You often talk about releasing early, building something people want, etc. Since you carry a lot of influence, and (I guess) invest in Etherpad, and they've adjusted it to your personal request - do you worry that this effect might give you a distorted view of how well Etherpad specifically, and YC startups generally, are building what people want?

To further drive that thought home, my company hereby offers to add any 1-day-effort feature request made by any popular tech figure, provided they're willing to share our work with their audience. (Sites for those would be http://bug.gd or http://www.yumbunny.com or any other site in my profile.)

Seriously, they'd be silly not to have taken pg up on this even if he wasn't helping to back them.

I don't think pg implied that he bullied them into implementing this, only that he talked them into it. And I can't imagine that YC would last long if pg got into the habit of bullying its start-ups into implementing his personal requests.

While solutions like DropBox or Sharepoint allow versioning and collaboration it's at a higher level. Maybe I've been spoiled by VMWare, but I love having branches of specific versions to fall back on and test things out. Why should writing be any different?

What a great way of bringing attention to a start-up: make something that's inherently fascinating due to features. This is unlike anything online I've seen before because the functionality just doesn't exist anywhere else. So it's a great thing to watch and it also tells you about the service.

I like that you keep constantly going back to eliminate unnecessary language. Something tells me that's the first and most obvious trait of watching good writers write: at the governor's school, when the writers would write on paper instead of on the computer, you could read a draft and see that emerging pattern of phrase after phrase, each building on the last. It's interesting seeing it happen on the computer.

Anything you write with it now you will be able to play back like this soon. Etherpad is already saving all the keystrokes. They just haven't released the player yet. (This is a one-off prototype.)

http://gvetchedintime.com/gvetchedintime/videos.php

This site shows how some fairly good illustrations are drawn (brushstroke playback). Unfortunately, not all drawings can be animated and the ability to search for playback-capable drawings is limited:

http://www.ratemydrawings.com/

http://www.ratemydrawings.com/drawings/animation/428912.html

It would be interesting to have the timestamp data (if it is recorded) of the keystrokes (adjusted to author's timezone - privacy withstanding) so you can see a clock/calendar [day|night icon] when the author wrote it (was it 3am in the morning? did they put it aside for a couple hours or was it a period of days?).

EDIT: Didn't see this comment. http://news.ycombinator.com/item?id=495812

However, it should be somewhere hidden and available only if people really want to be able to replay. Else it is going to be an over kill on their servers.

It might be kind of jerky during the editing stage, but it's better than scrolling around trying to find out where the action is. ;)

What I learned from this is that I write very differently from the way you do. My first draft is almost straight stream of consciousness, and then I go back and reread and edit several times. You seem to make several attempts at each sentence before finding one you like, and then edit the whole thing.

Of course, I may just not realize how I actually write, since I don't have a playback of anything of length to watch.

What I'd really be interested in watching is some famous programmers working on code with this kind of playback. Not sure if it'd actually be enlightening, but it would be fun to experiment with.

With a bit of Javascript I'm sure you could make it scroll smoothly.

"Launching lets you observe the collusion between your presuppositions and the needs of actual users."

I went back, and you are right, it is collision.

Collusion is more fun though. :)

If this were only available when you guys were writing code for Viaweb, and when you were writing On Lisp -- what a wonderful thing it would be to learn from. It would show the thought process behind what went into writing what you wrote!

They will probably give up Word for something, eventually. Especially among the current generation, writing is something done online; a blog, an email, a tweet, writing on someone's wall. At some point, it stops making sense to have a word processor distinct from the web. Right now, Word has more features that people like than online writing tools, but that will change over time. Online writing tools are a more natural fit for collaboration and integration with words' final destinations.

Microsoft could make the online tool that replaces Word, but it is probably too dangerous to their current revenue stream. I don't think they can get the same revenue for an online version of Word, and it would be difficult for them to rationalize trying a new business model that may never regain the revenue they get now.

(And yes, I'm basically just making the Innovator's Dilemma argument for word processors.)

Something like Windows Live Writer 2009?

http://jalaj.net/2009/01/22/enter-windows-live-writer-2009/

If it has an installer, it doesn't count as an online tool, as far as I'm concerned.

If high-resolution timestamps for each keystroke aren't too much of a burden it seems like this would be straightforward to implement, given an algorithm like BioPassword's.

There are already non-webapp mouselog/keylog and replay applications. And besides, think of the implications of letting students be on the internet while they are writing essays.

a) telling students to install a keylogger, write an essay with it turned on, then submit the essay plus the logs, and then the teacher figuring out how to replay the logs, versus

b) telling students to write their essay at etherpad.com, they submit the url, teacher clicks on "replay."

As for the matter of students being online, that's going to have to get solved anyway, because all the apps are moving online.

I think that pendulum has almost reached the end of it's arc. There are a number of trends that are starting to pull in the opposite direction. From concerns about privacy and security to the proliferation of mobile computing, intermittent disconnection from the cloud has it's uses.

Tools like Google Gears and the continued blurring of the line between desktop applications and web applications mean that it's easy to build applications where being connected enhances the experience considerably, but the essential functions are still there even if the network goes away.

Ironically I think it's the tech crowd/geeks (us) who are most resistant to webapps. We can see the privacy,security,reliability issues. The average users however, cannot. They're already embracing webapps.

The big question is what's next. I'm guessing it'll be mobile - the geeks are embracing it just as the mainstream is finally becoming comfortable with last decade's technology. But it's a bit too early to tell.

First, there's the simple privacy issue. If a professor told me to install a keylogger on my computer or if she told me to use etherpad to write an essay - either way I'd feel like my privacy was being violated.

Then there's the practical issue of how this system would affect writing ability. I personally do my best writing when I'm able to ignore the critic in my head and focus on the words. Being self-conscious about someone seeing all the stupid things I wrote along the way would make it much more difficult to get into this state of concentration.

Finally, there's the issue of making teacher's not only read two dozen boring essays but also watch replays of them being written. I personally think their time could be better spent.

...and even then I use WriteRoom, full screen green on black, no formatting options: http://www.hogbaysoftware.com/products/writeroom

  script -t 2> script.demo.time script.demo
  xemacs -nw
  # M-x tetris
  #            ; play tetris now
  # C-x C-c    ; exit emacs
  exit
  # watch tetris being played:
  scriptreplay script.demo.time script.demo