Contextual Runtimes

Benedict Evans is such a great analyst and his insight into the web>mobile transition has been consistently prescient and helpful to investors, including USV and me personally.

A couple days ago, he penned “16 mobile thesis” which is a must read for anyone building a mobile/internet company or investing in that sector. These 16 theses are organized roughly chronologically, starting with what has largely happened, followed by what is happening, and ending with what may happen.

I found myself most interested in the middle section, 7-9, in which Ben talks about where the action is turning to in the mobile ecosystem. And my favorite part is titled “Post Netscape, post PageRank, looking for the next run-time.” In this part Ben describes what used to be the dominant environment and the search for what is next. At the end he states:

Really, we’re looking for a new run-time – a new way, after the web and native apps, to build services. That might be Siri or Now or messaging or maps or notifications or something else again. But the underlying aim is to construct a new search and discovery model – a new way, different to the web or app stores, to get users.  

I agree with Ben but I think there won’t be one runtime in the mobile era. I think what is emerging is multiple runtimes depending on the context – “contextual runtimes.”

If I’m building a lunchtime meal delivery service for tech startups, that’s a Slack bot.

If I’m building a ridesharing service, that’s going to run in Google Maps and Apple Maps.

If I’m building a “how do I look” fashion advisor service, that’s going to run in Siri or Google Now.

If I’m building an “NBA dashboard app”, that is mostly going to run on the mobile notifications rails.

So the war for users in mobile and the race to be a platform is real and it is important. And Apple and Google are playing that game as well (the notification, map and voice runtimes are controlled by them already).

But it isn’t clear that all of these contextual runtime environments will be controlled by and or subsumed into the mobile OS. That’s what makes chat so interesting. Slack has emerged as the dominant chat app in the enterprise, but not the only one. Facebook Messenger has emerged as the dominant cross platform chat app in the US, but not the only one (our portfolio company Kik continues to grow and is already massive). Whatsapp and Telegram are very popular outside of the US.

In content, there is an entirely different set of “runtime environments.” Facebook and YouTube are huge content discovery and consumption environments. Twitter and Snapchat are trying like hell to join them. So are many other mobile social platforms. Content is a bit like chat. I don’t see this sector converging quickly into the mobile OS platforms.

So the thing that is a bit different in the mobile era of the Internet, as opposed to the desktop era, is not everything is built on top of a browser. The phase we are in now, phase one I guess, has two dominant “runtimes”, mobile web and native app.

But we are heading into a new era in which a few native apps, chat, maps, voice input, notifications, content/social, and surely a few more, will become the new browsers. And entrepreneurs will be building contextual services on top of them.

In this era context will be critical. The example I keep coming back to is the list of places I need to go to when I’m in a new place. Like this Foursquare list of top places where we are this week:

top places

Seeing that list on a map versus on a list makes a huge difference.

That is the power of context and that’s where I think the next big moves are to be made on the mobile internet.


Comments (Archived):

  1. LIAD

    All posts are created equal but some are more equal than others.

    1. fredwilson

      I generally prefer the ones I write looking at views like this

      1. William Mougayar

        Watching the sea or the mountains has something magical about it. There’s a famous french song by Alain Barriere, A regarder la mer. Looking at the Sea.”Whoever can’t understand me can’t understand the seaIn this great universe I will have beenAnd I spend hoursAnd I spend hoursLooking at the sea”

        1. fredwilson

          i kind of wanted to go to the mountains because they are getting dumped on right now. but the gotham gal wanted beach and she wins most, if not all, of these debates

          1. LE

            Doesn’t sound like much of a debate.

          2. awaldstein

            Just booked four days skiing with my son in Aspen in three weeks. Seriously excited about this.

          3. JLM

            .Both.Merry Christmas. Sunscreen.JLMwww.themusingsofthebigredca…

          4. LIAD

            Debates. Lol.

      2. Simone

        posting such images is dangerous (I am from the seaside), makes me want to emigrate to Australia tomorrow and just forget about anything else.

        1. sigmaalgebra

          At http://www.businessinsider….are about 10 just astounding views of old growth forests, some from Australia.Ah, in this thread, did we mention discovery?

          1. Simone

            for me, it has to be the sea. in a beautiful forest I feel like an intruder, it is with the sea that I feel connected

      3. sigmaalgebra

        Yes, there is something strangely attractive about nature scenes. So, get a nice place in the country with some good scenery. My backyard has had foxes, skunks, rabbits, squirrels, chipmunks, grass snakes, falcons, flocks of crows, geese, flocks of turkeys, deer, racoons, ground hogs, stray cats, huge varieties of various birds, and more, but each year the population thins out some — I don’t know why. Somewhere out there is an owl — I’ve heard him a lot but never seen him. Up the road a mile or so is a flock of guinea hens and a herd of cattle, and there is another herd of cattle a little closer. The wild life is fun. Need to up the population of the wild life.About Christmas, there isValery Gergiev, Mariinsky Theatre, Piotr Tchaikovsky, The Nutcracker…and it is not the usual! It’s way too good for that. There was a really serious, special effort behind that performance. A great Christmas treat.

  2. awaldstein

    Damn this is something to think about.I’m a big fan of Benedict’s work because unlike most any other analyst out there, he adds huge value without pretending to know the ultimate answer and–for me a big one–makes short lists to think within.

  3. jason wright

    this should have been our homework assignment, reading Ben’s 16 theses, and then your post, and then we’re all primed for this discussion. i like homework.i have stereoscopic vision and two feet. the web must map its information to my reality. i am the web.

  4. William Mougayar

    So, if you had written this, it would be titled “16 Contextual Runtimes.”I like where mobile is going, as the smartphone becomes the entry and notification point for any interaction. But this creates 2 challenges:1/ More bi-directional knowledge needs to be constantly shared between the user & the “data processor” or “service provider”.2/ Multiple contextual interaction points might complicate/fragment the user experience, and we’ll need to adjust to it.

  5. William Mougayar

    There’s a great article, Smart Messaging as Platform Shift, related to this, by Sarah Guo from Greylock. GUI isn’t going to be the only UI.

    1. jason wright

      “In 2014, college students (at least in this interesting study at Baylor) spent ~9 hours a day on their smartphones.”9, a day!

      1. William Mougayar

        wow…i wonder how much of it is on Facebook, Snapchat, Messaging. prob 90%.

        1. awaldstein

          For all the issues with texting I find that I use it more and more.Never would have forecasted this.

          1. William Mougayar

            I do too, but it’s fragmented across Skype, native, Hangout, WhatsApp, Kik, Tango, Twitter IM and Slack. Too many entry points!

          2. Anne Libby

            Online games, too.

          3. William Mougayar

            OK. I’m not a gamer.

          4. leapy

            Agreed. My children all link up with their friends on Minecraft – crafting together while texting about homework and “stuff”.

          5. laurie kalmanson

            the teachers were annoyed when the kids figured out that they could leave comments on google docs as a form of chat

          6. awaldstein

            YupVery little WhatsApp, no Kik or Tango in my world, more and more Slack,

          7. William Mougayar

            I meant Viber. Tango no more.

          8. Matt Zagaja

            I remember using AIM and GChat a ton back in the day. GChat ate AIM and then got eaten by iMessage at least for my friends in the Apple ecosystem. It’s too convenient to have everything sync’d on your computer and phone. Slack is more for organizational/professional stuff but at work Jabber still rules the day. I have a couple people that prefer Facebook Messenger or GChat still.

          9. William Mougayar

            It’s driven by the recipients who are communicating with me and their choices. If it was up to me, I would stick with one platform. Add Facebook Messenger to that list.

          10. LE

            I tell people “text for short stuff” “email for longer things” and “don’t leave me voicemails I can’t easily reply to them”. [1]For calls I typically say “approx time say ‘afternoon'” then I will text you first and if you are around I will call you”. I hate set meeting times approx. works better. [2][1] Now if there was an easy way to reply to a voicemail with a voicemail (and not have a phone answered) that would be good but there would have to be effective text to speech so I could memorialize all conversations which I typically need to do.[2] I am waiting for an app that allows two parties to say they are available and it’s a good time to talk. Would need granularity as to the parties (mom different then business and different then a friend).

          11. leapy

            Re your footnote [1] – My wife and her relatives and friends all use Viber and regularly embed short voice messages inline with text messages. Works very well.

      2. JLM

        .Well, at least they weren’t dancing, no?Inside joke. Sorry.JLMwww.themusingsofthebigredca…

        1. LE

          Maybe that is what she was doing last night that made her late to the podium.

          1. JLM

            .In the spirit of the season, I will not touch that.Happy Holidays from Steamboat Springs where we are getting ready to embrace a two foot dump of champagne powder.JLMwww.themusingsofthebigredca…

          2. LE

            I admire the fact that you are a man that will only do a job if it’s well done. So from your reply I figure either or both a or b is true.a) You did not watch the debate last night (you were out to dinner)b) You want to get out on the slopes.Have a good time, we are definitely jealous. Post pictures later if you can so we can be even more jealous.

          3. JLM

            .You possess incredible powers of deduction. I was at Christmas caroling and dinner.http://themusingsofthebigre…Happy Holidays and let’s have no broken limbs.JLMwww.themusingsofthebigredca…

          4. sigmaalgebra

            Of course, no one will know because the diva kept everyone else out of the little girl’s room.

      3. sigmaalgebra

        Sounds like their professors needed to assign more homework, a LOT more, give harder tests, etc.!

    2. Richard

      Good Read. Who are your 10 go to blogs (non-techcelebrity) each week?

      1. William Mougayar

        It’s hit and miss. I use my Feedly feeds & email newsletters that bring me a ton of stuff eg Mattermark’s, plus Twitter of course. I’m not religious about any particular blog, except AVC, and a 3rd newer one that I won’t divulge yet 😉

    3. sigmaalgebra

      Seems to me that the HTML GUI is still pretty good. Even if virtual gets traction, I’m reluctant to believe it will replace the HTML GUI. For one, the HTML GUI is a convenient half-way meeting ground for both the writer and the reader, easy enough for the writer and good enough for the reader. The user-computer interactions that can be done with the HTML GUI can be a lot of what we might want; that GUI is sufficient for a quite rich U/UX, rich enough that the content is more of the bottleneck than the GUI. My guess is that we need to concentrate on the content, information wine and not the HTML GUI bottle.

      1. William Mougayar

        It’s much harder to mess up a text UI 😉

  6. pointsnfigures

    multiple run times. The unbundling of a service so it becomes more on demand to individuals rather than one to many.

  7. Twain Twain

    Analysts see tech through a different lens from the developers, inventors & R+D folks making the tech.The key macro-requirement for the “next big thing” is coherency. That means coherency in programming languages & IDEs. Coherency across different browsers. Coherency across different media (PC, mobile, smart TV, IoT hardware etc.). Coherency of UX across those media.Coherency in terms of the machines understanding us and our inputs & information and being trained for better intelligence according to those inputs.The runtime would be most efficient if it was a neuro-chip embedded directly inside our brains and physiology (yes, that’s why Google R+D is looking into this under Ray Kurzweil and Regina Dugan).

    1. LE

      The runtime would be most efficient if it was a neuro-chip embedded directly inside our brains and physiology (yes, that’s why Google R+D is looking into this under Ray Kurzweil and Regina Dugan).You know that SNL skit “stop it you’re killing me!”. This sounds exactly like the type of idiocy emanating from the lifestyles of the idle rich and full of themselves.All of this is a big waste of time and money it’s a hobby not a serious pursuit. I’ll take the bet that nothing of the sort will happen in our lifetime.Anyway, Kurzweil is out to pasture on this one. Kurzweil in one paragraph:Bill Gates calls him “the best person I know at predicting the future of artificial intelligence”. He’s received 19 honorary doctorates, and he’s been widely recognised as a genius. But he’s the sort of genius, it turns out, who’s not very good at boiling a kettle. He offers me a cup of coffee and when I accept he heads into the kitchen to make it, filling a kettle with water, putting a teaspoon of instant coffee into a cup, and then moments later, pouring the unboiled water on top of it. He stirs the undissolving lumps and I wonder whether to say anything but instead let him add almond milk – not eating dairy is just one of his multiple dietary rules – and politely say thank you as he hands it to me. It is, by quite some way, the worst cup of coffee I have ever tasted.More about him:…He looks older than his 65 years, I guess this shit isn’t working for him that well:It’s hard to know where to start with Ray Kurzweil. With the fact that he takes 150 pills a day and is intravenously injected on a weekly basis with a dizzying list of vitamins, dietary supplements, and substances that sound about as scientifically effective as face cream: coenzyme Q10, phosphatidycholine, glutathione?

        1. LE

          Dugan added that the chip had already been FDA approved and could be taken 30 times a day for someone’s entire life without affecting their health.Someone would have to be 10 years old to believe anyone that tells them “without affecting their health”. I will wait for the long term studies on this. Maybe they can experiment on prisoners in jail on drug offenses as a start. 30 years later maybe they might have something.

      1. Twain Twain

        Kurzweil and neuron chips that can be connected to and manipulated by the Cloud that will make us more “god-like”:* http://www.huffingtonpost.c

        1. LE

          What’s best about “this shit” is the oh so practical examples they give on why these “dreams” have a place in our world (as opposed to maybe using their high intelligence to solve the spam problem or the poor people who commit crimes and take drugs and are losers).For example this:“So I’m walking along, and I see Larry Page coming, and I think, ‘I better think of something clever to say.’ But my 300 million modules in my neocortex isn’t going to cut it. I need a billion in two seconds. I’ll be able to access that in the cloud — just like I can multiply intelligence with my smartphone thousands fold today FOR SURE. Humor is that important!!! And if every guy can tell a joke and get the girl then the girl will no longer be impressed because all guys will be created equally as funny! [1] Maybe Larry Page has announced that it’s necessary to win his favor with a joke or clever remark. With humor survival of the fittest. Generally fat people have an advantage with this perhaps because they weren’t social when young so they had plenty of time to develop a humorous side to overcome social anxiety or something like that..[1] My joke that got a laugh this AM at Panera Bread while I was waiting in line. Guy tells his wife “I felt really good yelling at Jim last night at the party. Then 30 seconds later I felt bad!”. Me (overhearing this I couldn’t resist) “well you should have kept yelling at him then!”. (Instant genuine laughter from everyone who heard, and all from 300 million modules in my neocortex..).

    2. SubstrateUndertow

      I’m all on board with the “coherency” mantra but that can be framed at many levels and from multiple perspectives.First thing first !Before we start looking for wetware level embeds maybe we should try experimenting to find optimal “coherency” tactics via more easily retrofitted reiterations at the software/social (App) level. After that shakes out the stable reusables we could start thinking about embedding proven strategies that extend/amplify “coherency” at the wetware embed level ?It seems to me that “Messaging” more generically visualized as universal/generic “networked synchronization-signalling” of everything at every level (from quantum on up) is the core coherency-organizing-dynamic that informs reusable coherency-strategies all the way up through software-driven and social-driven systems.It is my (by now tiresomely repeated contention) that the top-down narratives, metaphors and lexicons exemplified by the organic living-systems-dynamic is the most fruitfully available/tangible gateway for homing in on a reusable bottom-up coherency-framing strategy that is capable of riding in on the coattails of our present App-driven social-meme landscape/organism.The core generically reusable metaphor here is the “neural-node/net” applied as nested constellations of mutually-adaptive purpose-driven Apps/social-processes.One could visualize our evolving global social-organism as made up of App constellations akin to organ-function analogues. Whatever reusable messaging evolves within App-function-constellations and between App-function-constellation at all levels of that social organism it must be universally consistent for optimal technical remix and widely accessible to mass-culture perceptual-ergonomics for social stability.What generic metaphoric-imaging palettesof “neural-node input function” categories,of “neural-node processing function” categoriesof “neural-node output function” categoriescan be collaboratively teased out to satisfy optimal RISK-based technical remix potentials while maintaining a maximum metaphoric overlay with widely accessible mass-culture perceptual-ergonomics ? ? ? ?

  8. Manolo Espinosa

    Getting complicated/fun as platforms try to control/subsume all these sectors – improve content positioning (Twitter moments, FB video, Snapchat discovery/stories) and/or morph into utility environment (FB Messenger), while staying true to core DNA/advantages (Snapchat kids sharing their real lives).Lots of opps, still early.

  9. Aviah Laor

    It reminds the famous story of how the first printed books were just copies of handwritten books. Took a few decades between the invention of print (the new runtime), and the first printed cookbook. The question is to what extent the runtimes will be open.

    1. fredwilson

      that’s why it is so important that there are at least two of everything (OS, chat, map, etc). then the one that is open would get richer and the one that is closed would not.

      1. Salt Shaker

        2 or even 3 of everything makes sense. Keeps things somewhat in check. Problems arise when there are 4, 5 or more fundamentally doing the same thing (e.g., music, ride share, home share, travel). Sectors and rev streams just not large enough to sustain all, and for some parity evolves into parody.

      2. LE

        then the one that is open would get richerNormals don’t care about that shit period. [1] They want something that works and looks nice and provides value. Open means nothing to would get richerExcept for Apple and Microsoft which did pretty well with closed.[1] In other forms “voters want the snow cleared after a storm”.

        1. Aviah Laor

          I believe that people are increasingly more aware to the long term price of this “easy now don’t care” attitude. It’s very similar to giving complete control of your career and financial future for a seemingly comfortable job, just to find out you are fired and it’s gone. BTW Microsoft (and Apple on Macs) was not closed compared to current mobile, you could always ship an exe file, send to it to whoever wants it, and double click – no approvals or guidelines or anything.

      3. Twain Twain

        What’s likelihood of “open” leading to the monopoly position Peter Thiel argues is necessary to create the next Google, FB, Apple etc.?Meanwhile, David Galbraith, partner at Anthemis VC, notes: “Europe’s sort of semi-open market has created the scenario where it doesn’t have any of its own platforms.””Open or closed?” is not a straightforward question to answer.

          1. SubstrateUndertow

            My mind is a closed system.Oh wait maybe that is why i”m not winning 🙂

        1. tralalalalalala51

          China won with closed. It took the right approach or else silicon valley would have dominated the entire world. The EU should have done the same thing.

        2. Frank Gilbane

          In general, it is not a useful question. The tension between “open” and “closed” has been a constant of computing systems and standards for decades, and this tension has fed creative development of both. Context determines the leverage of “open” or “closed” and is fluid.

      4. SubstrateUndertow

        Does history really support the thesis that open always wins ?Maybe in the long run ? Hope springs eternal !

      5. Benedict Evans

        I’d agree with that, but there are degrees of open, and tradeoffs. iOS and Android are both more closed than Windows, but more open than a featurephone.Clearly, iOS has had tens of billions of app installs – more than all of Windows did in however many decades. So it’s pretty rich. And the ‘closed’ issues have benefits – safety, power management etc – that actually enable more downloads, and so more innovation.There’s also a self-selection issue – the benefits of the closed aspects of iOS (amongst other things) draw in the ‘best users, who drive a rich ecosystem – or, at the least, those users mean that the iOS ecosystem is strong despite those closed aspects.Perhaps the key is to identify the point of maximum leverage for openess and open *that* part.

        1. fl1nty

          http://platformsandnetworks… this article comes to mind about Google’s selective openness and closeness at a high level

      6. Sarthak Haribhakti

        Would it be a too big a stretch to call the smartphones themselves a run-time? I cant imagine calling notifications on our phones a standalone run time. Or maps for that matter. Maps, Notifications and voice are good sub-sets of a big run-time system of a smartphone. Now, messaging is essentially the biggest sub-set of this smartphone runtime. In China, though, WeChat can be referred to as a runtime in itself. It essentially runs independent of the OS the phones are running. It is truly a magnificent runtime that started out as a sub-set of the smartphone runtime system but managed to break free.

      7. Sean Hull

        As engineers we understand how open-ness is so important. But do consumers choose it? Is it even on the menu?

    2. laurie kalmanson

      that is an awesome analogy. related: early cars looked just like carriages, but without horses. it took a while until cars stopped looking like the thing they were replacing and started looking like something new.

      1. fl1nty

        Kind of like how Tesla cars need to look like traditional cars instead of looking like the first prototype Google self driving car because that is what people expect cars to look like

        1. laurie kalmanson

          i would like for teslas — and the prius and the leaf — to look like space cars or something else different

  10. creative group

    Fred:Great post. Defined and linked what the next entrepreneurs should be concentrating and focusing on.

  11. falicon

    Agree with you. Context was the core idea behind my thesis (a good 5 years ago). We couldn’t nail the business model, but I still believe the thesis was/is spot on.Context is an insanely important part of both relevance and knowledge; until recently it’s been a huge missing part in the search for “true A.I.”…so I think we’ll see even more advances around context, relevance, and of course A.I. in the near future…

  12. jared zlotnick

    You conclude with greater optimism for the mobile internet… Is that specific to app versus web?The former offers more context but the latter is more open.

  13. Pete Griffiths

    Context = data science.The new platform = datasets (proprietary and public) + AI

    1. sigmaalgebra

      For a long time that is destined to be more of a workbench than a platform.

      1. Pete Griffiths

        I like that characterization. I think you’re right. I think DS is going to be rather more widespread and influential than what we typically refer to as a workbench which is what I was getting at, but I take your point.

        1. sigmaalgebra

          I have yet to see anything in DS that is new, correct, and significant — the usual criteria. To me, DS looks like a lot of routine computing, yes, with more cycles, bytes, and bandwidth than in the past, but otherwise nearly all just some 101-102 level material in elementary probability and applied statistics, with some help from Leo Breiman’s work on trees and forests.Breiman is the bright spot, a bright guy, great background, student of M. Loeve at Berkeley, co-student of J. Neveu student of Loeve.Most of that material is in QA section of a library, and there’s much, Much more there that is good, polished, out of great work on important real applications, powerful.For statistics, DS should learn about sufficiency, Chi-squared tests, not just regression but the rest of multi-variate statistics, re-sampling (Efron, Diaconis), Monte Carlo and its refinements, of course the central limit theorem and the strong law of large numbers, the Poisson process and the renewal theorem, Markov processes, and not just elementary Markov chains, martingales and the associated convergence result and inequalities, ergodic processes, stochastic processes, e.g., second order stationary. With that material in hand, move on to optimization, control, and stochastic optimal control. Then there is more in functional analysis, Hilbert and Banach spaces.Authors include Doob, Halmos (a Doob student and assistant to von Neumann), Rudin, Loeve, Breiman, Neveu, Dynkin, Nemhauser, Bertsekas, Cinlar, Tukey. Only a small fraction of the DS people have even heard of these authors and a much smaller fraction understand what they wrote.All I mentioned is from the real history of significant real applications and is from 20 to 90 years old.There is so much of this stuff, with so many applications, that the day to day work of actual applications has to be really an experts workbench and not really a platform. For a long time, only a toy version can be a platform.DS is a beginning student lost in the QA section.DS will collect a lot of data and start to make sense out of it todescribe some real world processes. Then they will start to understand some of those processes. Then they will want to look for what decisions to make for those processes. Then they will look into optimization — best decisions. Then they will notice that the rest of the world might respond and want to look into at least von Neumann-Morgenstern game theory (easy from the theory of linear programming).Then they will notice that the process they are trying to optimize also is subject to independent, exogenous inputs and want stochastic optimal control — I wrote my Ph.D. dissertation in it.For more, suppose X is a real valued random variable and its expectation E[X] exists. Consider the vector space over the reals of all such random variables X where E[X^2] is finite. Make this into a vector space and an inner product space and a normed linear space. Observe that it is both a Banach space and a Hilbert space. Nice! Think about that a little — it’s totally mind blowing that it could be true, but it is.Take the usual topology and define convexity the usual way.Then each non-empty closed convex subset has a unique element of minimum norm. It’s true. Count on it!We were talking about Ben’s desire for search and discovery, right? Well, that minimum norm result is one of the prerequisites for my original math for my startup for an answer to Ben’s future. Gee, this means that I’m in DS? Wow.

          1. Pete Griffiths

            I think the real breakthrough has been in the scale of the datasets that are now available and that we have the tools to handle that kind of scale.Peter Norvig’s talk deals with it well:’The Unreasonable Effectiveness of Data’…

          2. sigmaalgebra

            > Scale.Right, many times more bytes, bytes per second, cycles per second. And at times that can be very good up to crucial.Yes, for Norvig’s claim, sometimes it’s true, and that fact has been something of a surprise.I first saw Norvig’s talk some months ago, and, still, I don’t think much of the content. Gee, he borrowed the title from E. Wigner?Data alone just comes up quite short of what can do if know more about the data. E.g., can empirical analysis of user arrival times at a Web site, use big data for this, and still miss what can know about those arrival times just from knowing about the situation and system and using the renewal theorem to conclude with quite high accuracy a Poisson process. Then, for the arrival rate of that process, use some data, but don’t need terabytes to do well doing that.Broadly, the computer industry and the academic computer science community are way, way behind what I outlined. They just didn’t bother to take the right math courses in college and grad school; generally it’s too late for them to learn that material now; and they definitely will not reinvent the material.For my project, knowing that material is from important to crucial and my original work is still more important. But there is next to no one in the industry I could talk to about such things. That fact is an advantage, but the flip side is that I have sell results, not technology too few people can understand.For the future of DS, the the right stuff is like I outlined and in the pure and applied math sections of the libraries, not in the computer science sections; the computer people don’t have the right stuff. So, now you can be ahead and not bet on weak horses.E.g., I worked in an AI group and basically did an upchuck at the AI ideas and then to have something much better wrote a paper essentially a fancy version of resampling. My work totally blew the doors, and knocked the socks, off the AI work we did.In AI, the emphasis is on the A not the I.I think more of, say, Breiman and the other people I mentioned.

          3. Pete Griffiths

            Errr….yeah…I guess…

  14. Semil Shah

    Given these aren’t run in a traditional web browser means “permissionless innovation” will be harder to find on mobile as (1) there will be a smaller handful of apps which have enough scale to have new services built on top and (2) either them or the OS may restrict access if something takes off on their back a la the “be your own bitch” meme from an old post of yours. This makes VC investing in mobile even trickier.

    1. Oz Har Adir

      While there will be less apps (compared to websites), there are more entry points (compared to being dependent on just Google or just Facebook to be discovered). This will mean that the apps that would make it, will be able to get more valuable than they could on desktop.

      1. Semil Shah

        Besides the App Store and places like Product Hunt, what are those discovery entry points?

  15. hypermark

    Here’s a narrative. We are moving to an Applied model of thinking about Runtimes. Uber is a beautiful example of this applied model of thinking in action. The context of what the Uber run time experience is all about — from passenger to driver to Uber, and back — is beautifully thought out and more incredibly executed.In terms of how all of these runtimes are plumbed together, I think there is a third path that is emerging: A) Embeddable, so it’s Spreadable; B) Can run within a Web Environment; and C) Can be pulled into a Native App Environment from A) or B). This is a logical anchor for where HTML5 takes root, and where Flash once had a foothold. HTML is easily sucked into native environments, as it is.

  16. MatthiasCB

    Most of the examples given were in the utility space (meals for startups, ride-sharing, fashion advisor, NBA notifications,…).Though some of the contextual environments are social services (e.g. Chat), I am wondering what are the implications for the social space? Anyone can think of connecting people based on context on top of existing contextual environments? Or is “connecting people based on context” what social environment essentially do and therefore services on top of those contextual environments would not make any sense?

    1. MatthiasCB

      And btw. “Age of Context” by Robert Scoble and Shel Isreal is a good read in this context.

  17. sigmaalgebra

    Yes, I can have significant agreement with Ben’s 7: Search and discovery and there search can only give you what you already knew you wantedThe internet makes it possible to get anything you’ve ever heard of but also makes it impossible to have heard of everything. It allows anyone to be heard, but how do people hear of you? Right: Long ago the experts in information retrieval noticed: A user (1) needs to know what they want, (2) know that it exits, and (3) already have enough information, e.g., keywords/phrases, accurate enough to characterize exactly or nearly so what they want.E.g., if I keep remembering Oh, you left out a bunch of stuff. and right away at Google find…with the transcript of the movie Back to School. with the full quote.Then another obvious keyword search finds the video clip with the quote:…As a former B-school prof where that clip is far too close to the truth, that’s one of my favorite moments in the movies!Google’s keywords worked! Why? For that phrase, I had (1)-(3).And, as I’ve spouted often enough here at AVC, IMHO that old information retrieval works only for about 1/3rd of the content on the Internet, searches people want to do, and results they want to find.So, for the other 2/3rds, need to get around (1)-(3).Then as Fred quoted, at the end Ben has But the underlying aim is to construct a new search and discovery model Well, with Ben, and maybe Fred, and me, that makes 2-3 in all the world so far that think this way!For the rest of what Ben says, I’m not much impressed.Sorry, Ben — for me I see no need for a smartphone.Lots of things have microprocessors in them, and just because a smartphone does doesn’t mean that it is in competition with PCs.Yes, I very much want a good keyboard, mouse, and a big, bright screen. E.g., there’s no way I could have written my software on a smartphone. And I want all my other relevant books, papers, etc. near my computer.I can think of some ways to make PCs much better, but none of those ways are in smartphones.Sometimes I mail letters, and I address the envelopes on an old Xerox daisy wheel printer — works great. Sometimes I print on paper, and I have a terrific B&W laser printer for that.I like music, a lot, and my PC is terrific for music. E.g., I download music, stream music, and play music on CDs and DVDs. And I want, and have, some high quality speakers.And I want to be running Windows for the software I write and for the large collection of software that no doubt won’t run on a smart phone.And my computer has about 404,000 files that I backup regularly with some good ways to find the files and back them up. No, if only for security, I won’t backup to the cloud.There’s a larger point: My view of the main driver of the future of computing and the Internet is just information.

    1. Henry Thornton

      How did the 1/3, 2/3 split come about?

      1. sigmaalgebra

        Via a lot of hours on the Internet, judgment, and, as I wrote, IMHO!For more, currently the search technique that works really well is based on just keywords/phrases in text. Well, that means that the content to be found either has some good metadata or is based on text. So, that leaves out a lot of still images, video clips, and music recordings.Next, for long tail, e.g., specialized with each instance of content for a relatively small audience, even if the content is based on text, likely more so otherwise, keywords/phrases will give too many results, and a user needs something better. And that might be? Well, we can say with some safety that for such a search what the user wants is content with some meaning they have in mind. So, want a means of search, discovery, …, that lets a user get content with that meaning. Tilt! Halt! Doing anything with meaning has been something of a Holy Grail problem in computer science. Tough to specify meaning accurately with just a few keywords/phrases; tough to match such input with the relevant content even if the content is based on text (we don’t yet have real computer-based text understanding good enough to work with meaning); and still more difficult for content not based on text, e.g., still images. E.g., one important broad case of meaning, tough for a user to characterize with keywords/phrases and/or for a computer to match with content, is artistic taste. E.g., maybe a user feelsfor me, it has to be the sea. in a beautiful forest I feel like an intruder, it is with the sea that I feel connectedas in the comment in this thread at…So, call this consideration a case of artistic taste for images of nature. Or, beyond artistic taste, just what a user feels.So, where’d I get the 2/3rds? Well, with the considerations above, maybe 3/4ths is too large and 1/2 is too small! So, the 2/3rds is an IMHO guesstimate.Or, now that we have a guesstimate of the size of the market or user need, a question is how the heck to solve that problem? Well, we could try intuitive heuristics, image manipulation, speech recognition, singular value decomposition, clusters, wisdom of the masses voting, the interest graph, semantic analysis, artificial intelligence, machine learning, …. Yup, we could try those. I doubt that any of them would work at all well for this problem.

        1. Henry Thornton

          @sigmaalgebra:disqus Thanks for the clarification – I had overlooked the “IMHO” in your original note. Not an unsolved problem and can explain directly if that is possible.

          1. sigmaalgebra

            IMHO: Maybe everything I type into the Internet is IMHO? And that is a special case of the much more general situation that everything on the Internet is IMHO?On my 2/3rds guesstimate, likely I should emphasize that that is after removing the NSFW content on the Internet. I’m not interested in the NSFW content if only because my site should be ad supported, and the advertisers I want wouldn’t want their ads next to NSFW content.Not an unsolved problem and can explain directly if that is possible.Eager readers are ready to read! Explain away! Go for it!

  18. Mark Essel

    The challenge for context is (and has been) increasing personal relevance. Getting permission and access – and leveraging that information, aka caring about saving me time or enriching my life experience. Not me the user/id, me the person.How does a business grow on a foundation of empathy and concern for individuals? by starting with one person. Reminds me of virtual assistants, the next AI frontier

  19. Phil Simon

    I’ve been saying for a long time that we live in a multi-platform world. Those platforms may ultimately morph but the fundamental model is the same: the platform.

  20. Terry J Leach

    Correction to first paragraph his analysis has been help not just to you, but many others such as myself who follow his post. As a Systems Architect and Entrepreneur, his analysis gives me a great view of the computing landscape, which along with other blog post such as this one and Albert’s and few others inform my thinking.

  21. george

    Great read and leads me to believe that we’re transitioning through another significant science method stage – moving from collecting the data, to analyzing the data and drawing conclusions (contextual runtimes). What makes the next era Epic, the defining characteristics of mobile – it’s omnipresence and people now becoming more and more of the interface.

  22. Betty Greenwald

    Unless and until Kik announces its numbers I dont think you have a fair position to state they have massive user numbers. As I understand it Kik’s MAUs and DAUs are the lowest in the industry. Imho, Kik is dead (unable to earn revenue) and only holding out for a hail-mary buyout from Wechat. And yes I understand Kik has tried to reject the industry standard measurement of success in maus and daus. Facebook releases these numbers every quarter.

  23. Betty Greenwald

    What is a “runtime”? I have programmed console videogames since 1979 and I have never seen the word used as Ben Evans does. Are you talking about system software? An App? Youtube is certainly not system software. Are you referring to anything that can have something on top of it? Is a word processor a runtime?