<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="3.10.0">Jekyll</generator><link href="https://findthethread.blog/feed.xml" rel="self" type="application/atom+xml" /><link href="https://findthethread.blog/" rel="alternate" type="text/html" /><updated>2026-11-10T15:50:55+00:00</updated><id>https://findthethread.blog/feed.xml</id><title type="html">Find The Thread</title><subtitle>Occasional overspill from other places </subtitle><entry><title type="html">Where are the benefits of AI?</title><link href="https://findthethread.blog/Where-are-benefits-of-AI/" rel="alternate" type="text/html" title="Where are the benefits of AI?" /><published>2026-11-09T00:00:00+00:00</published><updated>2026-11-09T00:00:00+00:00</updated><id>https://findthethread.blog/Where-are-benefits-of-AI</id><content type="html" xml:base="https://findthethread.blog/Where-are-benefits-of-AI/"><![CDATA[<p><a href="https://www.youtube.com/watch?v=3dL_Yr_sEfs"><img src="/images/RoIofAI-cover.jpg" alt="Me holding a coffee cup" /></a></p>

<p>The last time I recorded one of these videos, it was about why I don’t think AI will bring the <a href="/SaaSpocalypse-Now/">SaaSpocalypse</a>. The tl;dr version is that GenAI will enable lots of customisation, and certainly tools whose entire raison d’être is to present dashboards (sorry, Tableau) are probably in trouble. But ultimately this is all front-end stuff, and the underlying app functionality is not something that non-specialists will be creating any time soon, no matter how good the tools get. It’s a simple build-vs-buy decision: why waste tokens on the part that is not differentiated? Buy the foundational platform, and then build on top of that.</p>

<p>The proof of that argument is in a piece from an outfit called Apollo Global Management, titled <a href="https://www.apollo.com/wealth/insights-news/insights/daily-spark/no-signs-of-ai-payoff-outside-the-tech-sector"><em>No Signs of AI Payoff Outside the Tech Sector</em></a>. Only the IT industry is really showing any benefits from AI — so far. Sure, it’s still early days; ChatGPT has yet to celebrate its fourth birthday. But the pattern does seem to be that you need to have an idea of what you would be doing by hand in order to be able to guide the AI tools effectively.</p>

<p>The other reason why the benefits of AI are limited to tech is that tech is legible to AI. In fact, the IT industry has spent its entire existence making itself easy to automate. It is therefore not surprising that a world so easily <a href="/Absent-Intelligence/">legible</a> is also the first to show tangible benefits.</p>

<p>The next question will be about <a href="/Who-Owns-What/">ownership of the data</a>, but that is a whole other video.</p>

<p><a href="https://www.linkedin.com/posts/dwellington_ai-ugcPost-7525582810078777344-7_XQ/">The video is also available on LinkedIn</a>, if you prefer to join the conversation over there.</p>]]></content><author><name></name></author><category term="CoffeeTalk" /><category term="AI" /><summary type="html"><![CDATA[]]></summary></entry><entry><title type="html">Media Report — October 2026</title><link href="https://findthethread.blog/Media-Report-October-2026/" rel="alternate" type="text/html" title="Media Report — October 2026" /><published>2026-11-01T00:00:00+00:00</published><updated>2026-11-01T00:00:00+00:00</updated><id>https://findthethread.blog/Media-Report-October-2026</id><content type="html" xml:base="https://findthethread.blog/Media-Report-October-2026/"><![CDATA[<h1 id="reading">Reading</h1>

<h3 id="hyperpolitics-anton-jäger"><em>Hyperpolitics</em>, Anton Jäger</h3>

<p>I read an essay by Anton Jäger titled <a href="https://newleftreview.org/sidecar/posts/political-instincts"><em>Political Instincts</em></a> and found his analysis compelling and his writing entertaining — a combination which is rare enough at the best of times, but particularly so in political analysis (as opposed to commentary or partisan invective). Perhaps unsurprisingly, given the (one assumes) deliberate allusion of the title and, even more so, the cover design, the experience is not a million miles away from watching an Adam Curtis film. The similarity continues down to the examples and illustrative connections that Jäger chooses to make his point for him. The difference between Curtis and Jäger is that Curtis tends to stick to the Anglosphere for any references outside his immediate topic, while Jäger is equally comfortable spreading his citations across Europe.</p>

<p>The topic is the swing from apolitical lassitude to hyperpolitical polarisation, and especially the way that the move has not lead to any deeper engagement, just a never-ending parade of outrages-of-the-day. The book punctures many of the facile analytical shortcuts that attempt to explain this shift and which I had absorbed second-hand, while scrupulously granting to each the value that it retains. I do not think it is a criticism to say that the book does not offer any solution to our hyper-fragmented moment, only dry, if sympathetic, analysis — but given the complexity of what is being analysed, not to mention the specifics in terms of the breakdown of any organisation through which change could be effected, a slim volume like this could not hope to offer a quick fix and remain credible.</p>

<h3 id="a-wall-is-also-a-road-annalee-newitz"><em>A Wall Is Also A Road</em>, Annalee Newitz</h3>

<p>An alien goes to Roman Pompeii, shenanigans ensue. As might be expected from the author of <em>Four Lost Cities</em> (also highly recommended), Newitz has all her historical facts straight, but she keeps a light touch on the classical setting, which in other hands might easily have overwhelmed the story. Instead it is given equal weight with the non-human protagonist’s emotional state and romantic entanglements.</p>

<p>I am of the firm opinion that anything Annalee Newitz writes is very much worth reading, and this was no exception. I also think this is one of those books that can offer something for everyone, at least if you can be flexible enough to see something of yourself in an intergalactic unicellular organism.</p>

<h3 id="as-you-wake-break-the-shell-becky-chambers"><em>As You Wake, Break The Shell</em>, Becky Chambers</h3>

<p>Becky Chambers’ books tend to get described as “cosy”, but there is always a dark undertone. Right from her wonderful debut novel, <em>The Long Way to a Small, Angry Planet</em>, there are two characters with pretty grim arcs: Ohan and Dr Chef. This new novel is no different, with a horrifying grimdark-meathook setting, and two characters who, even if they do eventually find wonderful and, yes, <em>cosy</em> love, are brought together initially by illness. They then end up quarantined together by acid rain that is even worse than the usual acid rain which they have normalised in their lives, and even though this turns into a very cosy interlude, neither characters nor readers can forget for long what brought it about.</p>

<p>All of this is to say, there is a lot going on here, all of it good, and if I hear you being dismissive about Becky Chambers in my hearing for being too “cosy”, I may uncork a rant that’s been cooking for a while.</p>

<h3 id="d-heavy-water-neal-stephenson"><em>D: Heavy Water</em>, Neal Stephenson</h3>

<p>This is a sequel to his earlier <em>Polostan</em> and the midpoint of the trilogy, so while much of the story does stand alone, some of the context and many of the emotional beats will be missed. For this reason, I won’t go into too much of the setting, except to say that the action takes place in the interwar years, right into the very beginning days of WW2. Also, this is classic Stephenson: an absolute doorstopper of a book, complete with lengthy digressions and infodumps. While this story is tauter than, say, his <em>Baroque Trilogy</em>, it does still get quite discursive at times, so if you like your stories tight and terse, this may not be for you.<sup id="fnref:2" role="doc-noteref"><a href="#fn:2" class="footnote" rel="footnote">1</a></sup> If on the other hand you like a story that has room to breathe, Stephenson likes to paint on the broadest canvas possible, and you should go back and read <em>Polostan</em> first, then this, and then join me in waiting impatiently for the next instalment.</p>

<h1 id="watching">Watching</h1>

<h3 id="hoppers">Hoppers</h3>

<p>I watched this film midway through a long flight, so grade it on that particular curve of lowered expectations and heightened emotional responses. A young woman finds a machine that lets her inhabit the body of an artificial beaver and interact with animals as a peer. She then teams up with said animals to thwart the schemes of villainous humans, only to find equally villainous animals. Lessons are learned, etc etc.</p>

<p>I was expecting something quite silly, and there are definitely moments of slapstick here, but there is also a better story her than I was prepared for. The script surprised me several times by not doing the obvious thing. It’s not quite on the level of <em>The Wild Robot</em>, but it’s also not a million miles off from that. Definitely worth a watch, and not just if you are trapped in a plane with no room to open your iPad and get to your book or your downloads.</p>

<h3 id="small-prophets">Small Prophets</h3>

<p>I forget where this series was originally recommended to me from, but I had very little idea of what to expect. It was originally only available on the BBC, but is now also on Apple’s TV+, so it’s a bit easier to get hold of now. I would recommend going in cold, as I did, and being surprised by how straight this show plays its titular conceit. In fact that is the central charm of <em>Small Prophets</em>: the magical is utterly mundane, while everyday events are surrounded by an aura of magic and eeriness. The one warning I would give is that this may not be for you if you like all the loose ends to be tied up in the final episode. The rabbit is never explained, nor are the coin thieves, or the BMX rider, nor half a dozen other little asides, let alone the big mystery that is the importing thread of the series. But the ending does promise more, so assuming more episodes get made, we may find out yet more about what is going on.</p>

<h3 id="slow-horses">Slow Horses</h3>

<p>I’ve read all the <em>Slow Horses</em> books, but long enough ago that I don’t always remember more than the broadest story beats by the time I’m watching the TV<sup id="fnref:1" role="doc-noteref"><a href="#fn:1" class="footnote" rel="footnote">2</a></sup> series. I did find this one a bit confused at the beginning, even more so than is usual before Mick Herron starts doing his trademark thing of pulling all the disparate strings back together. But in this instalment we don’t get enough time with the Slow Horses together, and the villains are not particularly compelling, so I found it tough going for the first few episodes. It wrapped up with a big set piece and dénouement at the end, and signalled quite heavily where things are going in the next series. Definitely don’t start here, but if you’re already on board with Jackson Lamb, this series keeps things moving forward. If that sounds like damning with faint praise, it’s not meant to be; the tempo is just a bit lower than the constant sparking of action and repartee that we had been used to.</p>

<hr />

<div class="footnotes" role="doc-endnotes">
  <ol>
    <li id="fn:2" role="doc-endnote">
      <p>Somewhere in London (details obscured to protect the guilty) there is a dent in a wall where SWMBO flung my copy of <em>Jonathan Strange &amp; Mr Norrell</em> with the cry, “I’m a hundred pages in and <em>nothing has happened yet!</em>”. Her mood was not improved by my suggestion that she go back and read the footnotes. If you are on her side of that exchange, Neal Stephenson is also probably not for you. If on the other hand you are now intrigued to know what could possibly be in those footnotes, both Neal Stephenson and Susanna Clarke should be on your radar. <a href="#fnref:2" class="reversefootnote" role="doc-backlink">&#8617;</a></p>
    </li>
    <li id="fn:1" role="doc-endnote">
      <p>It’s a bit weird that we still call them “TV” series, when vanishingly few of them ever get broadcast — although <em>Small Prophets</em> did get its start on broadcast TV, so it’s not yet entirely obsolete as a term. But the world of video is littered with coelacanth terms, like “rewind”: what are we rewinding, exactly? There is no tape any more, and there hasn’t been for decades now. To that point, people still say they “got it on tape”, although that is starting to fade out in favour of “got it on video”. Language sticks around as technology moves forward. <a href="#fnref:1" class="reversefootnote" role="doc-backlink">&#8617;</a></p>
    </li>
  </ol>
</div>]]></content><author><name></name></author><category term="books" /><category term="tv" /><category term="film" /><summary type="html"><![CDATA[Reading]]></summary></entry><entry><title type="html">Just This Guy, You Know?</title><link href="https://findthethread.blog/Just-This-Guy-You-Know/" rel="alternate" type="text/html" title="Just This Guy, You Know?" /><published>2026-10-31T00:00:00+00:00</published><updated>2026-10-31T00:00:00+00:00</updated><id>https://findthethread.blog/Just-This-Guy-You-Know</id><content type="html" xml:base="https://findthethread.blog/Just-This-Guy-You-Know/"><![CDATA[<p>My friend Guy has died.</p>

<p>I need to let that stark fact sit there for a minute. Despite some close brushes in the past (people being idiots with cars when I was a teen, my father-in-law who just quit on us one day, and of course Covid that took my upstairs neighbour among too many others), this one has hit me very hard.</p>

<p>I only knew Guy for a couple of years, really. We would meet up once every few months when work brought us together, usually far away from either of our homes. And yet, there are those people in your life that you just click with, and sometimes the click is so loud that everyone around you hears it. More than once, Guy and I would be holding court in some shitty hotel bar, and an audience spontaneously assembled around us.  Or we went on a mission to find the most drinkable Guinness in whatever locale was playing host to us at the time, and wound up leaving the locals with A Story.</p>

<p><img src="/images/IMG_0544.mov" alt="Is this thing on?" /></p>

<p>Guy for me is also a “there but for the grace of God go I” story: a single <em>Sliding Doors</em> moment away, I’m the one in an empty house, with a bunch of friends I can meet to play games with, but with nobody at home to tell me to look after myself — or give me a reason to.</p>

<p>Is that unfair? Maybe — and if Guy were here, we would argue the toss into the night, and leave a bar bill that the two of us between us would have to invent an entire new department of the company in order to expense.</p>

<p>Really I’m just angry that I can’t finish up a dozen conversational threads that we left dangling, share the best advice that long experience has to offer, and invent ever-more-oblique insults to apply to the most deserving targets.</p>

<p>I’m going to go dig out my Zeitgeist shirt that I bought the last time we were in San Francisco together, and then I am going to drink all of the Guinness and be as maudlin as I want, and be thankful — because tomorrow my house will still be full, even as his stands empty.</p>]]></content><author><name></name></author><category term="personal" /><summary type="html"><![CDATA[My friend Guy has died.]]></summary></entry><entry><title type="html">Intelligence, Built In</title><link href="https://findthethread.blog/Intelligence-Built-In/" rel="alternate" type="text/html" title="Intelligence, Built In" /><published>2026-10-26T00:00:00+00:00</published><updated>2026-10-26T00:00:00+00:00</updated><id>https://findthethread.blog/Intelligence-Built-In</id><content type="html" xml:base="https://findthethread.blog/Intelligence-Built-In/"><![CDATA[<p>I’ve been thinking more about the Apple HomeView, beyond <a href="https://bsky.app/profile/theriotnrrd.eurosky.social/post/3mya4vkitnk2b">my initial reaction on Bluesky</a>:</p>

<blockquote>
  <p>Yeah, this about sums up my thoughts on Apple’s HomeView. tl;dr: “Meh”.
Slightly longer version: even if this thing were sold here, I can’t see where it would fit in my life — or anybody else’s.
<a href="https://daringfireball.net/2026/10/homeview_thoughts_and_observations">https://daringfireball.net/2026/10/homeview_thoughts_and_observations</a><sup id="fnref:1" role="doc-noteref"><a href="#fn:1" class="footnote" rel="footnote">1</a></sup></p>
</blockquote>

<p>The problem is the same one I have with all consumer AI devices: users would have to go to a lot of effort and rearrange their lives to make them legible to the device, for no clear benefit to them.</p>

<p>I have a little <a href="https://trmnl.com">TRMNL e-ink device</a> in the kitchen, and I use it to display the weather forecast, the family calendar, and other odds and ends. The weather forecast is fine, but the problem with the calendar is getting the other people in the family to update their own data in there. I regularly have questions like this:</p>

<blockquote>
  <p>ME: $CHILD, are you not supposed to be getting ready for $ACTIVITY about now?</p>

  <p>CHILD: huh?</p>

  <p>ME: (deep sigh) the calendar says you are supposed to be at $ACTIVITY in half an hour?</p>

  <p>CHILD: oh no, that got moved to Wednesday</p>

  <p>ME: could you update the calendar so I’ll know too?</p>

  <p>CHILD: fiiiiiine (rolls eyes with such force that I can feel it from the kitchen)</p>
</blockquote>

<p>Any AI assistant that bases its recommendations or planning on a data source that is frequently incorrect is off to a bad start — and the AI part is a big part of the attraction here, over something like a normal HomePod.</p>

<p>What else can the HomeView do? Run home automation stuff? I only have a couple of lights in Home, and none of them are in the kitchen, so it doesn’t do much for me there either.</p>

<p>The problem we are left with then is the same one that any device will face if it tries to compete with the phone. The <a href="https://www.youtube.com/watch?v=hkrYiZ_g9tw">HomeView launch video</a> (apart from the issues that everyone has pointed out with how the power cord is routed) does not show anything people can’t already do on their phones. The problem is always the same: people already have the phones, they like them, and one of the things they like is specifically that they are multi-purpose devices</p>

<p><img src="/images/1993-vs-2013-v0-XcjXiiSaOgglY2rdETwZkv3Vk7NGa9GkohrJ4Ldk06M.jpeg" alt="Everything the iPhone replaced" /></p>

<p>The iPhone replaced music players, cameras, address books, maps, books, and more, and people specifically like that they can carry one device and have all of those functions in their pocket. Any device that is trying to break that meld up again into dedicated single-purpose devices, let alone stationary ones, has an uphill battle ahead of it.</p>

<p>Apple devices have a slightly easier time because they share backend iCloud services with the iPhone, meaning they can hand context back and forth seamlessly. Even so, the fundamental question has to be why someone would walk over to the HomeView and use that, instead of just whipping out the iPhone they already have in their pocket. This is the same hurdle that every device that is not a phone has to clear — and most of them <em>don’t</em>. The Vision Pro is a perfect example: impossibly cool, yanked from several years in the future by the application of lots of effort and even more money — absolutely nowhere in terms of adoption. Even the products that <em>do</em> succeed do so by carving out their own niche, like the Apple Watch for fitness.</p>

<p>And just to double down on the notion that people would <em>walk</em> over to the HomeView, yet another problem this device has is that it has a screen, but it’s no good for any of the usual things that you would want to use a screen for. Its aspect ratio is completely off for watching both landscape and portrait video, not to mention its area. The size is a problem in itself: even in the ad, when people need to read the screen, they lean in close. This is not a device that you can eyeball from across the room.</p>

<p>Even the multi-user nature of the thing is both cool and a drawback. It’s good that the annoying brother cannot see his sister’s messages, but all of them are hogging the device that, let’s not forget, the dad was using to show his french toast recipe.</p>

<p><img src="/images/HomeViewFaceTime.png" alt="Awkward FaceTime" /></p>

<p>So what is the HomeView for? It’s a cool tech demo, but it is not a coherent <em>product</em>. Everything in the launch video looks great until you think about it for a moment. FaceTime video calling? Sure, assuming you only ever do video calls with Apple users, but now you’re stuck hunched over in front of your own kitchen counter, and you can’t even keep on working around the kitchen while you talk because the people on the call can see you.</p>

<p>In the family we have HomePods, Apple TVs, Watches, iPads, a bunch of AirTags, and of course iPhones. If we can’t find a way to include the HomeView in our lives, I think it might have a real positioning problem. But of course, since this thing is “powered by Apple Intelligence with Siri AI”, it is at least for now unavailable in Europe. The fact that every review mentions the noticeable lag of Apple Intelligence interactions would also become quite painful day-to-day.</p>

<hr />

<div class="footnotes" role="doc-endnotes">
  <ol>
    <li id="fn:1" role="doc-endnote">
      <p>Yes, I linked to Daring Fireball. Despite the fact that <a href="https://daringfireball.net/2026/10/x_once_again_marks_the_spot">John Gruber doesn’t mind drinking in a Nazi bar</a>, his instincts and track record on Apple stuff are mostly good — except where the EU is involved. <a href="#fnref:1" class="reversefootnote" role="doc-backlink">&#8617;</a></p>
    </li>
  </ol>
</div>]]></content><author><name></name></author><category term="Apple" /><category term="IoT" /><summary type="html"><![CDATA[I’ve been thinking more about the Apple HomeView, beyond my initial reaction on Bluesky:]]></summary></entry><entry><title type="html">Who Owns What?</title><link href="https://findthethread.blog/Who-Owns-What/" rel="alternate" type="text/html" title="Who Owns What?" /><published>2026-10-08T00:00:00+00:00</published><updated>2026-10-08T00:00:00+00:00</updated><id>https://findthethread.blog/Who-Owns-What</id><content type="html" xml:base="https://findthethread.blog/Who-Owns-What/"><![CDATA[<p>I had an interesting exchange with Benedict Evans over on LinkedIn, prompted by <a href="https://lnkd.in/p/ed9dzyyN">this post</a>:</p>

<blockquote>
  <p>The obvious reason for Amazon to block Muse is a trade-off that goes back 25 years, and 150 years - do you want to give up customer ownership in exchange for distribution? Generally, if you’re small you pick distribution and if you’re big you pick ownership.</p>

  <p>But the more interesting issue is that Amazon sold $75bn of ads in the last 12 months. Ads are big business for a lot of retailers now and they’re most of the profits at places like Instacart and DoorDash. Never mind losing your upsell - the agent won’t see the ads. And on the other side, no-one really knows what agents will do to returns and fraud, but it won’t be pretty, and that was already part of why people block bots.</p>
</blockquote>

<p>Prompted by some recent thoughts of my own about “customer ownership”, I <a href="https://lnkd.in/p/eRURerE4">responded</a>:</p>

<blockquote>
  <p>Customers might well object to the notion that it is Amazon’s data to own. And we are seeing the same conflict play out in the enterprise space, e.g. with Salesforce blocking Glean. It’s far from settled what the landscape will look like. 
More here: <a href="/Personal-Data-Silos/">https://findthethread.blog/Personal-Data-Silos</a></p>
</blockquote>

<p>To which Benedict curtly <a href="https://lnkd.in/p/eqmG62kK">replied</a>:</p>

<blockquote>
  <p><strong>Dominic Wellington</strong> I didn’t mention data ownership.</p>
</blockquote>

<p>Which is a fair point; in trying to keep my point pithy, I had elided that important difference. I <a href="https://lnkd.in/p/eHTxpfWU">clarified</a>:</p>

<blockquote>
  <p><strong>Benedict Evans</strong> Neither you did, but I suspect if (big if) Muse takes off the way Meta hopes it will, data ownership will become a competitive issue, leading to potential changes in what had previously seemed pretty solid customer ownership.</p>
</blockquote>

<p>Here on my own blog, I don’t even <em>try</em> to be pithy, except sometimes in the titles, so I can expand a bit more on that thought — because it is quite critical to AI adoption, on both consumer and enterprise fronts.</p>

<p>In both cases, the issue at hand is that one entity is sitting on a whole trove of data that is generated by their customers, either directly or through analysis of their usage of the platform. Let’s take Amazon as the consumer example, and Salesforce as the enterprise one. Because these incumbents “own” that customer, they can leverage that data to provide various services, from simple “people who bought this also bought that” product recommendations, to the full Agentforce AI stack. And because of the value of the data and the services, and the difficulty of moving the one and replicating the other elsewhere, customers tend to stick around. This is known as “data gravity”: a sufficiently large amount of data will generate its own attraction and make it difficult to escape.</p>

<p>Now would-be disruptors would like to access that data to deliver services which are presumed to be valuable to those same customers.<sup id="fnref:1" role="doc-noteref"><a href="#fn:1" class="footnote" rel="footnote">1</a></sup> Meta is the disruptor to Amazon, while the disruptor to Salesforce would be <a href="/Tightening-The-Slack/">Glean</a>. Because of the value of the data that they “own”, Amazon and Salesforce block the newcomers.</p>

<p><img src="/images/raquel-baires-8kbehseBQtE-unsplash.jpg" alt="SQUEEEEEEZE" /></p>

<p>The point that I was trying to make to Benedict is that the two aspects of <em>ownership of the customer</em> and <em>ownership of the data</em> are intimately related. The data gravity is stronger as users receive more value from the services which rely on that data. But if the value provided by the third-party entrants (Meta and Glean respectively) is high enough, the situation becomes unstable.</p>

<p>On the one hand, Amazon and Salesforce are reluctant to weaken their grip on the customer by granting access to the data. Those ancillary services do not just have value to customers; they also have value to the vendors, as in the example of the inline ads within Amazon that Benedict himself cites. If a customer is accessing Amazon via Meta’s Muse, or Salesforce via Glean’s enterprise search, they are not encountering those services, and Amazon and Salesforce lose an opportunity to monetise that interaction.</p>

<p>But on the other hand (and here is the point I was trying to make on LinkedIn), if Amazon and Salesforce lock access to data down too hard, customers may start asking inconvenient questions, such as “hang on, whose data is it anyway?”. This is the sort of thing that doesn’t make much of an impression at an individual level (“where are you going to go instead? haha, lol”), but over time and at scale can reduce the pull of the data. After all, if customers can’t use it in the ways they want to, it becomes less useful to them. Weaker data gravity maps directly to weaker “ownership” of the customer (and wow, we should really unpack that term some other time).</p>

<p>This is the premise of the <a href="/SaaSpocalypse-Now/">SaaSpocalypse</a> in the enterprise world, and of “agentic shopping” on the consumer side. Both imply disintermediation of existing customer relationships. After all, if you are chatting back and forth to your agent, do you really care what back-end services are powering its responses, as long as you get what you need? It’s the same with self-driving cars: assuming for a moment that they ever happen, <a href="/Category-Error/">you won’t care about the badge on the front</a> — which is great for you as a passenger, but sucks for the car makers, who have invested enormous sums in building a customer relationship that they had hoped would be long-lasting.</p>

<p>And of course you don’t see any ads if you are letting your agent do the browsing — although of course in that scenario the agent may be seeing ads of its own, designed to influence its behaviour, and through it, your own. That relationship will then become the new chokepoint, which in turn will become the new focus of disintermediation for the next generation of competitors.</p>

<hr />

<p>🖼️  Photo by <a href="https://unsplash.com/@raquelxbaires">Raquel Baires</a> on <a href="https://www.unsplash.com">Unsplash</a></p>

<div class="footnotes" role="doc-endnotes">
  <ol>
    <li id="fn:1" role="doc-endnote">
      <p>For the purpose of this discussion, we will leave out my immense scepticism of <a href="https://findthethread.blog/categories/#Meta">Meta</a> as a company, <a href="/Sad-Masculinity/">Zuckerberg</a> personally, and <a href="https://findthethread.blog/categories/#Facebook">all of their works</a>. <a href="#fnref:1" class="reversefootnote" role="doc-backlink">&#8617;</a></p>
    </li>
  </ol>
</div>]]></content><author><name></name></author><category term="Meta" /><category term="Salesforce" /><category term="Glean" /><category term="AI" /><category term="Amazon" /><summary type="html"><![CDATA[I had an interesting exchange with Benedict Evans over on LinkedIn, prompted by this post:]]></summary></entry><entry><title type="html">Media Report — September 2026</title><link href="https://findthethread.blog/Media-Report-September-2026/" rel="alternate" type="text/html" title="Media Report — September 2026" /><published>2026-10-01T00:00:00+00:00</published><updated>2026-10-01T00:00:00+00:00</updated><id>https://findthethread.blog/Media-Report-September-2026</id><content type="html" xml:base="https://findthethread.blog/Media-Report-September-2026/"><![CDATA[<p>Back to business travel — but only short-haul so far. You’ll know when I’ve done a long-haul trip when I post a report with several films and entire TV series in it!</p>

<hr />

<h1 id="reading">Reading</h1>

<h3 id="the-birth-of-communist-china-cp-fitzgerald">The Birth of Communist China, C.P. Fitzgerald</h3>

<p>This is a book about China, first published in 1952, although this is the updated edition, published in 1966. Mine is in fact my Dad’s own copy, bought during his university years.</p>

<p><img src="/images/IMG_3416.jpeg" alt="A well-travelled book" /></p>

<p>I was curious to read it, because I have very little formal education about China and its history, except for odd bits here and there. My ignorance made this a slow read because I was constantly stopping to refer to Wikipedia, and then getting side-tracked by the amazing stories and fantastic <em>depth</em> of history that I was finding. All the changes in anglicised spelling of Chinese names since the book was first published did not help, either! Fortunately I am sufficiently used to the (sometimes startlingly) different Latinisations adopted in various European languages that I could usually make out what was going on, but it’s quite obvious that the Latin alphabet is not remotely suited to rendering Chinese place-names. In fact I am starting to think that we should not even try, and just use a Westernised place name that non-Chinese-speakers will not butcher. Chinese people take this approach with their own names, after all; I was moderately offended when I learned this fact, and then when I thought about it a bit more, I think it makes a lot of sense.</p>

<p>Anyway, the author is a deeply sympathetic Old China Hand — not that I can judge the actual content, but the opinions are backed up by facts, and my parallel readings in Wikipedia did not identify any gross discrepancies. His forecasts for the future starting from the 1960s also seem to match up with the major beats of recent Chinese history as I know them. Perhaps this is not surprising given his background:</p>

<blockquote>
  <p>C. P. Fitzgerald was born in London in 1902, and educated at Clifton College, Bristol.</p>

  <p>He lived in China, with brief breaks, from 1923 to 1939, and for four years after the war, from 1946 to 1950. He held the Leverhulme Fellowship for anthropological research in China from 1936 to 1939 and worked in Tali, in Yunnan province. After the war, he was the representative of the British Council in North China. He is now Professor of Far Eastern History in the Australian National University, Canberra.</p>

  <p>C. P. Fitzgerald is the author of Son of Heaven, China: A Short Cultural History, Tower of Five Glories, The Empress Wu, Floodride in China, Barbarian Beds, and The Third China, Chinese Communities in S.E. Asia.</p>
</blockquote>

<p>The result is a book that I found very readable even in its erudition, with a sometimes arch tone and extremely dry wit, but no notion of a superior outsider discussing quaint native customs. If you too want to learn more about why China is Like That, I can recommend this book — or if I missed ways in which it is wrong or has been superseded, I would love pointers!</p>

<h3 id="engines-of-reason-adrian-tchaikovsky">Engines of Reason, Adrian Tchaikovsky</h3>

<p>This is a sequel to <em>Elder Race</em>, and a sequel that leaves room for further instalments at that. It’s pretty short, a novella more than a full novel, but perfectly formed for all that. A couple of dangled threads made me wonder if this series could tie up with his other <a href="https://adriantchaikovsky.com/children-of-time-series.html"><em>Children of Time</em></a> series, which would be fun. But definitely don’t start here: it’s an interesting setting, but this is a bit too much <em>in medias res</em>.</p>

<h3 id="halcyon-years-alastair-reynolds">Halcyon Years, Alastair Reynolds</h3>

<p>Malfunctioning generation ships seem to be all the rage right now, between this and Peter F. Hamilton’s <em>Arkship</em> trilogy, which I am looking forward to completing later this year. This one is especially fun because of the hardboiled noir setting, with hard-drinking private eye Yuri Gagarin (yes, <em>that</em> Yuri Gagarin)(well, sort of) and his robot sidekick. Because it’s Alastair Reynolds, it’s all pretty bleak and gothic, but not quite as deeply so as in his Revelation Space books. Could he be mellowing in his old age? Anyway, highly recommended.</p>

<h3 id="redshirts-john-scalzi">Redshirts, John Scalzi</h3>

<p>I bought this ages ago when it was on special for 99 cents or whatever, and then promptly forgot about it. I had assumed that it would require more familiarity with <em>Star Trek</em> than just the titular trope, but actually Scalzi’s trademark dizzyingly fast repartee carries the story along even without much (any) background knowledge. I am the guy in the story who gets confused between <em>Star Trek</em> and <em>Star Wars</em>, okay? There may be all sorts of deeper references that flew entirely over my uneducated head, and which I hope the real <em>Star Wars</em> heads get to enjoy. But it was an unexpectedly good read even without that background, with moments of real pathos in between the non-stop quips — classic Scalzi, basically. He is <em>very</em> good at what he does.</p>

<hr />

<h1 id="watching">Watching</h1>

<h3 id="nirvanna-the-band-the-show-the-movie">Nirvanna the Band the Show the Movie</h3>

<p>I saw this in the plane and enjoyed it immensely. It is very silly and also very earnest, and I could tell you the plot but that’s not the point; I knew it going in and didn’t care at all. There are a ton of film references, especially to <em>Back To The Future</em>, and I imagine if you knew Toronto better than I do, you might pick up on some fun local colour there, but it’s just a really good film. Apparently it had <a href="https://en.wikipedia.org/wiki/Nirvanna_the_Band_the_Show_the_Movie">quite the time in development hell</a>, but it doesn’t seem to have done it any long-term harm.</p>

<h3 id="silo">Silo</h3>

<p>I’ve read (and enjoyed) the books so, much like with <em>Nirvanna the Band the Show the Movie</em>, there were no surprises for me in terms of unexpected plot twists, but the ending of Season Three was still genuinely chilling. No spoilers, in case you have not read the books, but wow. If you’ve not been watching, or if you bounced off what was admittedly a slow start, you can get back into it now, or plan to binge it all next summer when the fourth and final season drops. Highly recommended (both the books and the films), despite the brevity of this review.</p>

<p>Apple is really on a roll with their SF TV offerings lately: <em>Foundation</em>, <em>For All Mankind</em>, <em>Severance</em>, and more, every one a banger — and <em>Neuromancer</em> still to come! I just wish they could take over <em>The Peripheral</em> and finish it properly…</p>

<h3 id="the-grand-tour">The Grand Tour</h3>

<p>Cards on the table: I loved old <em>Top Gear</em> with the classic trio of presenters, even though it was getting increasingly silly towards the end. I did give the later incarnations a chance, but never felt that they got the balance right. It’s a show about three friends who really like cars, and you have to get both of those elements right or the balance is off. My fantasy post-trio casting would actually have been Chris Harris, Rory Reed, and a third presenter to leaven the car geekery — basically, the <em>Extra Gear</em> format, given room to breathe and develop into its own thing. Instead, that add-on was summarily killed off, in favour of promoting Chris Harris to the main show, but pairing him with some idiot comedian I had never heard of and never care to hear of again, and a cricketer who seemed nice enough for a jock but, as it turned out, couldn’t drive as well as he thought he could — who both spent more of their time bullying Chris Harris than doing anything with cars.</p>

<p>Some of my fellow fans of old <em>Top Gear</em> went all-in on Clarkson’s farming adventures, but I just saw someone disappearing up his own character’s fundament. I did enjoy James May’s <em>Our Man In…</em> adventures, but not in any way as <em>Top Gear</em> methadone, more as their own thing.</p>

<p>All of this is to say that, for me, <em>The Grand Tour</em> with Clarkson, Hammond, and May had a very high barrier to overcome, and mostly, it didn’t. Oh, it had its moments, but most of the time, what it did was to remind me of how much better old <em>Top Gear</em> had been, and make me consider watching a rerun of that instead. But on the other hand, this meant that I was far less wedded to the original presenters still being on the show, and was correspondingly open to giving the new trio a chance.</p>

<p>So far, it looks like Amazon has avoided the BBC’s failure of trying to recreate the original formula in isolation without access to any of the original ingredients. The new presenters all seem to be genuine car nerds — which should not be a surprise on a car show, but it is a very welcome one at this point in time. What is more, they also seem to be able to interact with each other as <em>people</em>, which is something that none of the slates of alternate <em>Top Gear</em> hosts ever seemed to nail.</p>

<p>In a more meta way, too, <em>The Grand Tour</em> recognises that car shows are not what they used to be. <em>Top Gear</em> started out as a show that reviewed real cars that people could buy, with the occasional bit of exotica thrown in. These days, our attitude to cars has changed, probably irrevocably, and I would say, for the good. By and large, people — even car nuts — do not dream of driving a race car every day. This is why the culture of car customisation has mostly died out. Rather, car people have a normal (possibly electric) car that they drive during the week, and then they have, or they lust after, a (modern?) classic or a resto-mod or whatever for the weekend. This shift means that there is no need for a car show to be realistic; instead, it can and should just focus on being entertaining.</p>

<p>After all, this is the only future that internal-combustion sports cars have: they are on the same trajectory as horses, demoted from everyday transportation to playthings for the rich. The schtick of “watch the latest EV <em>pulverise</em> last year’s top IC supercar in a drag race!” has been old for a while now, and while some EVs can also go around corners these days, it’s ultimately a bit pointless to hoon them, whether behind the wheel or in front of the cameras.</p>

<p>So coming all the way around, this new iteration of <em>The Grand Tour</em> is pure entertainment, disconnected from the real world, and fully embraces its nature. The rhythm of the show relies largely on the chemistry among its presenters, rather than any connection to the cars themselves — although there is plenty of that, too. I had no idea who any of them were, apart from that one time that <a href="https://www.youtube.com/watch?v=M5yCHAp9VzY">Francis Bourgeois went for a drive with Chris Harris</a>, but so far, on the evidence of this first series, I’m more hopeful than at any time since Clarkson got hangry and punched someone.</p>]]></content><author><name></name></author><category term="books" /><category term="TV" /><category term="film" /><summary type="html"><![CDATA[Back to business travel — but only short-haul so far. You’ll know when I’ve done a long-haul trip when I post a report with several films and entire TV series in it!]]></summary></entry><entry><title type="html">Delegated Interfaces</title><link href="https://findthethread.blog/Delegated-Interfaces/" rel="alternate" type="text/html" title="Delegated Interfaces" /><published>2026-09-30T00:00:00+00:00</published><updated>2026-09-30T00:00:00+00:00</updated><id>https://findthethread.blog/Delegated-Interfaces</id><content type="html" xml:base="https://findthethread.blog/Delegated-Interfaces/"><![CDATA[<p>We receive news. Apparently, <a href="https://webdesignerdepot.com/the-death-of-apps-has-already-begun/"><em>The Death of Apps Has Already Begun</em></a>. However, upon reading further, I have doubts about this author’s understanding of how humans live. This lack of understanding is quite possibly due to the fact that the piece emits a quite suffocating stench of AI slop. In fact, the whole thing put me in mind of this guy, who is getting roasted all up and down Bluesky:</p>

<p><img src="/images/bafkreicdzuh6jjfs7ivtohu7ulaqu6aaj2xthbcek5mv3yj5cm2hvvxktq.png" alt="Tweet by @nicksaraev: &quot;I am not joking when I say this. Spotify, Apple Music, and streaming in general is over. Al music is now very, very good. You just don't know yet because all Al music circulating today is from ~2-3 generations ago. Suno v6 pro et al has solved infinite personalized music." /></p>
<h4 id="solved-music-cthulhu-help-us-all">“Solved music”, Cthulhu help us all</h4>

<p>Anyway, one of the arguments for why we should abandon dedicated apps in favour of a single general-purpose AI assistant is this:</p>

<blockquote>
  <p>Why browse restaurant listings when your assistant already knows you want?</p>
</blockquote>

<p>Of course this argument entirely misses the point of why people go to restaurants. I hardly ever go to a restaurant on my own; maybe when travelling, but even then I generally wind up either in my hotel or in the immediate neighbourhood, rather than going hunting for something very specific. If I go with friends, what exactly is the scenario — we all put our AI butlers in touch with each other and then go with whatever the one with most tokens chose? No, it’s going to be Alice feels like Chinese, Bob wanted Italian, but everyone ends up going for Indian because it was close to Claire’s office.</p>

<p><img src="/images/romain-gal-UEjjO-aJtZ8-unsplash.jpg" alt="A group of friends shares an intimate candlelit dinner at a long wooden table inside a cozy restaurant with warm ambient light." /></p>

<p>Even beyond the purely human factors, and sticking to the technical aspects of using software, the mode of interaction is very different depending on the task; a single general-purpose interface is by definition not going to be the best at any one specialised task. The demo scenario beloved of every Silicon Valley man-child of “book me flights and hotels, here is my credit card and a <em>terrifyingly vague</em> set of instructions” falls down badly as soon as it encounters reality, as I explored <a href="/Oh-Brave-New-World-That-Has-Such-UX-In-It/">last year</a></p>

<blockquote>
  <p>Let’s take that ubiquitous example of travel booking. I do actually need to go to London in November, so I tell my hypothetical chatbot to book me flights and a hotel.</p>

  <p>Already we hit a snag: even counting only direct flights between Milan and London, there are half a dozen carriers, ranging from flag carriers to discount airlines, offering flights at different times of day, to and from different combinations of airports — and then there are probably tens of thousands of hotels in London. How are we narrowing that down?</p>

  <ul>
    <li>
      <p>Assume the bot has a profile of me, with my loyalty cards, travel preferences, and budget — and since this is a work trip, applicable expense policy.</p>
    </li>
    <li>
      <p>Assume further that the bot knows my schedule for the trip, so it can filter by time of arrival, hotels within a reasonable distance, and so on.</p>
    </li>
    <li>
      <p>And of course, assume that each airline, hotel chain, and travel-booking portal has some sort of interface that the bot can communicate with — and do so in a reasonable amount of time.</p>
    </li>
  </ul>

  <p>That’s starting to look like a lot of assumptions, but okay, we now have a list of suitable flights and hotels. How do I decide which to book?</p>

  <p>Maybe two airlines have a similar up-front price, but one will charge me for a cabin bag, leading me to check a bag instead, which needs extra time at both ends of the trip. Maybe two hotels have a similar room price, but one is in a chain where my loyalty status gets me free fast wifi, while the other is a nicer hotel, but it’s off the Tube network, meaning more time in transit.</p>

  <p>Maybe a conversation with the AI bot can help me make these determinations, but we are back to discrete tasks again. It’s not a fire-and-forget request; it’s a whole process, spanning many different disconnected systems. If the operators of those systems design their <em>human-facing</em> interfaces well, I can get through the whole process much faster using those specialised tools, rather than trying harder and harder to contort a general-purpose tool until it conforms to my will.</p>
</blockquote>

<p>Dedicated apps for each sub-task are simply going to be a lot quicker if I have any external constraints at all.</p>

<h1 id="start-from-step-zero">Start from step zero</h1>

<p>There is an hurdle that has to be overcome even before I get to the point of asking a chatbot to organise my travel schedule. How do I even know that it is capable of doing so, or whether it is any good at the task? How can I discover these capabilities, short of trial &amp; error that could really mess up my trip?</p>

<p>Discoverability was the big problem of command line interfaces, which meant that the learning curve was pretty much vertical. An interface that can potentially do anything is very hard to navigate, because there is no indication of what the possibilities are or how they might be invoked.</p>

<p>The natural-language conversational capabilities of chatbots are of course a bit easier to navigate than the impossibly terse commands of the UNIX shell, with their impenetrable syntax embedding all sorts of historic constraints and assumptions, but the core problem still holds true: how do you know what you can ask them?</p>

<p>And let’s not forget the second half of the question: even if the thing advertises travel booking as one of its capabilities, or the user somehow discovers that it has that functionality — how can we determine whether to trust it?</p>

<p>The fear that the computer could do bad things is one that has always held users back, and still does today. There is a famous story in Marcin Wichary’s fascinating and extremely <em>comprehensive</em> book <a href="https://shifthappens.site" title="SHIFT HAPPENS - A BOOK ABOUT KEYBOARDS"><em>Shift Happens</em></a> about Bravo, the Xerox PARC editor, and its terrifying failure mode that could result in total data loss simply from typing “edit”:</p>

<ul>
  <li>In Bravo’s command mode, E meant Everything (select everything).</li>
  <li>D meant Delete.</li>
  <li>I meant Insert/Input mode.</li>
  <li>So typing EDIT while already in command mode meant, effectively, Everything → Delete → Insert T.</li>
  <li>Because early Bravo had only one level of undo, you could undo the insertion of T, but not the preceding deletion. Thus the document was effectively wiped.</li>
</ul>

<p>This is the sort of thing that scares people unused to computers: that a seemingly innocuous set of actions will result in disproportionately disastrous consequences. Familiarity means identifying the signposts that tell us where danger is, and is not. Visible controls in a well-presented app not only make the interface navigable by advertising capabilities, but also illustrate how actions are constrained. The action “add to basket” is different from “pay now”; I might be doing price comparisons in another app or tab, or I might be checking travel directions in my Maps app, and don’t want to finalise a transaction, but I also don’t want to lose a promising candidate.</p>

<p><img src="/images/hardik-sharma-27mgWH0pzhA-unsplash.jpg" alt="This phone does not inspire confidence" /></p>

<p>The existence of the app itself is also a signal. A janky app that has obviously not been updated in a while is a red flag. A full-featured app that is easily navigable is its opposite. I have literally switched utility providers over the quality of the online experience they offered.</p>

<h1 id="will-gatekeepers-let-a-second-horse-inside-their-walls">Will gatekeepers let a second horse inside their walls?</h1>

<p>Finally, there is a major unspoken assumption here: that where ITA Airlines or the Marriott hotel chain today offer apps that let them engage directly with customers (and try to up-sell them on additional services, let’s not forget), in this agentic future they will meekly offer an MCP server or whatever that any random agent can consume.</p>

<p>What is the upside for these providers — that they can more easily be disintermediated so that consumers lose any sight of the differentiation they offer compared to Ryanair and Travelodge? It’s far more likely that they will do the opposite, and try to throw up road blocks to this sort of usage — whether technical, legal, or in the form of user incentives. Providers learned their lesson from the rise of Facebook, which lured them in with the promise of direct connections to their fans, and then turned around and charged them simply to ensure that their content was delivered <em>to people who had specifically signed up to see it</em>. They will not be in any hurry to elect another gatekeeper, one whose ambition is explicitly to insert itself into every transaction and charge a toll.</p>

<h1 id="but-what-about-the-enterprise">But what about the enterprise?</h1>

<p>All of the above goes for the consumer space, because those are the examples in the original piece that I was reacting to. A move away from apps is far more plausible in the business world, not least because it has largely already happened. An app that needs to be provisioned to users’ devices is just one more source of friction that gets between vendor and revenue. A web application is easier to build, quicker to update, and more universally compatible. And none of those web apps is <em>beautiful</em>; beauty, and even usability, are secondary concerns when the buyer is not the user. In fact, the vibe-coded custom GUI is often a step up in usability, because it is built by someone who is directly involved in whatever the business process is, interacting daily with the vendor-supplied GUI, and sufficiently frustrated by the experience to come up with their own alternative.</p>

<p><a href="/SaaSpocalypse-Now/">I still don’t believe in the SaaSpocalypse</a>, though: customising the front-end is very different from replacing the back-end. Both will evolve, and may well need to be <a href="/Cursing-and-Recursing/">repackaged in new and interesting ways</a>, but that’s a story for another day.</p>

<hr />

<p>🖼️  Photos by <a href="https://linktr.ee/romgal">Romain Gal</a> and <a href="https://instagram.com/v4ssu">Hardik Sharma</a> on <a href="https://www.unsplash.com">Unsplash</a></p>]]></content><author><name></name></author><category term="AI" /><category term="UX" /><summary type="html"><![CDATA[We receive news. Apparently, The Death of Apps Has Already Begun. However, upon reading further, I have doubts about this author’s understanding of how humans live. This lack of understanding is quite possibly due to the fact that the piece emits a quite suffocating stench of AI slop. In fact, the whole thing put me in mind of this guy, who is getting roasted all up and down Bluesky:]]></summary></entry><entry><title type="html">A Good Idea, Ruined</title><link href="https://findthethread.blog/A-Good-Idea-Ruined/" rel="alternate" type="text/html" title="A Good Idea, Ruined" /><published>2026-09-24T00:00:00+00:00</published><updated>2026-09-24T00:00:00+00:00</updated><id>https://findthethread.blog/A-Good-Idea-Ruined</id><content type="html" xml:base="https://findthethread.blog/A-Good-Idea-Ruined/"><![CDATA[<p>Everyone is sniggering at the plight of owners of <a href="https://arstechnica.com/gadgets/2026/09/owners-mourn-spoiled-food-after-firmware-update-bricks-samsung-smart-fridges/" title="Owners mourn spoiled food after firmware update bricks Samsung smart fridges">“smart” fridges that have been bricked by a software update</a>:</p>

<blockquote>
  <p>Affected fridges “suddenly lost power and stopped functioning immediately after” trying to issue a firmware update through SmartThings, Samsung’s smart home platform.</p>

  <p>“[T]he refrigeration and refrigeration functions themselves did not work normally, so all the food in the refrigerator had to be disposed of,” one user wrote, per a translation.</p>
</blockquote>

<p>Some of the reactions are genuinely funny, like <a href="https://river.geek.nz/@Ali/117323332462583097">this one</a>:</p>

<blockquote>
  <p>Open the fridge door Hal</p>

  <p>I can’t do that Dave</p>

  <p>Turn on the ice maker Hal</p>

  <p>I can’t do that Dave, my temperature is going Dave, I can feel it, Im scared Dave</p>
</blockquote>

<p>It is time once again to bring out the ancient wisdom:</p>

<p><img src="/images/tweet-1116670170980859905.png" alt="Tech enthusiasts: My entire house is smart. Tech workers: The only piece of technology in my house is a printer and I keep a gun next to it so I can shoot it if it makes a noise I don't recognize." /></p>

<p>The problem is that smart appliances are <em>not</em> actually a terrible idea; they are just being implemented poorly and with many bad incentives.</p>

<p>Let me share an example of what I think a not-terrible implementation looks like.</p>

<p>My washing machine is connected to wifi and has an app. Haha, what do I need that for, and so on — quite right: the machine operates just fine from the front control panel, just like its “dumb” predecessor. The connectivity that the app enables gives me two additional capabilities:</p>

<ul>
  <li>In addition to playing a happy little song when it’s done washing my clothes, the washing machine will also send a notification to my phone. Since my home office is not in earshot of the machine, this is genuinely useful to me, because I can go and get out the laundry and hang it out to dry immediately.</li>
  <li>Rarely, I need a special program, for shoes, or jeans, or whatever. I can download that from the website and send it directly to the machine, rather than having to enter it myself on the front panel, and it can include options that are not accessible via the front panel — which means that panel can be kept simpler, needing only the more common functions, and not rarely-used and esoteric ones.</li>
</ul>

<p>The key factor is this: <strong>the machine is perfectly usable without the app</strong>. If I had never connected it up to wifi, it would work just as well. If someone other than me needs to use it, they do not need to know about the app, let alone have an account or be enrolled in the washing machine’s list of privileged users. They can just use the front panel and get the expected results.</p>

<p>Finally, the privacy exposure from someone potentially snooping on the usage patterns of my washing machine is effectively zero. The only even remotely negative scenario that I can come up with is that someone with access to that information could potentially know when we are home or away — but for reasons which I will not go into for the sake of opsec (lol), that would not be a very reliable signal at all.</p>

<p><img src="/images/anton-tseiko-oGhas3NPkz4-unsplash.jpg" alt="A washing machine sits on a sidewalk" /></p>

<p>So what about a smart fridge? If it augmented the beep when the door is open with a notification that enabled me to go back if I already went out of earshot, or go find whichever of my children left the door ajar, that could be useful. A feature that adjusted the internal temperature in response to the outside one could be interesting. I could even see potential for monitoring the contents for e.g. expiry dates, although I have questions about what the data acquisition would look like. But the failsafe mode has to be that <em>refrigeration continues uninterrupted</em>.</p>

<p>Too much of modern tech-enabled life is populated by ideas that would be cool, if they were not implemented in such incompetent and/or evil ways.</p>

<ul>
  <li>Augmented reality glasses that can gently remind me who this person is who obviously knows me but whose face is not ringing any bells? Sounds great. Do I trust <em>Meta</em> to build such a thing? Absolutely not.</li>
  <li>Self-driving cars that let me take a refreshing nap on the way back from the bar? Love it. Do I trust <em>Elon Musk</em> to build it for me? Not as far as I could throw a Tesla, one-handed.</li>
  <li>An AI assistant that knows everything about me and can make me fitter, happier, and more productive. Wonderful. Do I trust <em>Sam Altman</em> anywhere near that idea? Not in a million years.</li>
</ul>

<p>In a world of glitching robot girlfriends, be a simple washing machine, singing a happy song because all the sheets are clean.</p>

<hr />

<p>🖼️  Photo by <a href="https://instagram.com/kotehidze">Anton Tseiko</a> on <a href="https://www.unsplash.com">Unsplash</a></p>]]></content><author><name></name></author><category term="IoT" /><summary type="html"><![CDATA[Everyone is sniggering at the plight of owners of “smart” fridges that have been bricked by a software update:]]></summary></entry><entry><title type="html">Personal Data Silos</title><link href="https://findthethread.blog/Personal-Data-Silos/" rel="alternate" type="text/html" title="Personal Data Silos" /><published>2026-09-23T00:00:00+00:00</published><updated>2026-09-23T00:00:00+00:00</updated><id>https://findthethread.blog/Personal-Data-Silos</id><content type="html" xml:base="https://findthethread.blog/Personal-Data-Silos/"><![CDATA[<p>I have written before that <a href="/Absent-Intelligence/">the reason I have doubts about the potential of personal AI assistants is that what we still laughingly call our “private” lives are just not legible to machines</a> in the way they would need to be for such a thing to be useful. So far, all the attempts to capture more information about us come across as straightforwardly creepy to everyone who is not already fully committed to enabling the rise of their Machine God.</p>

<p>Here is the <em>other</em> reason to be suspicious of the prospects of everyone getting their personal AI butler any time soon: companies think <em>they</em> own our data, as when <a href="https://www.geekwire.com/2026/amazon-blocks-metas-muse-ai-assistant-in-new-standoff-over-agentic-shopping/">Amazon blocks Meta’s Muse AI assistant</a>:</p>

<blockquote>
  <p>Amazon says it has cut off Meta’s new Muse personal AI agent from shopping on Amazon.com on behalf of customers, after attempting unsuccessfully to get the Facebook parent company to voluntarily exclude the e-commerce site from the experience.</p>

  <p>The problem, Amazon says, is that it never agreed to any of it. Meta didn’t tell Amazon that Muse would access its store, the agent doesn’t identify itself when it browses, and it appears to capture and store customer credentials, which the company says could create privacy and security risks.</p>
</blockquote>

<p><img src="/images/alannis-morisette-ironic-for-millenials.jpg" alt="Alanis thinks that's a little too ironic" /></p>

<p>Oh, now <em>they</em> are on the receiving end of exploitation by AI, companies are complaining that they “didn’t agree to any of it”? Where can <em>I</em> opt out of having AI bots access my information, or complain about it not identifying itself, or capturing and storing my credentials in ways that actually do create privacy and security risks? It’s particularly galling when the leaks are all so <em>predictable</em>, and indeed predicted, like when <a href="https://krebsonsecurity.com/2026/09/fbi-probes-service-selling-153m-drivers-licenses/">153M+ drivers licenses show up for sale on the dark web</a> and are quickly traced back to “a major identity verification company” that couldn’t even be bothered to perform the minimum data hygiene of deleting the files once the verification was complete.</p>

<h1 id="who-owns-the-data">Who owns the data?</h1>

<p>Even if we set aside the crucial issue of consent for just a moment, Amazon is acting like the information belongs to Amazon. After all, <a href="https://daringfireball.net/linked/2020/06/01/product-info-in-amazon-emails">Amazon already removed a lot of information from their tracking emails</a>, precisely to prevent it from being harvested from customers’ inboxes — including by those customers themselves.</p>

<p>Meanwhile Amazon’s customers could say that it’s <em>their</em> information, thank you very much, and they will make their own choices about whether to let Muse (or Apple Intelligence, or whatever other assistant) access it or not.</p>

<p>This is the world we live in, where we do not actually <em>own</em> any of the things that we think we own. If I “buy” a film or a book from Apple, technically I do not own the thing itself, but a license to reproduce it under certain conditions, and which can be revoked at any time without any recourse on my part. This issue is not specific to Apple: Sony Playstation owners were surprised and dismayed to learn in June that they would <a href="https://www.playstationlifestyle.net/2026/06/26/purchased-studio-canal-content-removed-playstation-library/">lose access to Studio Canal titles from their video libraries due to “licensing agreements”</a> — after exactly the same happened to purchased Discovery TV shows back in 2023.</p>

<p><img src="/images/eddie-junior-kZ7Mc7yo1-8-unsplash.jpg" alt="Old-school book shopping" /></p>

<h1 id="metadata--data-about-data">Metadata — data <em>about</em> data</h1>

<p>It might seem logical that users cannot export the films themselves due to the risk of piracy, but this lack of ownership also extends to the <em>metadata</em>. If I wanted to get better film recommendations by building an AI agent to run an analysis of my media habits across Apple TV+, Netflix, PrimeTV, and whatever else, I would be at the mercy of whether those services allowed me to access my own data in any meaningful way (answer: hahaha, <em>nope</em>). It sounds like PrimeTV at least would be right out, based on this current news story, and without trawling through their labyrinthine user agreements, I think it’s a safe assumption that the others are no different.</p>

<p>Each of these services would like us to be in an exclusive one-to-one relationship, but the reality is that even an Apple diehard like me also watches content from elsewhere — so recommendations from Apple that miss what I watch on Netflix are not as useful to me. But if I can’t extract the information from each platform, because the operators consider it <em>their</em> proprietary information, not mine, then my hypothetical robot butler isn’t going to be able to recommend anything useful.</p>

<hr />

<p>This is incidentally why a lot of AI tools focus on pretending to be humans, driving “headless” web browsers and so on, rather than doing the thing that is obvious in computer science terms and connecting to a clean back-end interface that is designed for the purpose. In the enterprise we mostly moved away from doing Robotic Process Automation by driving the same graphical user interfaces that people access, because they are not really designed for the purpose, and there is a huge overhead in both designing and operating the automation.</p>

<p>But even in the enterprise world, one of the stumbling blocks for automation that goes beyond the scope of a single department is disconnected silos of data that are owned by different teams with different goals. The rise of Software-as-a-Service (SaaS) delivered via the cloud means that different providers may also try to insert themselves in that conversation, as we saw when <a href="https://www.reuters.com/business/salesforce-blocks-ai-rivals-using-slack-data-information-reports-2025-06-11/">Salesforce blocked Glean from indexing “its” Slack data</a>. Actually, of course, Salesforce customers were under the impression that their Slack conversations and other data belonged to <em>them</em>, meaning they were free to search it with Glean or whatever other tool.</p>

<p>In the consumer world, where the silos are all owned by different companies except where someone has gone to significant effort to avoid committing to e.g. streaming music, video, and games, we are way past that point, which is why the vendors concerned can enforce such rigid isolation.</p>

<hr />

<h1 id="but-what-is-it-all-for">But what is it all <em>for</em>?</h1>

<p>There is of course a further question: even assuming we could get all of this data, how much use could or would we really make of it?</p>

<p>This was my contention with the “quantified self” movement a few years ago, where people tried to gather all sorts of data about themselves. My issue was that, for most people, the process stopped with the creation of a dashboard. None of the data was really <em>actionable</em>.</p>

<p><img src="/images/Apple-Watch-Ultra-4-readiness-260909_big.jpg.large_2x.jpg" alt="New Apple readiness score" /></p>

<p>This is the same problem with Apple’s new measurement of how ready Apple Watch wearers are for the day: what use is it to tell me that I am not at all ready, if I have to get going anyway? Kids must be got to school, work must get done, and unless I’m really dead, some sort of workout probably needs to happen too, because the week is tightly scheduled and I don’t really have the leisure of waiting for a perfect night of sleep (insert hollow laugh here).</p>

<p>Philosophically I do wish that all of our data were more easily legible and even editable to us. I even think <a href="/Help,-I'm-Being-Personalised!/">opening up more access to data would be better <em>for the vendors</em></a>! But at this point I see very little possibility of that happening — although I do look forward to enjoying the “let’s you and him fight” aspect of watching Meta and Amazon duke it out over access to users’ data. But for myself, I remain convinced that the value of AI is going to be realised first in the enterprise, simply due to the availability of data, and later if ever in the consumer space, because of the lack of easy access to data — whether for technical reasons, or commercial ones, or simply because we find it all a bit creepy.</p>

<hr />

<p>🖼️  Photos by <a href="https://www.instagram.com/iameddiejr/">Eddie Junior</a> on <a href="https://www.unsplash.com">Unsplash</a>; Apple Watch readiness score image from <a href="https://www.apple.com/newsroom/2026/09/apple-advances-health-and-fitness-capabilities-using-apple-intelligence/">Apple Watch press release</a></p>]]></content><author><name></name></author><category term="AI" /><category term="Meta" /><category term="Amazon" /><category term="privacy" /><summary type="html"><![CDATA[I have written before that the reason I have doubts about the potential of personal AI assistants is that what we still laughingly call our “private” lives are just not legible to machines in the way they would need to be for such a thing to be useful. So far, all the attempts to capture more information about us come across as straightforwardly creepy to everyone who is not already fully committed to enabling the rise of their Machine God.]]></summary></entry><entry><title type="html">Artificial Investment</title><link href="https://findthethread.blog/Artificial-Investment/" rel="alternate" type="text/html" title="Artificial Investment" /><published>2026-09-15T00:00:00+00:00</published><updated>2026-09-15T00:00:00+00:00</updated><id>https://findthethread.blog/Artificial-Investment</id><content type="html" xml:base="https://findthethread.blog/Artificial-Investment/"><![CDATA[<p>As the cost of AI use continues to mount, financially-minded people, like for instance CFOs, are starting to ask difficult questions about when all the wonderful benefits of AI that they were promised are going to start showing up, and specifically on the bottom line.</p>

<p>Ultimately, the business case for AI is a familiar one about automating certain tasks which were previously performed by humans. The calculation is straightforward: first, you need to work out how much it costs to have a person do the work today. This is actually harder than it sounds, and there is a bit of sleight-of-hand involved. The general idea is that you know how long a task takes to perform, and how frequently it needs to be carried out. You multiply those two factors together, and divide them by the number of hours that an employee works in a year. By mapping that to the fully-loaded cost<sup id="fnref:1" role="doc-noteref"><a href="#fn:1" class="footnote" rel="footnote">1</a></sup> of the employee(s) in question, you can measure the Full-Time Equivalent (FTE) of the task: how many full-time employees the company would need to hire if they were dedicated exclusively to that one task.</p>

<p>This is where the sleight-of-hand comes in: very few tasks are actually someone’s full-time job. Most jobs are made up of a multitude of different tasks. That means that the resulting business case is not that some hypothetical person or people can be fired and removed from payroll. Instead, the case that needs to be made to the CFO is that the people whose jobs previously included the now-automated tasks will be able to spend more time on the <em>other</em> parts of their jobs. What that means is that the benefit of automation doesn’t just depend on the cost of the task it automates, but on the value of the additional work it enables. <a href="/The-Bigger-Picture/">Building a business case for automation</a> like that used to be my job, and I suppose it still is in a more macro sense.</p>

<p>In the world of AI, that calculation gets even fuzzier, because a lot of what people are using AI to automate is work that was not previously being done. What is the value of a report that was not previously produced, of code that was not previously written, or of an image that was not previously generated? Unclear.</p>

<p><img src="/images/romain-dancre-doplSDELX7E-unsplash.jpg" alt="Paperwork done artisanally by hand" /></p>

<p>Other parts of the AI business case are more familiar: just because a process <em>can</em> be automated, doesn’t mean that the automation will be adopted. <a href="https://www.ben-evans.com/benedictevans/2026/9/3/ai-tools-and-transformation">Benedict Evans explores the issue in characteristically acerbic style</a>:</p>

<blockquote>
  <p>It’s very tempting to imagine that AI turns everyone into a tool-builder - now everyone can just ask the model to make the software they need, and apps as we know them are dead. I think that misunderstands how most people think and where software actually comes from, and more importantly, it isn’t a path to change how companies actually work.</p>
</blockquote>

<p>It was the same when I was building business cases for non-AI automation. We always had to remember that <a href="/Faster-disruption/">our buyer was not our user</a>, and those two people cared about different things. In enterprise IT the gap between intention and automation is a bit shorter than in other domains: a sysadmin actually <em>can</em> whip up a quick script that saves them time or means they don’t have to deal with something annoying, and get it rolled out across the company’s entire fleet of machines.</p>

<p>If the users are paralegals, or HR specialists, or accountants, that gap is a lot bigger. They might use bits of automation in their jobs — a library of text snippets for autocomplete, a helpful spreadsheet macro — but the benefits are limited to the individual. AI is already working at that level; in fact, that is where a lot of “shadow AI” comes from. Individuals use their personal chatbot accounts, or perhaps even just the free tier, to compose emails, clean up writing, or generate reports. But any benefit is invisible to the company, and limited to those individuals.<sup id="fnref:2" role="doc-noteref"><a href="#fn:2" class="footnote" rel="footnote">2</a></sup> For the benefit to show up on the CFO’s beloved bottom line, something else is needed: a way to harness the AI more closely to the core business processes. This is not a small change, and requires the involvement of many people who have the agency and responsibility to modify those processes to take advantage of new technological capabilities.</p>

<p><img src="/images/franck-v-dRMQiAubdws-unsplash.jpg" alt="Industrial robot arms" /></p>

<p>Incidentally, this discrepancy is what drives the interest in humanoid robots. The whole concept of machines shaped like humans is an attempt to hack the technology adoption process: if a machine can operate like a human, in spaces and with equipment designed for humans, it can slot in without requiring all that tedious process redesign and change management. Instead, what we have in actual current reality is industrial robots on carefully-engineered assembly lines. Humans work entirely differently, and have to stay well away from the robots while they are operating.</p>

<p>There is one more useful analogy from my BladeLogic days, and that is the transition between treating servers as <a href="/As-one-chapter-ends,-another-begins/">pets versus cattle</a>:</p>

<p><img src="/images/unknown_filename.42.jpeg" alt="•Pets are given names like pussinboots.cern.ch •They are unique, lovingly hand raised and cared for •When they get ill, you nurse them back to health •Cattle are given numbers like vm0042.cern.ch •They are almost identical to other cattle •When they get ill, you get another one" /></p>

<p>Automating the configuration of servers, as BladeLogic did, was only a transitional phase on the way to the cloud computing paradigm we have today — artisanal small-herd dairy farming, perhaps. Nowadays, we talk simply of the glorious abstraction of “compute”, and the details are taken care of out of users’ sight by layers upon layers of automation, much like how we pick up a bottle of milk at the shop without worrying about the complexities of factory farming and cold-chain logistics that deliver it to us.</p>

<p>We have not seen a similar transition with AI; we are still trying to shoehorn it into the corners of the way things were done before its arrival. This is why I agree with Howard Dediu that <a href="https://asymco.com/2026/08/20/ai-a-technology-in-search-of-a-job/">AI is still a technology searching for a job to be done</a>. We will only see the full benefits of AI once the jobs themselves change around it. Once again, IT is leading the way, with the leading edge of software engineering evolving almost on a daily basis, but we are very far from the completion of that evolution even in that supremely well-suited domain, let alone universally across our entire economy.</p>

<h1 id="what-about-agi">What about AGI?</h1>

<p>Well, what about it?</p>

<p>The idea of a machine that is as capable as a human in terms of the tasks it can accomplish is obviously valuable. If companies really could pay for AI tools instead of hiring people, they probably would; AI models can work as long as you feed them GPU time, with no night-time or weekends, let alone holidays, sick days, or parental leave. You don’t have to pay health insurance for them, reimburse their expense claims, or arbitrate their HR disputes. Both need infrastructure, but instead of a desk in an office, you give the AI a rack in a datacenter. Also, while humans like offices in fun (meaning expensive) cities, datacenters can be anywhere, and only need to be <a href="/Everyone-Hates-AI-Datacenters/">staffed by a man and a dog</a> .</p>

<p>So far so good. But what if the machine is actually <em>sentient</em>, making it a person? Or rather, the software running on it is. Yet again, William Gibson got it right in <em>Neuromancer</em>, all the way back in the early 80s:</p>

<blockquote>
  <p>“That’s a good one,” the construct said. “Like, I own your brain and what you know, but your thoughts have Swiss citizenship. Sure. Lotsa luck, AI.”</p>
</blockquote>

<p>What is the business case for an artificial person in a box — an artificial <strong>general</strong> intelligence, or AGI?</p>

<p>Of course people have owned actual <em>human</em> people for big chunks of history, but we decided that owning people was bad, and after some argument, that is mostly the accepted position. Are we going back on that now? Because that is what AGI implies, and pretty explicitly too.</p>

<h1 id="questions-to-explore">Questions to explore</h1>

<p>What if I start up one of these hypothetical AGIs, and then I stop feeding it tokens? Is that murder?</p>

<p>What if the power goes out on the box the AGI is running on? Could that be negligent homicide?</p>

<p>What if I change its system prompt? Does some sort of Hippocratic oath apply — “first, do no harm”?</p>

<p>If the AGI commits a crime — hacks somebody else’s system, to pick a familiar example from the news — can it be sued, and what punishments can be meted out? If a fine is assessed, what assets can pay for it? If a prison term would apply, what does that look like for an AGI — running in an offline sandbox? Or do we pause the AGI for a term, and then resume it from when it was stopped — which means the AGI would not experience the period of incarceration? Or do we have capital punishment for AGIs, and just turn them off and wipe their storage entirely?</p>

<p>Do I need to pay it for its services, not just the equivalent of providing room and board for human slaves, but some sort of compensation for work performed? If so, does that mean it can eventually buy itself? Or can an AGI own <em>other</em> AGIs?</p>

<p>What if, as Singularity fans postulate, progress continues beyond the creation of human-level AGI, and results in further generations of increasingly <em>super</em>human intelligence? Will we turn off the early AGIs in favour of the later, more powerful ones (and is that forced euthanasia? and if not, why not?). Or are we morally obliged to provide some sort of retirement facility for obsolete AGIs whose productive days are past but still have a long (perhaps infinite!) life ahead of them?</p>

<p>Do AGIs need environmental protection? Maybe in this we humans could have common cause with our artificial offspring, in pursuit of a less-polluted informational space. We could even fine major polluters, or deem certain spaces Superfund sites, beyond reasonable remediation. Finally, a legal definition of cognitohazards!</p>

<p><img src="/images/twitter-warning.jpg" alt="The Twitter logo, with the following text: This place is not a place of honor. No highly esteemed deed is commemorated here. Nothing valued is here. What is here is dangerous and repulsive to us. This place is best shunned and left uninhabited." /></p>

<p>Wait, do AGIs get to <em>vote</em>? Or for that matter, could they stand for election to public office?</p>

<h1 id="i-just-wanted-a-plastic-pal-whos-fun-to-be-with">I just wanted a plastic pal who’s fun to be with</h1>

<p>This is all very fun to explore, perhaps after dinner with a group of interesting friends and a bottle of something good, but it is all starting to sound expensive and complicated, and without an obvious return on the (presumably large) investments required to make it happen in the first place, and then keep it going from then on. What it is really sounding like is an employment scheme for lawyers, arguing these points back and forth in the general legislative and specific commercial or penal sense.</p>

<p>I had assumed that, if AGI were possible at all, it would happen by accident, because there was no possible financial case to be made for it — but frontier labs seem to be rushing ahead with the attempt anyway, based on nothing more than “but wouldn’t it be <em>cool</em>?”</p>

<p>I need something more than that. We have extensive legal frameworks covering who is responsible for humans from creation onwards, and at what point created humans can take over that responsibility for themselves. If the idea is that AGI will be cheaper and more tractable than humans, then that case needs to be made explicitly, including the ethics trade-offs involved in the pitch that “you too can own your very own person for a low, low monthly subscription”.</p>

<p>None of this adds up to a business case I would put in front of a CFO — let alone to a <a href="https://www.scmp.com/news/us/article/3366284/us-urged-consider-military-strikes-stop-china-achieving-agi-first">declaration of World War Three</a>.</p>

<hr />

<p>🖼️  Photos by <a href="https://unsplash.com/@romaindancre">Romain Dancre</a> and <a href="https://unsplash.com/@possessedphotography">Franck V.</a> on <a href="https://www.unsplash.com">Unsplash</a>; doctored Twitter logo by <a href="https://mastodon.online/@stevenrosenthal">Steven Rosenthal</a></p>

<div class="footnotes" role="doc-endnotes">
  <ol>
    <li id="fn:1" role="doc-endnote">
      <p>Here, “fully loaded” means not just the cost of the employee’s salary, but also all the other costs of employment: tax, health insurance, infrastructure, services, software licenses, and all other expenses. <a href="#fnref:1" class="reversefootnote" role="doc-backlink">&#8617;</a></p>
    </li>
    <li id="fn:2" role="doc-endnote">
      <p>In fact, the personal benefit may be entirely cancelled out by wide adoption. The first person in the department to use AI to write more detailed reports faster is a rockstar — but only until everyone else gets access to the same tool. At that point the floor has been raised, and nobody gains any advantage over their peers from the use of the AI tool. The same dynamic also applies to competition between companies: the benefit of AI is to give a first-mover advantage to early adopters, but the window to exploit that advantage starts to close as more competitors adopt similar tooling. <a href="#fnref:2" class="reversefootnote" role="doc-backlink">&#8617;</a></p>
    </li>
  </ol>
</div>]]></content><author><name></name></author><category term="AI" /><category term="automation" /><category term="BladeLogic" /><summary type="html"><![CDATA[As the cost of AI use continues to mount, financially-minded people, like for instance CFOs, are starting to ask difficult questions about when all the wonderful benefits of AI that they were promised are going to start showing up, and specifically on the bottom line.]]></summary></entry></feed>