Sure, I'd agree with that. Obviously companies in different verticals are going to have different benchmarks for what constitutes innovation. But designing "platforms" (to my mind), while certainly a product design undertaking, falls towards the MBA-end of the business maintenance vs. innovation spectrum.

There's a significant amount of deeply technical requirements in the back-ends of real-time platforms (which Valve is going to need to stop Google from stealing their pie) - especially if Google's super-real-time predictive magic is a believable seemless experience.

That will require a great deal of innovation - just not the sort of end user facing magic that the Valve of old produced.

Unfortunately I fully expect the external visage of the new Valve to be a smooth and greasy risk-adverse PR experience - but hopefully they can provide a platform that someone else can use to produce Half-Life caliber games in the future.

I guess I'm not sure exactly which product you're mainly referring to. The steambox/link product appears to have been a flop. If you're referring to some type of on-demand gaming product that they're developing I'm unaware of that, but in general I'm very bearish on cloud gaming. It very much feels like a solution looking for a problem. Most people interested in console/pc-style games will have a hardware budget anyway, and the whole idea seems like a revenue/subscription money grab. Who knows, maybe I'm wrong and that's the way gaming is going in the future, but I know for a fact that latency issues are a big stumbling block and it just doesn't seem like a reasonable use of bandwidth compared to, say, streaming video. That Google "pre-emptive" rendering thing sounded like pure marketing double-speak to me.

I'm not saying they have it, I'm guessing they'll need to. There's already some pretty complicated networking code in steam as it is - if steam adds cloud gaming to the mix, the infrastructure complexity is going to jump significantly.

If cloud gaming becomes something more than a dream, Valve is going to suffer hugely if the don't have a viable competitive offering.

Anyway, ignore me - I'm not sleeping too well today but now, and my brain's a bit melty.

Regarding the preemptive rendering, I think it is potentially viable.

What I'm thinking is branch prediction without the ability to roll back. You'll need to be able to minimise user noticable mistakes, as well as making any corrections minimally distracting.

I'm imagining auto-aim on steroids.

I think this is exactly right. They are always looking to the next thing, which leaves a false impression of “not doing anything” because they’re visibly not continuing to do the same thing.