Skip to main content

Mike Sugarbaker

The robot and the echo chamber

4 min read

Pandora is a music streaming service which takes the name of a band or song, and delivers you a “station” full of songs it thinks is musically related. By “musically related” it means musically: it uses a bunch of criteria cooked up by something called the Music Genome Project, like tempo, “subtle use of vocal harmony,” “extensive vamping,” or “syncopated beats.” It lets you say whether you particularly like or hate a song it comes up with, and also lets you skip to the next song without expressing an opinion really. So: yeah. Yay Pandora. is a music streaming service which takes the name of a band or song, or the name of one of’s users, and delivers you a “station” full of songs it thinks is musically related, or sometimes just a list of songs with brief previews, or sometimes neither… but you can usually get to a radio feed of some kind if you stumble around for a bit. By “musically related” it means liked by other users who profess to like the music you searched for. It lets you say whether you particularly like or hate a song it comes up with, and also lets you skip to the next song without expressing an opinion really. So: yeah. Yay la– OK, wait a sec.

I kind of have a problem, both because seems to be emerging as blogdom’s favorite of the two, and because the two seem to represent an old problem of the net.

On, you will never (in my admittedly brief experience) hear anything unpopular. You will probably hear a lot that you like, provided that you start on a similar-artists-radio channel of an artist you like. But: last night I searched on TV On The Radio, fired up the radio, and eventually got the Pixies. Now, do I like both artists? Yes. Would I describe them as similar? No… no, not really. They are only similar in that I like them, and so do other people who like TV On The Radio. The band that nobody’s heard of yet, that does the kind of intense, musical out-there stuff that TVOTR does? won’t find them unless everyone already has. It’s a feedback loop, an echo chamber, it’s the nightmare scenario of intelligent-agent info feeds: no one ever hears anything they don’t already agree with.

On the other hand, when I play Pandora, I get a lot of obscure, bad crap I have to skip. Like, a lot. The robotic correllation of somebody’s musical categories seems to result in a lot of misfires with respect to what actually relates two pieces of pop music. But: getting back to the “obscure” part, I actually almost never hear anything I have even heard of, let alone heard. I discover things. All of my lengthy Pandora listening sessions to date have resulted in the purchase of at least one album. (Sometimes after a long, long search.)

There’s trolling the whole world of information and letting a lot of crap into your attention, if only briefly, to find the wonderful unexpected stuff. That can be a lot of work for infrequent rewards, but somehow the rewards feel really, really good. There is also going to the sources you know, and not worrying that there is something else out there – just trusting that if something is really worthwhile, you will probably hear about it. Why can’t we get these two models of information filtering to work together more productively? Or, how could we cut the Gordian knot and approach the problem in an entirely different way? Or, am I just obsessing about something obvious again?

Anyway Pandora does let you say whether you like or hate things, so it must be using the data somehow. I hope they add the more personal, social-software-y features that has, to make their tool less austere and more human, and do it soon. We shouldn’t have to choose between bottom-up and top-down, between cathedral and bazaar – that’s the other thing, that Pandora’s categories were made by experts and presumably applied by professionals, whereas basically is just the product of what people do anyway, via the site and its associated Audioscrobbler tool. People say that the top-down, made-by-those-who-know-what’s-good-for-you approach is now outmoded, but in this case it seems to have what folksonomy will never get us: the element of surprise.

And then Epitonic reopened. So I guess mu.

Mike Sugarbaker

Fortunately, great minds think alike

2 min read

Ace British tech Matt Webb and former Game Neverending developer Ben Cerveny may have just saved me a great deal of work:

playsh is a MOO-like text environment that runs on your local computer. The basic object types and verbs are based on LambdaMOO. It’s organised geographically, so you can walk north and south and so on. You have a player, so you can take and drop items. You can create new things and dig to new rooms, and there are verbs attached to all of these. There are ssh interfaces so you can connect to playsh from other computers, and other folks can connect to your playsh instance from their own.


I wanted to build on a couple of things that our brain does really well – geography and narrative – that are sidelined in our current “direct manipulation” paradigm (distance and locality are really powerful metaphors to build representations of knowledge and social structure on. Note that people have PhDs in geography, but not in picking shit up). I wanted something that was social from the ground up, as much as View Source is part of the browser, but also let people see the world in their own way, and experiment and share these different ways. MOOs fit the bill in so many ways.

Mike Sugarbaker

Another blow struck in the war on sobriety

1 min read

Liquid Breakfast at the Cricket
(click to enlarge)

This sort of thing is why I moved to Portland. No, not the alcohol part as such. Well, okay, a little bit the alcohol part. But you just don’t find things like this in the Bay Area, things that just make you say, “Awesome.”

Mike Sugarbaker

The real mystery for web developers

1 min read

It’s not “Why is MySpace popular even though it’s ugly?” MySpace isn’t ugly; it’s just simple, and doesn’t use, you know, rounded corners and Verdana and shit.

The real mystery is, why is MySpace popular even though its central features break every other time you try to use them?

Maybe ideas really do matter more than execution.

Mike Sugarbaker

A plea for the techier among you (and not among you)

2 min read

Dear tech community: thank you. Thank you for refactoring the job of web development so elegantly and comprehensively over the last couple of years. In every language from Ruby and Python to PHP and Perl down(?) to JavaScript, we now have powerful new abstractions that dramatically lower the barrier to building a web site that does something. Thank you. We needed that.

Now: why don’t you go to work on synchronous apps? If I want to make a web app I don’t have to completely stack blocks from the bottom up anymore, but if I’m putting up a chat server, or the back end to anything that’s really, really meant to work in real time, I still have to start from scratch. Granted, the scratch I’m starting from is better scratch these days, thanks to libraries like Twisted (love their new front page design). But still, there’s no reason I shouldn’t be able to write code that looks a bit like the following:

class MyChatRoom < SyncServer
   def start
     port = 7777
     welcome_msg = "Hi! Welcome to my chat server, please log in."
     welcome_prompt = "Login: "
     require_password = true
     user_store = "my_users.db" # or whatever
     prompt = "Say (/? for help): "
     commands = "file_full_of_commands.xml"
     spaces = "file_full_of_rooms_defined_somehow.xml"

chatter = chatter.start

...and end up with a chat room that basically works and has a few different rooms. Probably some of you can come up with a much better abstraction than the above, even. And I think it's clear how this is generalizable to a hundred different flavors of game server and interactive networked useful thingy.

Why has there been a lack of interest in this problem? Is it because chat is for the lower classes? Is it because lowering the barrier to making whatever sort of server you want is perceived as dangerous? Is it just because more people want to make web sites than want to make arbitrary other weirdnesses? Is it because people think Twisted is really all you need?

Mike Sugarbaker

Whispers in the darkness

1 min read

Does anyone know if those focused things, that just project sound into a certain spot, have ever been used in a theatrical production?

Update: more on focused sound projection from Make magazine (warning: PDF).

Mike Sugarbaker

LazyBar request

1 min read

Someone really ought to create a cocktail called the Florida Voter.

Mike Sugarbaker

For lovers of text from magazines

1 min read

If you’re like me, you enjoy reading text articles in a web browser. So you’ll be happy to hear that we relaunched KeepMedia, and the relaunch includes a great deal more free stuff. For instance, right now, our archive page for the sporadically correct Reason magazine offers a great article on the war between Prozac and a society with an increasingly Calvinist medical outlook. And it’s free, for now and for always. You’ll see good stuff in Psychology Today, the Atlantic and sometimes Esquire as well.

This relaunch is a start, and no doubt contains things that (still) suck, but we worked hard on it and I’m pretty happy.

Mike Sugarbaker


1 min read

I’m giving serious thought to unraveling this site’s current page-at-a-time-of-text-bits, newest-at-the-top model (which model is of course known as a weblog). ‘Cause, you know, a design that always emphasizes my latest thing really just reminds me how scattered and dillettante-ish I am most of the time.

The nice thing about modern blog frameworks is I can change all this with a little code-shuffling. I’d keep the RSS feeds, of course; their whole purpose is to tell you when something is new, so it’d be foolish to mess with that.

Any thoughts on what kind of ontology would actually make sense here? I fear divvying things up into “Tech” and “Other”… wouldn’t want anyone to think I’m one-dimensional. Or something.

Mike Sugarbaker

I got some code on you, sorry

1 min read

For lovers of convenience (that is, users of Firefox, Greasemonkey, and, Delicious Keys lets you hit ‘Alt + <N>’ to load the <N>th bookmark on a standard list of bookmarks. (The zero key gets you the tenth. After the tenth, you'll just have to touch the mouse – sorry.)

For lovers of other convenience who found Fictionsuit a little daunting when we first launched, you might not have noticed that we’ve introduced assignments: log in, and the front page greets you with a few sentence-long opportunities to make stuff up real quick.

Neither the “real quick” part nor the “assignment” part are really obligations, mind you. Assignments are seeds for your imagination if you’re not sure how to get going. So get going! We’ve got two exciting projects and more on the way.