I migrated from Thunderbird to GMail, and from Word/Excel to Google apps.
The only desktop apps I'm still using, besides development tools and the actual browser, is a PDF reader + Skype + iTunes + the file manager + Gimp / Photoshop for editing my personal pictures (before uploading them to Flickr).
On my Android phone I use native apps for online services, mostly because bandwidth is a problem and native apps have better caching. But the web interfaces for Facebook, Twitter and Flickr are actually usable.
I'm not saying that everything should or will move to the web. It would actually be stupid implementing Photoshop in the browser.
But the web is clearly dominating my time spent with computers.
"It would actually be stupid implementing Photoshop in the browser."
You say that now... In truth I don't think there's a fundamental reason why that's such a bad idea, and I think it will happen in the coming years. JS compilation is getting better and better, eventually it'll reach the theoretical limits, which is to run more or less as fast as client-side C. Similarly, web standards are improving at an amazing rate, given the development base and momentum behind web applications (which are growing to be as big a business as any platform) eventually we're likely to end up with all the bits and pieces for supporting web apps (UI elements, frameworks, etc.) at a level matching that of desktop development.
At some point it's not only possible but indeed very likely that a photoshop web app will be built. But by then it probably won't seem like a crazy idea.
I'm not talking about image editing in general, I'm talking about Photoshop -- the industry standard for professionals, software that should be providing everything needed for professional-level retouches.
This is an important difference -- Moore's law is still applied -- but image resolutions are getting bigger and bigger, algorithms get more complicated and more fancy, people also want to be more and more productive.
What I'm saying is that image retouching for professionals needs every CPU cycle you can spare + screen real estate + really good integration with input devices.
I'm not seeing the browser (which stagnated for years), as a powerful enough medium for that, unless you can make Javascript as-fast-as-C without sacrificing too much RAM (since you need that too) and unless you provide lower-level access to hardware (OpenGL ES is a good start, but you need more).
Now surely, if Photoshop can't be built in the browser, that doesn't mean something less powerful will be less useful. But it won't be Photoshop.
I know what you mean, I mean the same thing. In 20 years time I fully expect that a Adobe Photoshop (or something 100% equivalent) will be a web app. Sure, it can't be done reasonably today, but for every single roadblock preventing it there will be many competent engineers capable of removing that one roadblock and many people (developers and users) providing a strong incentive to do so.
Adobe is already trying to create a web version of Photoshop. Eventually, they'll get to full compatability but I don't think the web adds a unique advantage to Photoshop; likely they're just doing it to cut down on piracy figures.
One place where the web apps can excel is in replacing 3D modelers. Once we get to the point that the frontend can be done quickly in JS, then the entire backend (e.g. rendering) can be done on a server-side cloud.
Imagine being able to lease server time so that what was a 10hour job, is now a 1 hour job?
People spend 8 hours a day using desktop apps -- word, excel, outlook, GIS software, development tools, whatever.
They read their email and twitter and consume content throughout the day on their mobile devices.
At home they play video games, read books, comics, videos -- all using 'apps' on their TVs, mobile devices, and playstations.
Color me unconvinced -- I just took a break from 3 hours reading in iBooks to write this.