Reminds me of Paul Graham's observation:

"Twitter is important because it's a new protocol. Fundamentally it's a messaging protocol where you don't specify the recipients. It's really more of a discovery than an invention; that square was always there in the periodic table of protocols, but no one had quite hit it squarely."

http://ycombinator.com/rfs3.html

Its not really a new discovery, the very first instant messaging system was half publish/subscribe: http://en.wikipedia.org/wiki/Zephyr_(protocol)

NNTP?

I'd like to see it go further, and see a machine pretending to be a person tweeting, a Twitter version of the Turing test. Or you could go one better, and apply machine learning to the social media output and other electronic communications of a class of users, or a single user, and use it to create a software personality that autonomously tweets, posts to G+ and Facebook, texts, maybe even posts its own code and forks other users' code on github. Then let human users see if they can tell software-based users from humans - or from a particular person it's been programmed to emulate. Does anyone know of anyone trying something like that yet?

Earlier this year there was a "Socialbot" project where bots controlled Twitter accounts and were awarded points for follows and replies from humans. The team that won (http://aerofade.rk.net.nz/?p=152) programmed the bot to ask questions and use some Twitter culture (Follow Friday). There are also quite a few funny conversations in the post.

Ohmygod, a LOLCAT based Turing machine - many thanks for posting this reply.

My favorite of these is http://twitter.com/#!/big_ben_clock, which tweets on the hour in London. (I copied this idea and made @sf_o_clock). The idea of devices broadcasting information to provide ambient data to anyone who cares to listen is so fascinating. This could go way beyond the obvious things like time, stocks and weather, and into... well, things I can't imagine yet.

off-topic: I have a perverse desire to put some microphones up, see if there's any birdsong recognition software out there, and tweet whenever a bird in range tweets :oP

That's not perverse, it's a great idea. Ornithologists rely on crowdsourcing to conduct bird population surveys and track changes in bird populations from year to year. Let the volunteer surveyors sign up for twitter feeds for nearby detection of less common birds to go and confirm.

This sounds like a less depressing version of the gunshot detectors Oakland put up (http://www.wired.com/wired/archive/15.04/shotspotter.html). Unfortunately, the ShotSpotters are not, to my knowledge, tweeting.

Pretty much everything that tweets in an automated fashion gets unfollowed by me pretty quick. The twitter stream is cluttered enough as is, no need to make it worse by having weather vanes and thermometers screaming alongside my friends who see cute cats.

Maybe I'm missing the point, but I have yet to find a Twitter bot that provides anything of real value.

I can see a few instances of automated tweeting generating value. I have a bot that tweets the latest headlines from the most popular blogs in a small niche. I'd say a large % of the active people in said niche follow the bot and it gets a lot of retweets and sharing.

Twitter is another way to consume information. RSS aggregators are popular which are automated, twitter can do the same thing over a medium people may be more comfortable with.

I also see value of these without you following them in your stream. I may not want to hear every weather update but when I want to know the weather, it could be nice to know where I can find it updating in real-time in a place I already am.

And to counter the inevitable 'but there is product/site/place X that already does Y': so what? If you look at user behavior just because something exists doesn't mean multiple competitors and ways to consume the same information won't be popular or useful.

I agree that it's unfair to say "so-and-so already does X", since innovating often just takes the form of making something slightly less crappy than it was before.

Unfortunately, my experience with Twitter, and especially automated Twitbots, is that Twitter is the crappy experience that needs to be improved upon. People push data into it because it's easy to do, not because it adds any real value.

Let's take the weather example: Auto-tweeting weather vanes are too noisy to include in your normal stream, so you unfollow them. Then, if you do want that information you have to use a search client, or build some kind of special client/consumer that compiles that info for you. At that point, surely it's easier to use one of the other services. Getting the data out of Twitter in a usable fashion is simply more work than getting it otherwise.

I agree with some of the commenters - this information is useful but only when I need it. Being bombarded with the weather every 15 minutes is pointless - I only need it when I'm about to head out for the day.

That being said, filtering all of this data via geofencing, time fencing, etc would be awesome.

I've always thought there could be an interesting use for Twitter just as a transport system for psuhing/pulling data between applications. You wouldn't follow those accounts, they'd just be data feeds. It's definitely not the use case Twitter started with, but it's viable now.

The advantage of using Twitter over another messaging system is that it's public and open. Anyone can easily push or pull from it without you having to build any additional infrastructure. On top of that it's reliable and can handle a lot of data.

I wonder about the honesty of machine sourced Tweets, not in the machines themselves but in the fact it would be super easy for a human to duplicate the tweets with faulty information. As a personal tool it could be amazingly useful (plants tweeting when they need water) but I don't see it working so well on a crowdsourced scale.