pick when you order, which one you want.

And lastly, a team of consistency police who do nothing but go through all functions throughout the entire package and make sure they all work, look and act in a constant fashion to minimize learning time and frustration.

If all three teams at AD were put together so they're not all re-inventing the wheel I think a lot could be done.

I know I don't want much

Would be nice though...

And so many redundant positions would cause how many people to lose their jobs? Not sure I'd want that.

that's the nature of a competive market (companies are forced to be streamlined an efficient). I guarantee there are other employers from programmers out there.

@Daniel Brassard - that's a nice piece of infographic, although it seems to perfectly illustrate the point of (part of) this discussion: from a pipeline POV Max, Maya & Softimage seem to be almost "perfectly" interchangable, even Autodesk doesn't seem to attempt to highlight their individual qualities in a pipeline demonstration anymore. And if that's so, some might argue, Autodesk does have a slight redundancy issue, while others might point out that this interchangability isn't reflected in the way the product Softimage is being marketed, considering the focus on ICE & FR, while the infographic seems to clearly show, Softimage might fill almost any spot in the pipeline otherwise reserved for Max and Maya.
And we're back to "square one"...

I'm just happy Softimage does appear on that chart^^

My point exactly!

Nice graph. Indeed it is nice to see SI in there. Ironically, they seem to have forgotten their own "Flame" package in the compositing section. It used to be a big deal but maybe not anymore. Combustion and Inferno seem to be gone completely.

I think Softimage weak point is that it doesn't have enough modifier like 3DS Max has. There are only a few of them. Things like a surface modifier to convert curves or splines into a solid mesh.
Perhaps some better retopology tools. Gizmo based modifiers like in 3DS Max. As mentoined before something like Fumefx. A muscle system -> although you could apply softbody dynamics to clusters. A nice scattering tool like object paint in -> you know the other application.

More modifiers that's what we need.