Things That Tweet

I was at breakfast with a friend yesterday who told me about a project he did with some Twitter data around weather. He said as he was pouring through the data, he saw that there were bursts of tweets at certain times. He dug into the data and saw that it was weather vanes and thermometers that were tweeting out their data.

It got me thinking about things that tweet (like weather vanes, refridgerators, traffic lights, etc) and their role in the land of social media. I believe that devices and sensors that broadcast their data via social media channels are an important source of social data and engagement. And for some reason, they are way more common on Twitter than any other social platform.

Have you ever seen a weather vane on Facebook? I have not. If they exist I'd love to know about them. I want to understand the Internet of Things and its role in social media. I suspect that the symmetric friending model and the use of real names/real people in the Facebook system is a hindrance to devices updating Facebook pages, but I could be wrong.

Services that are too determinent in their use case are ultimately limiting in their extensability to important new uses. Machines are reliable sources of information and the social services that are friendly to them have a number of interesting opportunities in front of them.


Comments (Archived):

  1. William Mougayar

    There is lots going on in the Internet of Things. GE for eg is all over it. Eqentia has a portal on that to stay updated:… And this is the Twitter account to follow @InternetThings

    1. Vasudev Ram

      Eqentia looks interesting. Will check out the portal, thanks for the link, and I’ve followed the InternetThings Twitter account.

  2. Joe Yevoli

    The new engineer on board HomeField created a product similar to what you are talking about –  The description : An application platform designed to enable meaningful connections between things and people.I think its a very impressive app.  Thought you’d be interested in checking it out.

    1. fredwilson

      thankswe will check it out

  3. Dan Thornton

    Sadly one of my favourites was London Bridge, the Thames crossing which has to raise or lower for ships to pass underneath, and which had its Twitter username claimed by (If I remember correctly), some tourist body or travel agency which didn’t have much claim to it in the first place…It was a real shame, because there was something curiously engaging and wonderful about having a bridge announce that it was raising/lowering with the name of particular ships passing by, and it was an interesting example of transferring some human characteristics onto a particularly large impersonal object. I wonder if anyone was more patient when waiting to cross because they’d seen the Twitter account?

  4. JB

    There’s some really shady ads showing up when reading your posts in Google Reader:

  5. Matt A. Myers

    Quick response here (shocking, I know) –From my own view it doesn’t seem like Facebook users create this kind of update, but that could be because of bubbles created for only showing me content that I’m more likely to interact with.

  6. ErikSchwartz

    I’m not sure the term “social” is appropriate for things that communicate.

    1. awaldstein

      Good point….social data comes as a result of interaction. It’s the offshoot. Data from things is fodder for socialization.They meet and are part of the same ecosystem but they are not the same.

      1. Fernando Gutierrez

        And because they are part of the same ecosystem I also would like to be able to separate them when I want. In the case of Twitter, when I check my stream I’m usually in one of my three modes, and things are not always welcomed :-General reading. Links to content and a few tweets by themselves. People/things I follow.-Looking for specific information/news. Search is great and I don’t mind if I get results from things.-Socializing. I only want people.Some apps help with this, but I haven’t found the good one yet. Maybe lists could do it, but maintaining them is a pain.(edited for some format)

        1. awaldstein

          True…being able to parse the streams of the Twitter matrix to suit your needs is powerful. It’s our world to jump in and out of and to stylize to suit our needs. You are talking about curation of the Twitter streams at will. We need that.

          1. saranyan

            Yeah, and the curation should be automatic. I believe that reinforced learning methods go a great length in solving this problem.

        2. Boris Wertz

          Check out Summify to get a social summary of the most important news / content

          1. saranyan

            Isn’t “most important” subjective? I also feel that it varies from time to time based on what communities we are plugged into and what we are working on.

          2. awaldstein

            Yes…I agree. We need control of our filters. Social referrals. Contextual filters. What I call ‘adjacent on the interest footprint’ filters. One view is narrowing. Adjustable views are enlightening and broadening.

          3. Boris Wertz

            Agreed ­ most important is defined by what you feed into the system (i.e. Twitter followers, RSS feeds) so if your interests change, so should your feeds

          4. awaldstein

            Summify has been the early best one (I was an early user) but I need more adjustable filters to really get the job done.Check out Eqentia from @wmoug:disqus He’s built a platform for the enterprise that has a lot of flexibility and worth looking at.Honestly though, my big question is not only the versatility of the curation but what we do with it. If I can’t connect and discover communities of interest, then it’s only a small part of the way there.

          5. Boris Wertz

            Fully agreed ­ a social layer on top of the curation is extremely important.

          6. Fernando Gutierrez

            Summify looks great, thanks. For similar purposes I use @falicon:disqus (Kevin Marshall)’s and I’m quite satisfied. But still it doesn’t read my mind! 

          7. falicon

            Hey thanks for making sure people – I love having users like you!Oh and we are actively working on that ‘reading your mind’ part… 😀

        3. Donna Brewington White

          “Socializing. I only want people.”ha!!!

      2. Ed Borden

        ^^^^^ This is slam dunk right here.  No one cares to talk to machines or for machines to talk to them. What I think is far more interesting, and I think what Fred is really tuned into as someone who invests in “communities of engaged users”, is how machine data can drive engagement.We saw this play out first-hand at Pachube around radiation data from Japan.  It’s amazing how a tens of thousands of people from all over the globe can all of a sudden spotlight on an old analog Geiger counter connected to an Arduino sitting on a lab bench in the backwoods of Japan. Why do people tweet their Pachube feeds? To spur a conversation about their data.  Figuring out how to facilitate that conversation is the opportunity.

        1. fredwilson

          yup. that’s my point. the weathervane tweets, and i RT it or reply and then people are taking

          1. FAKE GRIMLOCK


          2. Chris Kalima

            This is exactly what I’m working on at Buoy Alarm, building a social component around real-time ocean buoy observations with the intent of creating additional value for users.Granted this is a specific niche in the Internet of Things, not a broad spectrum of “sensors, devices and environments,” but it serves a passionate community that is not only interested in the real-time data, but the historical data as well.”Things” really begin to get interesting when you link their data (in our case ocean conditions) to a historical record of social engagement (notes, likes, ratings), which in turn creates long-term value for both the individual as well as the community.These concepts are nothing new, but their application to the Internet of Things is just beginning, and will continue to grow as systems and devices get smarter and better connected. Great post!

        2. awaldstein

          Agree. What I like about Twitter is once your parse the data stream to create a contextual group or situation, you already have the dynamics of a community in place. That’s its power to me. You can curate information together but you still need to connect and have the core components of conversation. Twitter is a dynamic stream at its core.

        3. fredwilson


        4. Werner Weihs-Sedivy

          full pull 🙂

    2. Guest

      I agree with your thoughts @ErikSchwartz:disqus and @awaldstein:disqus reply below. Data like that Fred mentions above could be meaningful* but I am not sure it is ‘social’ in any manner or true semblance.Note(s):(1) While I do think data generated via a Tweet like Fred suggests above could be meaningful I think @ Taka makes EXCELLENT points in his reply thread and discussions w/ @L1ad:disqus above. Really took their discussion to heart.  [Updated]

    3. Dave Pinsen

      I’m not sure the term “social” is appropriate for lots of people who communicate. Consider all the use of bots, auto replies, auto tweets by people. I’d rather get automated tweets from things than automated tweets from people. 

      1. Donna Brewington White

        ” I’d rather get automated tweets from things than automated tweets from people.”Word

    4. ShanaC

      Isn’t the basis of all things that we do that is social the ability to communicate?I mean, your toaster doesn’t have a personality quite yet, but by talking to you, it does “communicate” in some sort of limited way



        1. ShanaC

          Toasters communicate when your toast is done- no?

    5. fredwilson

      What is social is our engagement with these things that tweet

      1. ErikSchwartz

        You’re not engaging with these things, these things are incapable of engagement on a social level. A social relationship has a give and take.What is interesting is engagement with other people in regards to the data streams these things provide. The box score of a baseball game is not social, the discussion around the box score is. That does not mean that box scores and other automated tweeting things do not add value to the stream. They are a jumping off point for people to talk to one another.

        1. fredwilson

          Yup. That’s what I meant but not what I said

        2. Werner Weihs-Sedivy

          In my opinion things (via their avatars) are getting spirit if you as a human could talk to them in natural language. thats a main focus in our development.

  7. RichardF

    I’ve seen quite a few applications of machine data emitted by sms.  I like the idea of it being transmitted over the net but preferably using an open standard like xml.The possibilities for using the data are far greater if a third party like Twitter don’t own the firehose.

  8. LIAD

    I think the Facebook comparison is inappropriate.Unlike Twitter, they’re not trying to be ‘the world in your pocket’ or to ‘instantly connect you with the *things* most important to you’ – for good or for bad they have a much more defined use case.For Twitter to deliver on it’s value prop it needs to support and encourage data streams coming from anything and anyone. Information emanating from inanimate objects can be equally important to information coming from living breathing ones. Twitter is the perfect tool to use to plug into these streams. Asymmetric relationships, bitesized data, delivered instantly, with minimal social norms baggage attached.

  9. Simon de la Rouviere

    Twitter’s ephemeral nature (and asynchronous follow model) is what allows this type of behaviour. It is what happens ‘now’. Facebook’s interaction feels like a billboard, while Twitter feels like a flow of information. Another interesting question related to this is why radio dj’s use Twitter more than their Facebook page when on air (something which I’ve seen): the interaction is about what happens now. This is also why Obama used Twitter to answer questions and not Facebook.With Jack Dorsey laying off product guys, I hope Twitter doesn’t change too much. This is what makes it, it is. Facebook is where information goes to die and stagnate, while Twitter is where information flows and happens.That’s the way I see it. 🙂

    1. fredwilson

      Jack is responsible for its creation. I can’t imagine a better caretaker

  10. Taka

    I’m interested to know why you think that “devices and sensors that broadcast their data via social media channels are an important source of social data and engagement”.I do a lot of work with RSS and have been musing for a long time now about the importance of RSS in the rise of machine-to-machine communications e.g. “blogjects”…One of the key ideas is the broadcasting of information in a form that is easy for machines to understand and do something useful with.Now, I’d be the first to admit that I’m not a big fan of Twitter and I find this idea of having devices automatically tweet information somewhat pointless e.g.…The London Bridge channel is a good example – it might be cute and fleetingly interesting to have a bridge announcing that it’s going up or down, but how is this socially useful or engaging? A webcam pointed at the bridge would provide the same information, but would anyone tune in? They might have when webcams were new and a novelty :shrug:It seems to me that information that is genuinely useful in a social, networking sense, is that which has meaning and relevance to both the sender and receiver. Having a machine mindlessly shouting out across the ether what it’s doing every hour, on the hour, doesn’t quite seem all that engaging. Other machines may well be interested, but people…?

    1. LIAD

      tons of automated things i’d like to hear from in near real-time.train delays. temperature of the pool in the gym. queue length at hairdresser. kid’s arriving at school safely. there needs to be a channel to receive all these feeds – the twitter protocol works great.

      1. Taka

        >>> tons of automated things i’d like to hear from in near real-time.train delays. temperature of the pool in the gym. queue length at hairdresser. kid’s arriving at school safely.Yah, but I bet you don’t want to be hearing about all of these things all the time! You only want to know about these things when you need to i.e. the point I was making about information needing to have relevance.Twitter is dreadful and completely unsuitable for this kind of thing. It’s a firehose of raw data – as if we don’t have enough information overload already :-/ – and it’s free-form, un-structured data i.e. impossible for another machine to decipher. But if this information was published in a structured form, that’s a completely different story. You could have an app on your phone that subscribed to this data feed and you could check it, when you needed to, to get the current status of your train, pool, hairdresser. Much more useful.Machines are supposed to make our lives easier. Having them spit out enormous amounts of raw data and then have us to wade through it looking for something that might be of use seems somewhat sub-optimal :-/

        1. LIAD

          Going forward i will be offered personalised versions of those kind of accounts. where i could specify frequency etc or set parameters to trigger the tweet. – only inform me ifa. kids haven’t checked into school by 10amb. temp at pool falls below X celsiusc. queue at hairdresser is above 3

          1. Taka

            >>> i will be offered personalised versions of those kind of accountBut who’s going to offer that personalisation? Twitter simply transmits information being submitted from other sources, and it’s un-structured data so they would have no way of interpreting it anyway. You can’t expect every publisher to offer personalisation services, so it has be done client-side. But again, because the information is un-structured, it’s very difficult to process.Twitter is the wrong tool for this particular job. It’s OK for human-to-human communication (and in particular, interaction) but the applications you describe are not that, they are “tell me the current status of…” applications and accordingly, would work better with a different framework.

          2. LIAD

            I hear you.I think twitter is a perfect conduit for that kind of info. I appreciate the tech difficulties you mentioned though.

          3. Hamutal Meridor

            That’s exactly the premise behind, which is why I find it so brilliant.Twitter is perfect in what it does, it’s simply not designed to do what you’re looking for. They’re merely a conduit for data. BUT – that’s why their API is open! When companies such as ifttt start using it to bring the right person the right data at the right time, then we’ve got something, don’t ya think?

          4. Boris Wertz

            Agree that this has to be done client-side. Summify for example does a nice job on aggregating the most important news based on social signals of your Twitter feed but news is much easier to figure out that some of the other information verticals.

          5. saranyan

            But what if there is a tweeting standard? Some convention for devices transmitting such data? Then, there is possibly a structure to this data. No?

          6. Rick Bullotta

            We built that filtering into a layer between our client apps (mobile or web) and the streams of data and events. On the mobile app, it supports notifications. A maintenance technician can follow machine events, a salesperson can follow a production order through to the customer, a physician or caregiver can monitor important events for a homebound patient, etc…

        2. Vasudev Ram

          Taka:Re. your above paragraph that starts with “Twitter is dreadful and completely unsuitable for this kind of thing. It’s a firehose of raw data” …The points below are based on what I understand currently, could be somewhat wrong in some cases:1) Twitter *is* meant to be a firehose of data.2) The data is not exactly raw – it only looks that way in Twitter clients like the official Twitter Web client ( or other clients that present the data for human viewing (most of them present it in HTML format, but HTML, though actually a structured format like XML, when rendered in a browser, looks like a flat page of text to humans, except for the clickable links).The actual Twitter data is also available in different *structured* formats (JSON for sure, and XML probably – need to check, but likely) for machines to decipher. See, as just one example, the docs for the Twitter API to get the public timeline, at:a)…and the example shown therein, i.e.b)…The above link is a standard HTTP GET call, which can be easily generated and run by most programming languages. Your browser can run that link too 🙂 Paste the link into your browser address bar and take a look at the output you get.The output is in JSON format, which can very easily be parsed and processed by most mainstream programming languages, because they all have JSON libraries available – see for the libraries; in most languages it takes just a single function or method call to convert the JSON data into a hierarchical data structure native to the language, such as a list of dictionaries in Python, or the equivalent in other languages. And in JavaScript, in fact, it is already in native format, because JSON *is* (a subset of) JavaScript:From”JSON (JavaScript Object Notation) is a lightweight data-interchange format. It is easy for humans to read and write. It is easy for machines to parse and generate. It is based on a subset of the JavaScript Programming Language”See also:…So apps can easily be written to parse that raw – but structured – data, and extract various types of information from it that is more meaningful and relevant to humans, including various types of filtering, sorting, aggregating, etc. of the raw data. And many such apps already exist.

          1. Taka

            Yes, you could be doing all of this, saranyan suggests a “tweeting standard”, but my question would be “why on earth would you bother”? It’s like trying to force a square peg into a round hole by filing off the corners, sanding it down, then whacking it in with a hammer. Why not just use a round peg?We already have mechanisms for transmitting this kind of information, HTTP over TCP/IP, any number of data formats (XML, JSON, RDF, etc.), that are well-understood, proven and widely supported. Why do we need to twist ourselves into knots coming up with a way to do this kind of thing over Twitter, that was never really designed to support it, when we already have better ways? Because it’s the new, cool thing? Honestly, does anyone really think these guys are going to be around and relevant in 10 years time? MySpace, anyone?

          2. Vasudev Ram

            Taka:Hoo boy (or girl) ………… :)You really don’t seem to be clued in to what’s going on, do you? Also, you don’t seem to read comments fully before replying to them. And why no info about who you are? You don’t need to register on Disqus to comment on this blog, but you can give some info such as a real name and email address, even if you don’t have a web site. Seems shady to me …To address your reply: You clearly didn’t notice that I mentioned the very same things – HTTP, JSON, XML – that you are talking about in your reply. And that those things and how to use them, *programmatically*, using the Twitter API and JSON parsers, etc., was the main thrust of my comment.All in all, your comments lead me to wonder whether you know jackshit about tech or life.EDIT: And don’t use phrases like “Why do we need to twist ourselves into knots coming” without showing a shred of evidence or even statements about your actually twisting yourself into a knot.

          3. Taka

            In response to vasudevram’s post below (it has no Reply button – does Disqus have a max.depth for replies? :-/)…>>> Also, you don’t seem to read comments fully before replying to them.I might say the same about you :-/ If you read the comment I originally posted that started this thread, I posted links to some of my weblog posts that discuss the same issues Fred was talking about in his OP. Taka is my real name. And I supplied my email address when posting comments, so if Disqus doesn’t show that in the UI, you’ll have to take it up with them.>>> You clearly didn’t notice that I mentioned the very same things – HTTP, JSON, XML – that you are talking about in your reply. And that those things and how to use them, *programmatically*, using the Twitter API and JSON parsers, etc., was the main thrust of my comment.Simply because something is being sent using JSON or XML doesn’t mean it’s structured. If a swimming pool is emitting a feed that looks like this (substituted angle brackets for the square brackets, I don’t know if Disqus is going to mangle it):[temperature]It’s a lovely 22 C degrees![/temperature]then this is unstructured data (despite being XML) – you still have to parse the string to figure out what it means and it’s fundamentally no different to scraping a web page. If the pool issues this:[pool][temperature]22[/temperature][unit]centigrade[/unit][/pool]this is structured data, it’s unambigious. In the first example, if they change the format of the sentence, change “C” to “centigrade”, you’re screwed. In the second example, they will (or should) publish a schema and it’s far less likely to change, and even if it does, the schema can still be versioned. Much more robust, much more suited for machine-to-machine communication.And if you’re arguing that we should be sending structured data over Twitter, why on earth would you do that? What’s wrong with plain old HTTP? It’s light-weight, it’s widely adopted, it’s non-proprietary, it’s proven, there are any number of optimization techniques that are well-understood and widely deployed. What advantage does Twitter offer other then being the cool new thing?>>> All in all, your comments lead me to wonder whether you know jackshit about tech or life.As a former (assembly language) kernel developer, I think I know a thing or two about technology. As a former technical lead for a commercial search engine, I know a bit about information, meaning and semantics of data. And as for life, I know the importance of civility.

          4. Vasudev Ram

            Replying to Taka’s comment below (due to no Reply button), the one with the 2 data examples.Just wanted to say, right away, that you were right and I was wrong. I apologize for the statements I made. However, there were  at least some (though mistaken) reasons for it, details to follow soon.UPDATE: Here are the above-mentioned details:Well, I seem to have quite misjudged you. I apologize, without reservations. A few partial explanations for my earlier words, below:>>>>>I might say the same about you :-/ If you read the comment I originally posted that started this thread, I posted links to some of my weblog posts that discuss the same issues Fred was talking about in his OP. Taka is my real name. And I supplied my email address when posting comments, so if Disqus doesn’t show that in the UI, you’ll have to take it up with them.Actually, I didn’t read your original (_first_)comment in detail, though I must have scanned it. I was initially scanning everyone’s comments, somewhat fast, after reading through the main post, looking for particularly interesting points of view by commenters. So, though I must have (subconciously, at least) noticed your blog links, I did not remember about them when I came to your 2nd comment in the tree – the one to which I replied, because all my attention was on the 2nd comment, while composing my reply. Alternatively, it could also be that I did not read your very first comment until later, sometme after posting my first reply (to your second comment), because of these two factors, combined: a) I sometimes refresh the page after reading (and maybe replying to) some comments, and b) Disqus has a default comment sort order of “by popular now”, due to which the positions of comments can change over time as comments change in popularity, and that position change takes effect when you refresh the page.But I did hover my mouse over your name (above your 2nd comment) and did not see any hyperlink, nor any email address – yes, that seems to be a Disqus isssue, at least for me (and I also checked again just now, it still shows that way). So, after I read your comment (the one in which you say Twitter is a firehose, and unstructured data), and given the fact that I seem to have misunderstood your meaning of unstructured data (I now agree with your view – see below), and hence thought that _you_ were wrong in calling Twitter data unstructured (whereas, as you correctly explained later with those two examples, it _is_ unstructured, so it was me who was wrong on that), plus the fact that you _seemed_ to be (not are) a troll, based on apparently no email / URL of yours being supplied, and also given the fact that I’ve come across trolls in the past – you’ll know well enough what they are and what they do, so no need to describe that; well, given all these above points, my conclusion was that you were a troll and/or a person who had very half-baked knowledge of what he/she was talking about and/or a person who was just throwing technical terms around with much knowing what they meant, maybe just to impress. My later uncivil language was caused by this impression; it was an attempt to push you off this discussion and blog, since I thought you really were a troll who was talking meaningless stuff, until I read your last reply, in which you explain what you mean by structured data, gave details about yourself and mentioned your blog links.>>>>>Simply because something is being sent using JSON or XML doesn’t mean it’s structured. If a swimming pool is emitting a feed that looks like this (substituted angle brackets for the square brackets, I don’t know if Disqus is going to mangle it):[temperature]It’s a lovely 22 C degrees![/temperature]then this is unstructured data (despite being XML) – you still have to parse the string to figure out what it means and it’s fundamentally no different to scraping a web page. If the pool issues this:[pool][temperature]22[/temperature][unit]centigrade[/unit][/pool]this is structured data, it’s unambigious. In the first example, if they change the format of the sentence, change “C” to “centigrade”, you’re screwed. In the second example, they will (or should) publish a schema and it’s far less likely to change, and even if it does, the schema can still be versioned. Much more robust, much more suited for machine-to-machine communication.I agree with your explanation of what you mean by unstructured vs. structured data now, and with your conclusion based on them, that structured data is more robust, with a minor dissent: I would say that if they change the “C” to “centigrade” or any other error-causing change, for that matter, it must be done either accidentally or maliciously, and the same kind of error-causing change can equally well happen with the structured version, for either of those reasons (accident or malice), and not just with data within the XML elements but even within the element tags, e.g. [pool] could be accidentally mispelled [poll] during an edit of their code. But overall, it is true that structured data, as you describe, is more suited for machine-to-machine communication, and also does not require hand-coded parsing, which the unstructured example does. In fact, I have often thought that there should be no need for screen-scraping at all, i.e., if all HTML data was structured and with proper semantic markup, computer processing of HTML pages to get meaningful information would be much easier. But of course, that is easier said than done, given the legacy of existing pages, and the fact that different people may use different semantic tags for the same meaning, and vice versa, and many other such issues …>>>>>And if you’re arguing that we should be sending structured data over Twitter, why on earth would you do that? What’s wrong with plain old HTTP? It’s light-weight, it’s widely adopted, it’s non-proprietary, it’s proven, there are any number of optimization techniques that are well-understood and widely deployed. What advantage does Twitter offer other then being the cool new thing?I was not actually arguing that we should send structured data over Twitter (when HTTP will suffice). As you point out, it can, of course, be directly sent over HTTP, the server can serve it and the client can fetch it with a GET call. Simple HTTP client-server stuff. My point was that the post’s discussion was already about Twitter, and I thought you were only talking about Twitter (which uses HTTP and TCP/IP under the hood), i.e. its JSON/XML data, and calling it unstructured, which I initially disagreed with, till I read your last reply where you use those two data examples. When, in your 2nd-last comment, you said why Twitter, why not HTTP over TCP/IP, etc., due to not seeing an email id or a URL of yours, and also due to not understanding what you meant, I thought you were just throwing words around.

    2. Joe Zydeco

      “it might be cute and fleetingly interesting to have a bridge announcing that it’s going up or down, but how is this socially useful or engaging?”If you’re trying to drive across that bridge and it’s up, you might want to take another route and avoid the traffic delay.  THAT’S how it’s socially useful.

    3. saranyan

      That is an interesting take. I feel that there is value in having these “social beacons” communicating their status is some way. It will be useful in some cases and I do agree that they are pointless in others.I would love to check on my devices or sprinkler systems or stuff like that. But, you have nailed the point, it has to be useful to me. But, people doing it is no fun. 

    4. Dave W Baldwin

      Think of it this way.  It all comes down to the end user and the machine delivering what it knows the user wants.  Yes, the bridge going up and down will have low ratings, but something of interest to a bigger group will have higher and the user can have a scan/read on what is going on in real time.

    5. fredwilson

      I like the broadcast model better than the telephone model

  11. William Mougayar

    There are more devices than people on earth connected to the Internet, ref. this Infographic from  @cisco:disqus… 

    1. ndebock

      Yes and that is why I always fought that Cisco is the one that is going to buy twitter and mke it the inetnet of things OS on top of its routeurs! 

  12. andyswan

    Does this one count?…

    1. Morgan Warstler

      Joan is definitely vain.  Speaking of vain… what’s with Carly Simon singing “You’re So Vain”…to Warren Betty????   The song IS about him.Women are weird.

    2. fredwilson

      Not sure

  13. Julien

    I think you stress one of the many facets of Twitter as a communication backbone (infrastructure) rather than a media company. I (as an engineer) love this aspect of Twitter, and this is the fruit of the early year’s push toward APIs, and a platform play.However, I wonder if this is still a priority for Twitter which seems to favor its “media” business facet more, with a business model that seems more directed toward advertising…

    1. LIAD

      completely become the broadcast hub for the ‘internet of things’  – twitter would need a ton of grass-root developer buy-in and love.I think those days may be long gone.

      1. fredwilson

        You kidding me? I suspect, but don’t know for sure, that the rate of uptake of API keys is faster than ever. The reality is often very different than what the media and blogosphere make it out to be

    2. fredwilson

      I think you can do both

  14. John Canosa

    While there is utility in the asymmetric Twitter model for these things, as a previous comment suggested there is a need for the subscriber to be able to pause and filter these updates.  Machines can be quite chatty, droning on and on about events you only care about during certain times.  One of the main problems I have with this model is that it is not bidirectional.  For an end user who may be just consuming the data from some sensors this may be OK. However  for a business who is relying on a certain piece of equipment (sensors+actuators+intelligence) it is not.  The ability to reach out and touch these machines to update, maintain, configure, and even interact with remote devices and end users is just as important.  Check out for an example of a company that does just this and integrates all of this functionality into a true business context.That being said there can also be great value in the more symmetric Facebook model.  As an example think of a soda or ice cream vending machine.  Why not be able to create a relationship with the end user.  What if the vending machine knew it was your birthday and  gave you a free ice cream?  What if you could use your ‘friendship’ with a vending machine to treat a remote friend to a soda to brighten up their day?  There are some pretty powerful marketing possibilities with this model.

  15. Christina Cacioppo

    Way back in Facebook history – summer of 2004 – one of the most popular profiles on the site seemed to be Keggy the Keg, Dartmouth’s unofficial mascot:…It didn’t take long before Facebook took the page off their network because Keggy “wasn’t a real person.” The Dartmouth kids promptly started a new Facebook group called “We miss Keggy the Keg!” which attracted even more members than Keggy had friends. I’d bet it was early policy decisions like these that set the tone of the network: real names, real identities, real people, all the time. 

  16. jmcaddell

    Fred, two points here.One reason that Twitter is an attractive platform for machine messaging is that its protocol is simple and compact. You don’t need much bandwidth to transmit 140 chars.As far as all that data being noisy and pointless, as some commenters have observed, it may depend on who is consuming it. If I am approaching London Bridge and it is about to go up, that’s highly useful information for me.So part of the value equation are various front-ends that make sense of this torrent of information for particular groups. For them, having a predominant method of machine communication and the ability to subscribe and unsubscribe at will, as Twitter provides, is useful and valuable.regards, John

    1. fredwilson

      I totally agree with you

  17. Kasi Viswanathan Agilandam

    There are billions of things that talk (with data or visual). Make them all tweet is twitters challenge (adaptability). Taking those 20 or 50 signals out of that that zillion noise is the biggest challenge for the rest of the world. 

  18. Barry Nolan

    Sidestepping social, what about ‘things that tweet converse’.  Specifically the ability for humans and things to converse (preferably in mobile apps).  Take a BoA email alert (thing=fraud system) I got last week – my cards were used in London. Great – but its read only.  What I want is to converse two way, in app, and to Tap of ‘ok’ or ‘call me’.  3 seconds, and I’m done. File  under convenience – “a thing that contributes to an easy and effortless way of life.”

  19. Morgan Schwartz

    Some of my favorite early prototypes:Noise pollution sensing coconuts:…Air pollution detecting pigeons: http://www.pigeonblog.mapyo…opportunities for sensing and reporting data are all around us – will there be a common protocol, will the data be open?…

    1. fredwilson

      Yes and yes because that’s what is necessary for this model to work

  20. Jose Muanis

    Check that’s really good.

  21. gregory

    This reminds me of a funny story just as I was getting into the Twitter ecosystem. We’d launched the Shorty Awards on a whim in December 2008 and much to our surprise it took off and became the top trending term on Twitter within 24 hours. Nate invited us to present at the NY Tech Meetup in January 2009 along with Stocktwits, Cotweet and Klout — all of which had just a couple people at the time.I was demoing how you could compose a tweet nomination at and click a button that’d push you to to send the prepopulated tweet (oAuth wasn’t around yet).  All of the presenters had to use the same laptop, so I figured to save time I’d just send the tweet with whatever Twitter account was already signed in on that computer. Just before clicking the button to send the tweet, two women from the front row screamed out “No, it’s a plant!”. I thought they were crazy, so I sent the tweet.It turned out they were the next presenters. They’d built a device called Botanicalls… that you attach to a plant — the device tweets as the plant letting followers know if it needs water, sunlight, nutrients, etc.So I’d tweeted a Shorty nomination as a plant. Needless to say I felt awful about it but they graciously accepted my apology.

    1. terrycojones

      Here’s a plant that pushes structured data about humidity, water etc. info Fluidinfo…

    2. fredwilson

      I remember those early days. Lots of fun

  22. Antonio Pintus

    Hi,interesting thoughts. I’d want to point you to Paraimpu, a social tool for the Web of Things we are developing with the aim to allow people to connect, inter-connect, use and share (yep!) things, sensors, objects, appliances, social networks and services in the Web to create personalized applications: For example, connecting weather data coming from remote sensors to facebook or twitter  it’s a matter of few clicks.Just for completeness, take a look to the blog: and the YouTube channel:…Cheers,Antonio

    1. fredwilson

      Thanks. I will check these out

    1. ShanaC

      I was just going to mention that…

  23. whitneymcn

    The information is certainly interesting, but I have one request to the makers of all devices that send tweets (or post data on other services) on behalf of their users: If the device sends a scheduled (daily, weekly, what-have-you) “status” notification, rather than event-driven updates, then use something like the day and time of the user’s activation to set the default time for that update — please, PLEASE don’t default to the same day/time for all users.Few things are more irritating than having a Twitter or Tumblr timeline suddenly filled with a stream of virtually identical messages from everyone’s toaster or rowing machine.

  24. Will Grant

    This is really interesting, but it’s only the tip of the iceberg I think.Devices publishing their data is one thing: sensors, weather vanes, etc – but what happens when *everything* can talk with a standardised language over a standardised protocol?IPv6 gives us enough address space for every object – I firmly beleive it’s a matter of time before the ‘Internet of things’ is just a normal part of everyday life. I’m giving a lecture about IoT in Manchester (UK) on the 21st September. Details are here:… 

  25. Devin A. Brown

    We’ve have some research on this from last year, “The Internet of Things” http://www.mckinseyquarterl

    1. fredwilson

      Thanks for the links

  26. sflite

    Check out the Tech Review article from Christopher Mims:Cape Cod is Tweeting, Thanks to the Internet of Thingshttp://www.technologyreview…

    1. fredwilson


  27. ds

    Fred,First-time, long-time.  I think you have highlighted a critical whole area for Twitter, i.e. the notion of Twitter as infrastructure to disseminate real-time information.  This, however, *can be* different from all of its social media and social networking use cases.  I wonder whether you, the Board and management may even want to contemplate (dare I say it) cleaving the company ?  Hang on, hear it out.  In essence, Twitter has two huge, enabling efforts.  One is all of the social and consumer interactions it fosters.  The other is the underlying communications platform it has created and allows for.  Allowing machines (or plants) to tweet out their sensor information or status is hugely valuable!  But not to me, who would rather just see if Seinfeld said anything funny today.  One problem, as noted by many comments thus far, is the deluge of information.  Which will only get much worse as more and more machines, bridges, and sensors tweet out bits of mostly useless information.  For a person and consumer, have Twitter exist as it does.  But for machines, sensors, bridges, it may be even better if this raw data was focused via a whole different effort in-house.  One that could focus on the appropriate protocols and messaging, as well as the right enterprises and industry groups with which to align.  For example, does anyone at Twitter (sales/product management/BD) interact with (let’s just say) GE, to help them get their fridges to tweet out appropriately, or their power gen equipment to send the appropriate 140 character formatted messages?  If not, perhaps Twitter should!  But this feels like a very different effort and perhaps-different company, doesn’t it?  How can Twitter do this type of stuff, and also focus on the user experiences that allow for political disruption in Iran or me to read Seinfeld’s quirky tweets.This is a big, complicated idea to leave in a blog comment, so we can pick it up offline if you’d like, too.  –  another vc

    1. fredwilson

      As Clay Shirky says “it is filter failure”We can now quite reliably filter spam. I think we can do the same with tweets

      1. William Mougayar

        True for programmatically, but the end-user has limited Twitter filtering manipulations today. I’m sure it will get better. 

        1. fredwilson

          If twitter doesn’t address this, it will miss its largest and best opportunity

          1. William Mougayar

            Yes & they have to rise above “the tweet” as the main unit of value. Keep adding value on top of it and around it. Eco-system is a bit gun shy unless there’s a really close relationship with Twitter.

          2. William Mougayar

            Actually, I expanded on our exchange relating to whether Content leads to people or people lead to content, or Both…or people leading to people (Twitter’s angle):…

  28. Francesca Krihely

    As many people have said, Twitter works because it is a real time information delivery system that allows people to interact with information quickly. Facebook is a little more bulky and as Christina mentioned, Facebook’s consumers prefer real people, not for robots. I think it would be fascinating to have YouTube channels that report real-time information. Programming that involves creating more advanced AI features that can process and understand data through a camera and shoot out valuable information.And then I think of Quora’s model of voting up information. Is there any way to use this structure for real time data reporting, connecting the robots and the people to filter out useless information?

  29. Chaochi Alan Chang

    Some experience with using twitter on Internet of Things. Back to 2009 when Orange Labs Boston was still alive, our group utilized twitter as M2M communication channel because its syntax is simple but good enough for our purpose. By using @ and # properly, we can get individual and group data easily. The only problem we encountered with twitter at that time is you can’t tweet to much in a short time. Therefore, I think the simple but good enough syntax is the main reason why so many “things” try to share data they collected with others via twitter.My 2 cents.

  30. Andreas Klinger

    An CEE Startup called Twingz is working on this idea since a few years. far as i know they are working on a more open standard and work closely with manufacturers.One of our Angelinvestors is involved with them. Happy to connect. 

  31. Gary Sharma

    i would love for my router to tweet me when someone is stealing my wifi

    1. fredwilson

      It would be alerting nonstop!

    2. Fernando Gutierrez

      There is a company in Spain called Fon ( which sells you a router that lets other members in the network use your wifi (and you can use theirs). The router (the call it fonera) can be configured in many ways and one of them is that it can tweet from its own account the events you choose from a set of available ones. More info here:

      1. Gary Sharma

        Thanks for the info. That sounds interesting. And I just realized I actually have an old fonera router lying around! Am going to try hook it up to my twitter and play with it a bit

  32. I agree with you completely, but it’s “determinant” and “extensibility.”

    1. fredwilson

      Ugh. My 6th grade English teacher is cringing somewhere. Thanks for the corrections. I will fix when I get to a computer

  33. asjs

    Botnets use IRC in much the same way, to post command and control messages to/from the slave nodes in the botnet.

  34. hypermark

    I’d put this in the bucket of “everything old is new again.” Back in the late 90s, concurrent with the rise of web-manageable devices, it was fairly popular to embed an SMTP server into a device like a printer (‘I am out of paper’) or a router (‘hacker attack’) specifically so you could push thresholded events to an IS administrator.The logical end-game on this stuff, as others have noted, is that that message has enough structure so that you can actually “do something” with that information:1. Green Event: fix the problem, let me know that it was handled2. Yellow Event: diagnose the problem, tell me my options3. Red Event: package up the data, and give me some potential root causes from which to debugEmail, SMS, RSS, tweets; it is fundamentally a message (sometime with a payload, sometime not). The real question is how do you filter that message so signal is separated from noise, and how do you federate aggregate messages so that “more than the sum of the parts” types of events get logically organized (there’s major rain storm in Des Moines, these are the clusters of conversational threads around that event)?

  35. Guest

    I would like to see babies auto-tweet when they need a diaper change.

  36. Graham Siener

    When I was faced with the task of building a data monitoring system for a sustainable school campus, I wanted a platform to share this data.  Twitter seemed like a potential fit but felt way too forced.Then I stumbled onto — they refer to themselves as a platform for the internet of things, which is a fantastic concept.  I sent all of the energy, water, weather, etc. data out to Pachube in what felt like a command line pipe.  Really clever people have gone a step further and consume those various feeds to make things happen.For example, here’s a feed of the Radiation dose in Minami Soma, Fukushima:…You can consume that via xml, rss, even json.  Paired with the power of hardware like Arduino, we’re getting into interesting territory! 

    1. Charlie Crystle

      I was just gonna say, “arduino”. Fred you should clear you schedule for Maker Faire. Arduino + Twitter + makers = America dreaming. 

      1. fredwilson

        I will try

  37. aguynamedben

    “I suspect that the symmetric friending model and the use of real names/real people in the Facebook system is a hindrance to devices updating Facebook pages, but I could be wrong.”It’s a little more low-brow, but I have found that getting a Tumblr account to tweet upon a new post is much easier with the Twitter API than it is with the Facebook API.  Getting simple data to Tweet took about 10 minutes.  Getting that same data to post in Facebook was non-trivial and I gave up. This insight is un-exciting, but for DIY hardware hackers that are throwing up weather sensors in their windowsills and tracking their fridge with XBee’s, ease of use in getting the data to another source is a major factor, and Twitter wins in that category, especially when the data is small, frequent, and innately “status”.I think the asymmetry of Twitter does have an impact on hackers as well though because it allows them to publish their hackery to the world at large. Many hackers want to show off their work by airing things publicly. The proof is in the public broadcast, not the broadcast to your friends who are unexcited by the fact you’re a nerd.

  38. ShanaC

    You mean like what is going on at the Moma:eg: http://newyork.grubstreet.c…The oven tweets when new things come out of the oven… (yum)…see this twitter feed:!/momab

  39. LukeG

    Is your friend going to share that data?

    1. fredwilson

      Very soon. Actually his client will

  40. Charlie Crystle

    My toilet tweets every time I take a pee. Gives new meaning to the term ‘tweet stream”

  41. Tim Thomas

    I have a ticketing site – – and I’ve attempted Facebook integration several times. Just to basically say “Hey, I’m going to see this show. Come with me.” Every time has been pain and suffering. Facebook’s API is painful to use, and frequently if you come up with an innovative way to use it you find that they have blocked you.I honestly think it’s the tech getting in the way. Google+ should definitely take this to heart and try to copy twitter’s model when they roll out their API. Doing a really heavy gData style model could potentially sink Google’s Ship.

  42. Gonzalo Garcia-Perate

    Twitter is widely used by machines to send data for two reasons:- Is an API that does not feel like an API, so people can just get on with building things. – Is a protocol that does not feel like it. Just use (# and @) and you’re rolling.For early examples of application see IBM’s scientist “Andy” who wired up windows, doors and energy feed of his house to twitter. He then used the feed to monitor whether energy consumption was shooting up and a window had been left open etc.. there are lots of examples like it.The business implications of machines tweeting (talking) are huge – regardless of whether it happens via twitter or not.At a macro level you have people like IBM with their Smarter Planet initiative… trying to build value out of data streams (potentially) coming out of cities to create new types of intelligent services for cities. GE, Philips, Cisco etc all have similar initiatives. Also see funding coming out of governments into academia and startups under the label IoT.At a smaller scale you have lots of new companies innovating and creating new product and service propositions based on the idea of objects communicating. Some over twitter others over their own proprietary and open networks and protocols. In this space there are two main types of companies, middleware and appliances.In the middleware camp you have people like the recently acquired Pachube and many others like IOBridge and Thingspeak,, Twingz, (not sure what is this) Open sense, Simplio,… ThingWorx, etc etc etcAt the appliance level there are many small and not so small companies that are exploring the IoT space. See Nike +… Vitality Glowcaps Wifi scale and body monitors by Withings fitbit, greengoose and many energy monitors.Are objects tweeting part of the social media landscape? Definitely. The places we live and work are as much part of our (offline) social context as our blogs, twitter, facebook, google+ pages are part of our (online) social context. The merging of the two is inevitable. For instance, the same way a large number of comments in a post like this is indicative of the size, interests, and popularity of the community that surrounds your blog – your online social context – the amount of activity that takes place in the shops, restaurants, transport systems around where you live and work are indicative of the activity and interests/preferences in your offline social context. Having data on what takes place offline can help you decide where to go, how, and what to have/buy when you get there. And that will be accessed via the cloud like everything else, and it will be part of your “Profile”.For instance, at the moment you may get an approximate of those signals via explicit foursquare (or whatever) checkins (you know a party or a good restaurant is nearby). Soon that explicit action of checking in, will not be necessary as the place you’ll go will tweet for you. And then someone will be aggregating that data and generating value out of it, I’m working on that exact problem.

    1. Rick Bullotta

      FYI, ThingWorx is considerably more than “middleware” and substantially more than “data”.  I’d be delighted to show you sometime.  Contact me offline.

    2. fredwilson

      Lots of great stuff in this comment

    3. tamberg

      You might want to add to the middleware camp. Yaler enables direct Web access to embedded systems behind a firewall/NAT. (Disclosure: I’m one of the founders)

    4. Alberto Brizio

      Since you mentioned IBM’s SmarterPlanet initiative, one of the finalists at their recent NY event was EnvEve. They have tons of expertise in collecting, aggregating and processing remote sensor data for environmental applications. Marco Brini, the CEO, is a brilliant fellow and worth connecting with if you are interested in the space.

  43. Mike Mahoney

    It seems that most of this conversation is focused on very simple ‘things’, such as sensors (thermostats, barometers, etc).  I think the IoT becomes much more interesting when you consider higher order things like buildings, manufacturing equipment, power generation facilities, etc.For basic sensors Twitter might make sense, but for more complicated equipment the idea of interaction becomes more valuable.  In that case a more ‘social’ platform like ThingWorx makes more sense.

  44. Rick Bullotta

    Hi, Fred.Nice to see the “internet of things” on your radar.  Back when we met a few months ago, I had a feeling it might resurface. We’ve built these kind of capabilities directly into our ThingWorx platform – whether its integrating the physical world with people, with social media (Twitter, FB, etc.), with other web APIs (Twilio, etc.), or other applications, we’ve got it covered.I agree with one of the commenters who suggested that twitter is not necessarily the right destination for certain types of machine/real world events.  The ability to capture more context, with more structure, in a way that people and other machines can easily consume the data streams, is very important, and what we’ve built into our stream engine in ThingWorx.Many are tackling the internet of things as a “data” problem, and it isn’t.  It is so much more than that.  The data, the events, analytics, visualizations, and the resultant useful *applications* that can be enabled are what matter, and we’ve created a platform + APIs + developer tools + end user apps to make it possible and cut the time to create apps by 10-100X.The data/events/services that are exposed are definitely of value for human collaboration, but they are also extremely valuable for machine/system collaboration as well. The ability to directly control and affect other “systems” based on real world events is of massive value and probably has more immediate and measurable (read: convertible to revenues) value than on the consumer side, but we’re playing both sides of the fence anyway.We now have active customers and partners in smart grid, healthcare, building/energy management, mining/resources, manufacturing (chemical, food/beverage, pharma), and many other sectors.Let’s catch up again sometime and I can give you an update.Best,Rick

    1. fredwilson

      Hi rick. Glad to hear you guys are doing so well

    2. Justine Lam

      Rick,I’m at Singularity University this summer working on a project to secure the supply chain. Can I talk to you about your platform?Justine [email protected]

      1. Rick Bullotta

        Absolutely, Justine.  In fact, we use the word “Thingularity” often… 😉

  45. Vasudev Ram

    Tim O’Reilly seems to be into the Internet of Things these days, as I saw in one of his profiles, maybe his Google Plus profile, recently. Following him on G+ and/or Twitter may be useful for those interested in the subject, since he generally tweets – a lot (@timoreilly) – and also blogs (on O’Reilly Radar – about stuff he’s interested in.

  46. Douglas Crets

    Another way to say it is that Facebook is an impediment to Facebook. 

  47. Kunashe Tarusenga

    People are like sensors, particularly on Twitter.A sensor is a device that is sensitive to a particular condition & responds with a calibrated reaction. When something interesting happens, tweet numbers peak (think resonance). You need to rock the boat more on fb to get a reading!

  48. sigmaalgebra

    This thread is heavily dipping its toe in a large, old body of work.Twitter may have an advantage in that it has many senders and receivers. Generally receivers join because so many senders are there; senders join because so many receivers are there; and the ‘middle man’ Twitter gets a ‘natural monopoly’.But if Twitter actually went very far with data from/to machines, then quickly there would have to be some big changes in what Twitter sends and receives, how it works with ‘followers’, etc.It’s an old story back to the Unix world of ‘simple network management protocol’ (SNMP) and associated work in ‘managed systems’, ‘managing systems’, ‘abstract syntax notation 1’ (ASN.1), ‘common management information service’ (CMIS), etc., e.g., as at:…with:”The common management information protocol (CMIP) is a protocol for network management. It provides an implementation for the services defined by CMIS, allowing communication between network management applications and management agents. CMIS/CMIP emerged out of the ISO/OSI Network management model and is defined by the ITU-T X.700 series of recommendations, its more popular correspondent designed by the IETF being SNMP.”So, over there on the right can have lots of ‘systems’ to be ‘managed’. Over here on the left can have lots of ‘systems’ doing the ‘managing’.The managed systems on the right can report data of various kinds, and the managing systems on the left can request data and issue ‘management’ commands.Okay, then that’s a nice generalization of what Twitter has now. That originally the ISO/OSI CMIS/CMIP and SNMP work had to do with managing computer communications networks is next to irrelevant.For the ISO/OSI CMIS/CMIP work, a LOT has been done. If Twitter wants to grow to do something important with data from/to machines, moving toward the old ISO/OSI CMIS/CMIP work will be nearly inevitable.An early question will be, what the heck is the format of the data that is sent? Since might have 100 million different machines sending data, maybe we have to pay a little attention to the ‘format’ and have more than one?Okay, people have been there, thought about that: One solution is for each of the, say, 100 million senders to publish their format — call it their data ‘schema’ which, say, describes the data ‘types’ — numbers, character strings, arrays, lists, lists of key-value pairs, collections of such things and collections, etc., to be sent. So, such a schema describes a tree.Then would have 100 million such? Then each programmer would want to be able to parse and work with the meaning of 100 million different schemata? Tilt!So, arrange the schema in an ‘inheritance hierarchy’ with a central ‘schema registration authority’. Then have a schema for, say, a car as a ‘managed system’ registered with the registration authority. GM inherits from this schema and has their own relatively minor additions and/or changes, so minor that software that works for data from cars will be close to working for GM cars. Then the GM Chevy division inherits from the GM schema. Then the Chevy Corvette for 2012 inherits from the base Corvette schema which inherits from the Chevy schema.Such a schema might be called a data ‘type’ or data ‘class’; an instance of data in this class might be called a data ‘object’; and we might say we have ‘object-oriented’ system management.Note: All this data is just ‘passive’; it’s just data, not software! While the point is simple enough, some people can get really confused about such things! What else do they get confused about?Given the tree of one schema, there is a simple, standard way to do a ‘depth-first traversal’ of the tree and, thus, write out the data ‘sequentially’ ready for transmission over a communications channel. The old ISO/OSI CMIS/CMIP way to send the data was ASN.1, and it is simple and obvious.To receive the data, parse the ASN.1 data and rebuild the tree, easy enough with current software. Really just need software classes for the elementary data types, say, numbers, character strings, times, dates, bit strings, etc., and then various aggregates — arrays, lists, lists of key-value pairs, where the elements of these aggregates can also be aggregates. An HTML document object model or the ASP.NET instances for a Web page is similar.Okay, that is a start, but soon enough too many managing systems on the left are being too active with the managed systems on the right. There are two problems:(1) Bandwidth. Can be sending too much data, commonly much more than necessary.(2) Conflicts. Separate managing systems can get into conflicts over their control of a common set of managed systems. So, managing system A can be turning the lights on at managed system X while managing system B is turning them off. Bummer.So, there needs to be some security, some sense of ‘ownership’, e.g., authentication, capabilities, access control lists, encryption (right: Kerberos) and, then, ‘transactions’, with, say, a monotone locking protocol, and deadlock detection and resolution.So, in the middle between the left and the right have a ‘resource object data manager’ (RODM) that handles all that. Where to get one of these? Gee, wondered if you would ask! There is at least one, although I don’t really know just what happened to it since the last time I saw it!Yes, somewhere in here will want some software, say, as an ‘agent’ in a managed system or inside a RODM.Next, one of our roles will be to monitor a managed system. Okay. Suppose we have data from the past three months and convince ourselves that during that time the system was ‘healthy’. So, anything too different might not be ‘healthy’ and, thus, might be ‘sick’.So, continually we want to do a statistical hypothesis test. But we have (1) data on each of several variables and (2) next to no way to know the probability distribution of such multi-variate data even when the managed system is healthy and much less when it is sick. So, we need a test that makes no assumptions about probability distributions; we need a ‘distribution-free’ test.So, we go to the books on statistical hypothesis testing looking for multi-variate, distribution-free hypothesis tests and see none. Bummer.But, then, maybe if we get a finite algebraic group of measure preserving transformations as in ergodic theory and sum over the group, then we can calculate and adjust the probability of Type I error in our hypothesis test? Then we might use an old result of S. Ulam (as in the ‘Teller-Ulam configuration’, a different subject!) to argue that our test is not ‘trivial’! If we strain and work asymptotically, we might convince ourselves that we can do nearly as well as the classic, best possible Neyman-Pearson result!Gee, I thought you’d never ask!So, that’s a start on the ‘data architecture’ for the future of Twitter in the world of data from/to machines.For the ‘social’ and ‘human’ part, that would make use of the ‘data architecture’.Other uses? Bloomberg-like feeds of data on the economy driving program trading, Homeland Security, motion sensors at home while at work, traffic on the Tappan Zee bridge, etc.

  49. Jeremy Justice

    As soon as I read this, I could help but think of @BeijingAir, which many of us expats in China follow. It’s gives an hourly report detailing pollution and ozone levels. Sadly, with the constant reminders of high pollution, it also forces us to ask ourselves, “Why are we here again?!?” 

  50. e.p.c.

    Andy Stanford-Clark (@andysc:twitter ) of IBM has been doing a lot of work to use twitter with various devices and things (see the links about his twittering house or the Isle of Wight ferries off’s API made it trivially easy to set things up like this. The switch to oAuth has had a definite impact which ruled out tweeting from low-power simple devices (i’m not arguing against oAuth…just that the computational requirements imposed made it much more costly for some of the initial tweeting devices).  Couple the API with twitter’s low-bar for creating accounts (honestly for all I know twitter requires the account to be a real person or tied to someone but if so they don’t enforce it, no “real names” policy nor “humans only” restrictions) and you’ve got an easy, ready made platform fpr “internet of things” apps.



  52. andyidsinga

    internet of things …that tweet and sh*t 😉

  53. Tereza

    My site corroborates this ( @honestlynowinc)Although 70% female, on the first warm day in March, we had a run on questions from guys asking the ladies if they should shave their beards. Interesting!After the Casey Anthony verdict, tons of people asking about the justice system.I brought in an expert on Balance this week. Hit a nerve.Yes –a real-time stream is the perfect way to gauge what people are thinking and asking, and may be doing soon.

  54. ShanaC

    probably because i’m so relaxed from being about to crash:I was rereading and it occurs to me the reason we want an internet of things is because when items have “a voice”, it is a lot easier to interface with them.  We want low volume notices that just tells us what we need is a short space (hence the appeal of twitter and and the notification capture pulldown box on android) – we’re just having a very hard time adapting the same concept from a UX and computer architecture pov for machines that we interact with quite differently – like your washer and dryer (which now have computers in them to control how you do with laundry – but users don’t see, to figure out what these settings are).I guess what I am trying to say is that the internet of things helps us understand how to use our things better long term…(off to bed I go)

    1. Guest

      I get where you are coming from Shana and agree that a better interface may result (something that allows we humanoids a way to better utilize) but I still do not think social is the appropriate terminology and social interaction is not what is going on. Tweeting is not conversing, though it might create a conversation – eventually or possibly.

  55. Donna Brewington White

    This post proves that you will never run out of topics.  

    1. Fernando Gutierrez

      Yeah, and if he does he only has to do a post about being out of topics like the one he did a few months ago and we’ll all go crazy again 🙂

      1. Donna Brewington White

        oh yeah…forgot about that time

  56. Donna Brewington White

    off topic:BTW, did you see this post from a Germany-based VC in response to your post on venture-backed start-ups reaching exits greater than $100 millionEarlybird Partner: Why Returns In Europe Are Now Outpacing The U.S.

    1. fredwilson

      Had not seen it. Thanks!

  57. Jeff Pundyk

    Fred, check out the Talk to Me show at MOMA.… review here:…best,Jeff

  58. Luis Argandona

    Hi, Very nice post. Here is an interesting case: a young kid here in Chile invented a way for twitter to warns us of earth tremors and earthquake moments before they happened. There you go: things twitting and taking care of us.…have a nice day, Luis Argandona

  59. My Showcase

    Nodeable may be of interest, “a Twitter for machines… that generates activity streams for important system events [and] builds in tools for managing common processes.”  They raised $2 million in funding from True Ventures.

  60. Andy Piper

    Hi Fred. As e.p.c. mentions, my colleague Dr Andy Stanford-Clark from IBM has been talking about “tweetjects” (google that) – objects that tweet – for some time, and the MQ Telemetry Transport or MQTT is also a great protocol for enabling this kind of interaction from “things” – anything with a small amount of processing power and the ability to transmit or receive data, even over unreliable networks. Take a look at Andy’s website, the various pieces of research IBM and IBMers have been doing in this space (including his Twittering house, the instrumented Isle of Wight Ferries system… etc), and It’s a royalty-free protocol which can be easily bridged to others. Look out for much more activity in this space in the very very near future.

  61. #Pablo On Tech

    Fred, data is going mainstream, 

  62. #Pablo On Tech

    Fred, data is going mainstream; you’ve posted some great spreadsheets yourself over the year.  I’d love to connect with you to have your data get a 10x in social visibility.

  63. jodyreale

    Can you imagine what would happen if commercial jets started tweeting?

  64. Melinda Byerley

    Hi Fred,  We here at PlantSense ( have an internet connected device that measures all sorts of information about your indoor and outdoor garden and harnesses the power of the cloud to analyze that data.  We are actively moving into the social media space and thinking about what aspects of gardening are most social and how to use them effectively, especially as we move into wireless devices. For example:  customers could see the top plants thriving in their immediate area, or have competitions with neighbors or friends across the country for the healthiest garden with prizes virtual and real.    We see a future of all kinds of interaction via social media–the device is simply a means to the end.  Of course we need the social services to be open to this–but the power of network is quite strong so we appreciate your call to action for them to support the internet of things!   

  65. Werner Weihs-Sedivy

    Hi Fred!We of – the twitter for things believe in the power of things becoming communicatable, customizable and controlable via short texting. Our plfm is primarily integrated with twitter, but other social networks are already connected as well. Our platform (currently in closed alpha, existing prototypes – pls check… for instance) is currently in several PoCs in the areas of B2C Communities (special interest groups), energy management, small appliances and vending machines.We’d be glad to provide further info if you like us, WernerPS.: We have not come across Facebook thing or weather vanes, too 🙂

  66. Meredith Finkelstein/Chang

    This is such a deceptively simple idea, but powerful idea.  I suppose once you hook up your sensor to the net most people hope that another party will do something useful with the data. I have a scale that (optionally) tweets my weight.  I’m not so sure I want to blast that into the twitterverse- but I could definitely imagine being part of a weight watchers group on fb – and hooking into that (and then hooking my wifi food scale into that too).I guess the question is – how do we connect people with the ‘internet of things’ (social or otherwise)