Another possibility is that YC could maintain a higher percentage of success, at a lower individual return. Because YC attracts some of the best young talent around, is very selective in funding, and seriously optimizes their fundees chances of success, it's not unreasonable that YC could maintain a better batting average than the VCs.
Another possibility is that because YC tends to form definite relationships with the founders they fund, they can count on the law of large numbers within, not just between, teams. For instance, YC lost money on Kiko, but is also invested in justin.tv--if justin.tv makes it, the YC[kiko folk] relationship could still end in the black.
N.B. This is all complete speculation; any more insight from pg would be appreciated.
This is an interesting adaptation of the notion of glogging / moblogging; I have to wonder to what extent these guys were inspired by the much more theoretical work done with eyetaps (amazing technology, but not really scalable). The interesting thing to see will be whether, since this is meant to be outward-facing (as opposed glogging's introspective bent), it will suffer from the lack of editorial control. Basically, does a person's life have a high enough density of interesting content that it is worth watching? If the answer is yes, then there is definite potential for radical developments in social media.
This article made me smile. Perhaps more importantly, it made me reread WhyFP, and the pastiche WhyHaskell. I'm a member of the choir, of course, but it's often good to read over a classic text to glean fresh understanding, and perhaps most importantly to see how "timeless wisdom" applies to your current adventures. (It's for that reason that I'm rereading The Mythical Man-Month.)
O'Reilly clearly gets to a point that I tried to make in reference to one of the first round of articles about freebase: namely, that by adopting the ideas of web2.0, namely relying on users to generate information because it is natural and convenient for them, we have a chance to take another shot at the semantic web, and perhaps actually make it past arguing in the W3C this time.
A programmer for google or yahoo, sure. A programmer for Microsoft or Adobe, not so much. Moreover, outside the big names, software companies are largely unknown; to most of us, hearing that someone works for, e.g, Fog Creek, is pretty nifty, but to the average person, it's meaningless.
This is an interesting variation on the model that Fon is using. If I'm reading it correctly, Meraki is assuming that only fraction of the routers will be hardwired to the internet, with the remainder acting only to extend the mesh. In that sense, it's similar to a local, community based project (alacloud.net) that I'm working with now.
Unlike with Fon, however, there is no clear way to offer users any compensation[1], or really to monetize the network (other than something like ads). I'm intrigued to see how this "experiment" turns out.
[1] It's left as an exercise to the reader to decide whether Fon's model is a horrible idea or not.
It works, but only under the assumption that user liveness is constant or increasing over time. A liveness metric should differentiate strongly between a user who has commented 30 times in the past 4 days, and someone who commented 25 times in the first 4 days of news.yc. and 5 times since. Perhaps I'll hack something together.
I'd actually go one step further, and say that user interface is /more/ important than code. In almost all case with a start-up, you users don't really care about your code, they care about the functionality, and its usability. That means that HCI trumps code, to some extent.
Of course, interface matters only in so far as it exposes functionality; that's why craigslist, which is a monster of design, but highly usable, can do so well. The interface facilitates the user's interaction with the functionality.
Even if there isn't a cut-off for adding features to software, there's probably a cut-off for where adding features is the best occupation for the founder. Either the software fails, and its moot, or it takes off, and if it takes off, there is probably a point at which it makes more sense for the founder to leave and work on a new project--one where the hard problems still have to be solved.
Do we have any metrics on liveness of accounts? I've certainly felt that the community of active, thoughtful commentators is pretty small (not a bad thing!).
Looking at old comments[1], though, indicates that the overall volume of karma flow was much higher in the early days. Of course, it could be that the bug-driven hyperkarma on posts led users to be more generous with karma on comments.
I run into this problem all the time--not good for diminishing by reload-addiction! It would be nice if pages included some js to rewrite history[1] via dom, so that we could always see the freshest versions of pages.
[1] It pains me to advocate breaking the "show me exactly what I was just seeing" semantics of the back button, but I think in this case the user clearly conceives the back button as "show me the abstract resource I was just seeing."
More important, I think, than displaying the number of new comments is making it possible to /find/ them. The reordering of comments is usually a great thing, but in a relatively involved discussion, it can become quite a chore to find that new comment.
I'm not sure what the best way to implement it would be, from either an algorithmic or HCI standpoint, but it certainly would be nice if you could come up with a way to make new comments stand out in threads. (Preferably with new defined by when the user last viewed the page, rather than being a static global definition.)
While the legal principle backing their idea (first sale rights) is pretty solid, I doubt it will hold up here. The courts have not been kind to first sale and fair use rights in the past 5 years. This case is also complicated by the fact that there is no way to ensure that there is no concurrent use. While the system can ensure that no two people simultaneously use my copy of a song, it cannot ensure that I am not using my copy while someone else is.
University. In my opinion, college is the best thing to do when you don't have something better to do, if you'll pardon the tautology. Universities put you in contact with lots of other smart people, with the resources and freedom to devote your effort to whatever serves your interests.
Wrong, perhaps, but a necessary wrong. Smart people that you know are perhaps the most effective filter to gauge purportedly smart people you don't know. While it would be nice if connections didn't matter, such a world would be infeasible. Thus, you enter the social network on merit: impress someone who is known to be smart, and that's your in. Eventually the line of contact reaches the VC you need to take you seriously.
I tend to think that it is more the result of a bit of cognitive dissonance, particularly with respect to what the points values mean. They can be viewed either as a metric of popularity, or a metric of quality. Most people {{handwaving}} conceive of points as some mixture of the two, hence there is a reluctance to give a point to a story with, say, 20 points, even if you liked it, because you don't think it is a "21-point quality" story.
Interesting...could someone who's been around longer than I have take a stab at assigning some of those peaks to real-world events? I'm particularly interested in the peak near the end of February, and the following trough.
There was once upon a time a bit of rumbling about getting Jython going on javame. I don't think anything ever came of it, and if standard Jython is any indication, it would have been, as you said, /very/ slow.
Based on the fund size they are discussing, I'd say that's a pretty good bet. For the startup community, I think that's a good thing. While competition in the seed/angel space is fine and dandy, having a VC firm headed by someone has hacker-aware as Tim O'Reilly seems like a definite win.
You're right, of course (and I need not point out the irony that DH was in deep with AI), but I'm holding out hope (and not just because I'm bitter that the Connection Machine failed!). The main reason (in my opinion) that the semantic web has failed so far is the reliance on 1) predefined, very limited ontologies, and 2) creator-generated rich metadata. Freebase seems to me like, if not the solution, at least a good step in the right direction in terms of addressing these. The real question, of course, will be whether a userbase springs up to keep it up to date, and whether the logic behind it can keep it reasonably self-consistent.
I'm going to go ahead and cross-post this from the other article about freebase.
Disclaimer: I admire Daniel Hillis greatly, perhaps irrationally.
That said, Metaweb sounds amazing. If it takes off, it will be not so much a useful product in itself, but an enabling technology for an entirely new class of products. Imagine the possibilities that arise from having a relatively complete, constantly evolving ontology of everything. All of those nifty tricks that computational linguists and proof theoreticians can do over finite domains, using very limited ontologies, become possible for all sorts of other applications. Armed with an ontology that closely matches the real world, it is finally possible to write code that can reason about the real world, rather than just throw Bayes law at it and hope something useful falls out.
From an HCI standpoint this would be tricky, but it would be great to have a system for handling dupes and near-dupes by merging their comment pages. For instance, right now there are two articles about the recent demo of Metaweb/freebase. I commented on one of them, returned to the mainpage, and realized that the other (which also had no comments) was now higher ranked.
This raises a quandry as to whether I should cross-post, or just move on and hope that the discussion happens in the one I picked.
(Note, this is the harder case, of near-dupes: the two articles are different, but \begin{precog} most of the discussion will be about the product they reference.\end{precog} Hence it is semantically reasonable to merge their comment threads.)
So, as a oneliner: add a way to merge duplicate articles.
Disclaimer: I admire Daniel Hillis greatly, perhaps irrationally.
That said, Metaweb sounds amazing. If it takes off, it will be not so much a useful product in itself, but an enabling technology for an entirely new class of products. Imagine the possibilities that arise from having a relatively complete, constantly evolving ontology of everything. All of those nifty tricks that computational linguists and proof theoreticians can do over finite domains, using very limited ontologies, become possible for all sorts of other applications. Armed with an ontology that closely matches the real world, it is finally possible to write code that can reason about the real world, rather than just throw Bayes law at it and hope something useful falls out.
I'm prone to think that pg's algorithm is, to some extent, content based. I have no real evidence to support this, just the evangelism in "A Plan for Spam."
Aha! I'd not had the realization that the the downmod arrow only appears probabilistically. It meshes interestingly with a notion that dfranke and I were discussing earlier tonight, namely the appropriate ratio of up to down mods. By controlling the prevalence of options to downmod, you can exert some control over that ration--very sly. :)
An addendum, stimulated by dinner with dfranke: under certain assumptions about the growth rate of the community, karma ages without any algorithmic intervention: as the community grown, the number of moderators grows, and so the mean and maximum moderation per post increases. As a result, new post are effectively weighted relative to old posts.
The problems, of course, are that there is some fixed point, and also that news.yc may not fit the growth model that makes the above true for a community like Reddit--and that may be a good thing.