Home

Search

What would you build on Twitter and Pushbutton?

Author's Note: This post is a work in progress. It is subject to revision, updates and possible deletion, so please, share your thoughts and help me better thing through it.-MM

After a night of Entourage and PBR, we got to talking about Anil Dash's Pushbutton Manifesto (my term, his ideas) which would basically have a Twitter-like server in every home, or at least every home that wanted one, with a distributed, RSS-like real-time status update network.

The essay and ideas fascinated me, and in something of a brainstorming session, I tried to come up with ways to make such a system useful. My first impulse, and the one that's stuck around all night, is a sort of Yahoo! Pipes for the real-time Web.

Both my friends are engineers, and both have Twitter accounts they almost never use. I spent the evening idly sketching out use cases. For example, parsing through:

  • A user's current location (city or more specific), local Tweets, local event info and listings, friend's recent Tweets and locations
to get:
  • An event recommendation

Then, I figured it would be relatively trivial to take take a CloudRSS OPML, for example, and parse that for popular URLs on a given subject or within a given community, kind of like an automagical niche Digg that is resistant to the dumbness of crowds.

For example, rather than wading through all of the tweets from people Jay Rosen's following (as he suggest in one podcast) to get a pulse of what's going on in the Journalism Twitterverse, you take his Twitter OPML (which doesn't exist currently), use a Pushbutton server to grab all the recent updates, pull out the URLs people are Tweeting, and discover what articles are the most talked about in these categories.

This creates a hand-curated social news feed that avoids the gaming and dumbing down that have hurt Digg as well as Reddit and Slashdot as the communities get more popular (I mean "hurt" quality and utility wise, as traffic has certainly gone up). Since it's a very controlled list of submitters (in this case, people who Jay Rosen thinks carry weight on journalism) I'm getting a very high signal-to-noise ratio. While a few people may post pictures of their cat, kids, or breakfast, few will retweet it so those posts won't make the news feed, and the signal will remain high (the number of RT's required to make the feed would be set in the Pipes-like application).

With this Pushbutton pipes, I could then take, for example, the OPML list of people Jay Rosen is following, people the Boston Globe is following, and people the Boston Herald is following, and then parse out, for example, all the people more than 50 miles away. Then, if I wanted, only allow through Tweets with the word or tag "Boston," to get a local, Boston view from season journalists and sources.

A couple more use cases:

Bargain hunting

User drags a OPML list of Twitter accounts like @BostonTweet or GroupOnBoston. Maybe the list is national.

The user then narrows down the location they interested in, say, within 5 miles of Cambridge. Then, the user sorts for hash tags like #drink or #food from these users. Tweets with other hashtags might even get demoted for being "off topic," then Tweets without tags in the middle, then Tweets with the most tags on top. At the very top, the Pushbutton Pipes setup might sort the output by most times Re-Tweeted.

Apartment hunting

Apartment hunting is a relatively inefficient, horrible process, so this is just a first iteration but there's plenty of potential here. First, perhaps do a local search for #apartmentsforrent metadata, in Tweets, yes, but also Waves, another, more extensive real-time data storage.

The user narrows it down to 4 choices that come up, some Tweets that lead to traditional Craigslist-style advertisements and other that are live, editable Waves. Since a certain amount of real-world relationships are shown, the renter sees he has several mutual connections with 2 of the landlords. He pings them, getting negative feedback on one landlord and positive on the other.

One of the apartments the renter wasn't able to get a reference has a Wave, and so the renter asks some qualifying questions not addressed yet: Are pets allowed? What about parking? The answers are given out in the open, on the Wave, so the community can all see the updated apartment info. The Wave is associated with a showing time, which users can sign up for right on that page.

The fourth apartment, with neither references nor a Wave to ask follow up questions, is left for later consideration if none of the earlier choices work out.

Make sense? Think of other useful cases for a Pushbutton Pipes set up?

Further Reading:

Nada Bada Bing? Microsoft search "porn" easily blocked by schools

BadaBingFound via BingThere are a lot of things not like about Microsoft's new search engine, Bing, but children getting access to porn through its video preview shouldn't be on the list.

The fact is, every filter (read: porn blocker) is going to have loopholes or workarounds, or else it will signal so many false positives as to make the Internet useless. Blame it on pornography's subjective nature. Blame it on how big and complex the Internet is. But all this logic seems to get thrown out the window as soon as someone whispers "But think of the children," and soon you have not only TechCrunch issuing warnings about Bing's video porn preview (they should know better), but also CBS News ("Parents Beware" the headline warns) and, inevitably, Fox News.

Please. Microsoft has too much riding on this to let it stay a black eye for long, and indeed, shortly after the Bing porn previews became public knowledge, Microsoft offered one way to close the porn hack, and then another.

Most telling, however, has been the response on an educational IT listserve I'm on. One administrator issued a challenge: "For those that have really good filters, try logging on as a student then BING and go to video - search porn - then put the cursor on one of the videos. Preview time."

Not so fast, one respondent who, you know, actually did this:

Our filter, Web Gateway by McAfee (formerly Webwasher by Secure Computing) appropriately blocks the content. Web Gateway enforces "safe search" and students are not able to change the option to turn off safe search. Additionally, searching porn related terms resulted in content that stopped the pages from being displayed.

And then three other school administrators, all using different filters, reported similar blocking success. One had to tweak some manual settings, but the other two worked fine blocking errant Bing queries out of box.

If school IT administrators aren't worried about Bing porn when it's their job to stop this stuff, why the hell are Fox and CBS News in an outright, end-of-days panic?

Oh, right. Ratings.

WolframAlpha thinks I need to drop 25 pounds

Actually, just 24 pounds, but it still stings.

I tried a lot of queries without success on WolframAlpha, the new "computational knowledge engine," ranging from rates of homelessness and AIDS in various regions to salaries for college graduates. None of the first 20 or so queries gave me back meaningful results. Then I decided to hold myself under Wolfram's unflinching light by putting in some of my vitals. After all, Googling oneself is a time-honored tradition, so why shouldn't Wolfram'ing oneself become the same?

Here's why:

WolframAlpha thinks I'm fat

While I know I could afford to drop a few pounds, 186 lbs for a 6'3" male sounds pretty downright unhealthy, and is about 20 pounds south of what my doctor recommended with his weight chart.

Obviously, WolframAlpha's databases could still use expansion (that or I really am a lot worse off than I thought).

But while these results weren't exactly pretty for anyone involved, I still see some strong potential for the service (which is NOT, as has been reported again and again, a "search engine," at least not as the term is commonly used today).

Mr. Wolfram's introductory screen cast alone was enough to get me excited, even if, as a friend remarked, playing around with it for a few minutes shows just exactly how close you have to stay to its databases to get meaningful results.

That said, the empty handed results on AIDS, homelessness, global warming ("Functionality for this topic is under development..."), post-graduate pay, etc. etc. shows not the engine's weaknesses, but how much potential it has as more databases, licensed or under public domain, become available. Having an engine that calculates and beautifully displays previously impenetrable data sets could be a huge asset to journalists, researchers, and everyone else with an inquiring mind.

In the meantime, there are plenty of cool things WolframAlpha can do. I suggest:

I really disagree with Silicon Alley Insider founder Henry Blodget's thoughts, that WolframAlpha's just another search engine destined to fade:

Our prediction: Wolfram Alpha (terrible name) will see a nice spike in traffic for a few days, then it will disappear unnoticed along with all the other "next-generation" search engines.

Why?

Because search isn't broken. It can be improved, yes, and companies like Wolfram Alpha will show Google how to improve it. But no search engine we've seen, including this one, comes close to making the quantum leap in performance required to get real volumes of Internet users to switch.

Specialized data base search engines have actually done quite well. Orbitz, Travelocity, and Priceline are some prime examples (and ones with great business models to boot), but Zillow's also a fan favorite, despite supposedly terribly inaccurate house price estimates that realtor's routinely gripe about.

Is Google broken? No, but neither was Yahoo or Lycos, really, when Google burst onto the scene, and to be honest, calling WolframAlpha a "Google-killer" mostly misses the service's point, which is not to search but to calculate specific results based on verified data.

Further Reading:

Syndicate content