This is a loose collection of thoughts based on a brief conversation from the other night. I can’t claim any credit for any of the smart bits.

Let’s take it as a given that Samsung’s attempt to try to pass off high resolution, synthetic moon photos as some kind of enhancement to its phones’ camera systems is boorish and lame. Have some self-respect, Samsung!

The moon is notoriously difficult to photograph, even by professionals with thousands of dollars of equipment. The moon looks different sizes depending on its place in the sky but that’s just our brain playing tricks on us and we don’t even know why. None of this stops people from trying to take pictures of the moon, even (especially!) with their phones. The moon is probably the most photographed single object across everyone’s photo libraries.

Back in 2007 (the year the iPhone and it’s 2 megapixel camera were released), researchers scraped photos of landmarks like the Statue of Liberty and the Notre Dame Cathedral from Flickr and used those to render fairly accurate 3D models just from the flat photos. Fifteen years later, there are incalculably more photos of even relatively obscure landmarks, taken at higher resolutions, and even more powerful synthetic image generators that can be run on your phone.

I learned to shoot on film cameras. I even learned to process film and post-process in a dark room, actually dodging and burning. In journalism school, we still shot film but scanned the negatives and used Photoshop for cropping, cleanup, and color correction. The threat of failing the entire class hung over everyone who even thought about digitally altering their shots beyond basic darkroom techniques. Online, though, the world blew past that and now we’ve lived through epochs of ‘SHOPPED, Instagram filters, then #nofilter, Snapchat lenses, magic erasers, and now fully synthetic images. Even the supposedly real, unfiltered images we all take on our phones are the result of billions of calculations that compose “computational photographs”. It’s hard to argue with the results.

I certainly won’t be the one to argue that an AI-generated moon is a computational photograph. I also won’t begrudge anyone for wanting to remove that random stranger from the background of their otherwise perfectly composed selfie. I abuse portrait mode on my phone because it makes pictures of my kids look amazing. It won’t be long before we’re all carrying a neural net around in our phones that will compose perfect — too perfect — photos of anything. Even the moon.

It was my birthday this weekend so, short of a quick joke, I didn’t follow the collapse of Silicon Valley Bank all that closely. Here’s a good rundown with a healthy dose of strident skepticism to boot. In the end, it seems like the post Dodd-Frank system worked as it should have (and had it not been kneecapped during the Trump administration, may not have been needed in the first place), with the depositors getting their money back, SVB’s leadership out of a job, and an admonishment from future-Senator Katie Porter.

I keep thinking about something Paul Ford said a few years ago. I can’t find the quote (I think it was in a podcast), but in essence it was about how tech hates the fact that it still has to rely on finance. Founders need VCs and VCs, for all their world-changing bluster, still need LPs to fill their coffers. Silicon Valley won’t admit what’s been obvious for years — that it’s fully beholden to Wall St. — and the collapse of an institution that by all accounts served them incredibly well is the end result. Crypto was another attempt to try to renege on their faustian bargain, and that’s played out in predictably terrible ways as well.

I’ve been gathering string for a while about all of the machine learning / LLM / generative AI¹ excitement that’s been happening over the past year² in the hopes of having something interesting to say. Thankfully, Robin Sloan has been deep in the work and has a piece that very much mirrors a lot of where I’m coming down (or probably more accurately, I’ve rather artlessly copied his takes).

The post is fairly wide ranging, in the best way, and this bit struck me in particular:

I think it’s really chal­lenging to find the appro­priate stance towards this stuff.

On one hand, I find critical deflation, of the kind you’ll hear from Ted Chiang, Simon Willison, and Claire Leibowicz in this recent episode of KQED Forum, appro­priate and useful. The hype is so powerful that any correc­tive is welcome.

However! On the critical side, the eval­u­a­tion of what’s presently before us isn’t sufficient; not even close. If we appre­ciate humility from AI engineers, then we ought to have some humility ourselves. Humility — and imagination.

An important fact about these language models — one that sets them apart from, say, the personal computer, or the iPhone — is that their capa­bil­i­ties have been surprising, often confounding, even to their creators.

AI at this moment feels like a mash-up of programming and biology. The program­ming part is obvious; the biology part becomes apparent when you see AI engineers probing their own creations the way scien­tists might probe a mouse in a lab.

The simple fact is: even at the highest levels of theory and practice, no one knows how these language models are doing what they’re doing.

Programming and biology! In my own mental models, I’ve replaced the biology bits with something more akin to raising a child, which is fairly hackneyed at this point. The training, the frustrations, the joy at something completely unexpected happening. Child rearing, though, happens so slowly over the course of a lifetime, whereas the language model advances are more like the humble Drosophila melanogaster, that workhorse of biotechnology that somehow manages to be cheap, fast, and good.

For my own part, something as hyped as generative AI almost requires me to turn up my nose in distaste (see also: crypto, NFTs, web3). And yet! My small explorations are some of the most fun I’ve had ever with computers. I’ve got a few decades worth of digitized thoughts and to be able to coax some meaning out of them with a handful of Python scripts (in a thus far unsuccesful effort to post not just more but better) is deeply intriguing. The idea of a universally accessible interface to the computer — an electric bicycle for the mind, as Simon Willison put it — is exciting not just to me personally but as someone who sees the often unfulfilled promise of technology that empowers.

Multiverses and split timelines are so hot right now, even if it’s often in the service of stuffing more intellectual property into amusement park movies. Rarer is to get a unique, tiltshift glimpse into a more mundane world that might have been, if only as a bit of winking humor.

I don’t remember when I started following @nanoraptor, aka Dana Sibera, but her posts — mutated retro tech recreations — have never failed to amuse me. For one, she’s an artist with photoshop, with an auteur’s sensibility, an impeccable eye for detail, and a deep knowledge of computer history. She’s also hilarious, provided you are a nerd of a certain vintage and take a moment to peel back the many layers of her art. And there’s something truly delightful about Sibera’s artisanal anachronisms that stands out in a world where Midjourney or DALL•E could probably churn out some artless approximation with relative ease.

Marcin Wichary (recently seen here) curated a number of her creations and wrote a really wonderful appreciation, naturally with an emphasis on the keyboards. I thought I’d seen most of Sibera’s work, but many of these were new to me, and several lead me down explorations into actual computer history I’d never seen, like the stunning Apple Jonathan. What I love about both Sibera’s work and Wichary’s appreciation is you can tell they ache for many of these to exist.

Take few moments to live in a world that never really could, but absolutely should, have been.

One of my complaints about the new Artifact AI newsreader is how it seems to ignore the reinvigorated (at least in nerdier corners) open web, like RSS and federated social media.

So, it’s interesting to see one of the original newsreader/aggregators, Flipboard, has added Mastodon support to their app. It’s been literally years since I used Flipboard, so I downloaded it, reset my password, and gave it a go. It works!

Flipboard, as a concept, never really worked for me, though. The unique interactivity — that eponymous flip in lieu of a scroll — seems like a nice demo that simply never held up to regular usage. I seem to recall a lot of excitement about their layout engine, which promised to deliver magazine quality page configurations but is ultimately sadled with the content it’s ingesting. This is doubly true for social media feeds, where the app is trying to create a genuine pastiche¹ from … well, tweets. I ended up having the same problem with the Readwise Reader app — it’s just not my thing in the way NetNewsWire is.

But we’re not here to review a twelve-year old newsreader app, we’re here to celebrate their embrace of the open web. Which makes it a bit more interesting that Flipboard has also set up a Mastodon instance of their very own to make it easier for anyone to get going with federated social media. Flipboard co-founder and CEO Mike McCue has a post detailing their embrace of the “federated future”.

Like I said, Flipboard the app never really stuck for me. But clearly plenty of people use it, and that’s great. Some of them might even start posting interesting quips over on flipboard.social and I’ll give them a follow. Just not on Flipboard.

Artifact, the social news feed app built and funded by the co-creators of Instagram, launched to the public today. I’ve been using the beta version for a few weeks, and I remain underwhelmed, both with the product and the ethos of the app.

Unlike seemingly everyone else who nodded to the beta announcement, I was decidedly not excited to hear the duo of Kevin Systrom and Mike Krieger were building a news app. I stopped using Instagram not long after they sold to Facebook, I’ve never considered Instagram “the good social network” and, in fact, think it’s pretty obvious that everything that makes Instagram bad was baked in from its founding. So, I was skeptical from the start Artifact offered much that was new, let alone would contribute to a better informed citizenry.

Even with this admittedly critical eye, I was to see if they’d cracked some code the many, many attempts at personalized news feeds hadn’t. Maybe the tech that Artifact is hanging its hat on really is as transformative¹ as Systrom believes and the long sought ideal of delivering a perfect, personal stream of news is at hand, or at least close. Even after putting in the work (and training Artifact often feels like work), I continue to find its recommendations to be somewhere on a spectrum between inscrutable to unhelpful.

Most of what Artifact considers alert-worthy, for example, have turned out to be the articles I’m most likely to tell it to send me less of. Until you’ve silenced enough of the content mills, the app feels like one big chum box. Once I’m in the app, I will usually find a link worth reading — but is that really a problem anyone has these days? Does anyone need, or even want, an endlessly scrolling “TikTok for news”? Even as a fairly diligent tester (tech + media is very much my beat), I haven’t found Artifact to be any more useful than Prismatic (RIP), Nuzzle (RIP), Flipboard, or even Apple News before it; my creaky collection of RSS feeds remains more serendipitous.

They’ve limited their dataset to fairly reputable news sources and blogs (instead of the entire internet), and have added only the most basic of social features (which require you to upload your entire address book, of course). These feel more like cover-your-ass maneuvers that reveal only the most surface-level lessons learned from the past decade or so. And with so much energy moving towards federated social networks, RSS, and a renewed interest in the open web, doubling down on a walled garden approach² feels downright antiquated.

All of which gets to my deeper problem with both Artifact and why I started off so deeply skeptical of its provenance. What is the purpose of this app? According to Systrom, his big revelation was in seeing how large language models and transformers would fundamentally change how recommendation systems work. Artifact seems to be answering the question of “what if we ingested all the news and ran it through this new tech, then showed people headlines they were more likely to read.” At no point does the app, or anyone working on it, seem to be interested in answering the question of “how do we help people better understand the world they live in.” Maybe I’m just too old school, but it seems to me that’s the fundamental purpose of all that news to begin with.

One of my many annoyances with tech is how they refuse to understand history, especially their own; they don’t even seem capable of listening to their own lionized heroes. It was none other than Steve Jobs who said, “you’ve got to start with the customer experience and work backwards to the technology,” yet Artifact doesn’t seem to know what it’s for. It has all the hallmarks of a well-built app — it’s performant, the layouts are crisp, the onboarding flowed along just fine — but that’s only the most shallow understanding of UX, let alone design.

We’ve seen this same playbook from tech time and time again. If you actually wanted to make the world a more open and connected place, you wouldn’t build Facebook, you’d make it easier to build on the open web. If you actually wanted to reimagine transportation, you wouldn’t build Uber or Tesla or hyperloops, you’d build trains. Tech continues to offer these non-solutions because they are easy, or fence in more of the commons, or they’re just enamored with some shiny feature and don’t give a damn about any of the knock-on effects.

If all of this sounds a touch dramatic for an app that feeds you headlines, it’s because I was there arguing with my bosses against adding facebook widgets to story pages or committing to platform lock-in. If it sounds like I’m hating on LLMs or transformers or what’s being hailed as AI, I’m actually as excited as I’ve been since I first learned to build a web page. But I want tools to help journalists do their jobs better — an ML-assistant that can summarize public meetings, investigate massive troves of financial data to ferret out white-collar crime, help human journalists understand and explain arcane technical topics, or offer personalized context about a news story to every reader. Instead, we get an infinite number of auto-generated Buzzfeed listicles and a startup that once again wants to take all the benefits of the open news ecosystem without offering anything back. It’s as cynical as it is unsurprising at this point.

UPDATE
February 23, 2022

Mere hours after I hit publish on this post, Ben Thompson published an interview with Systrom and Krieger about the launch of Artifact³. In the final section of the interview, they discuss pretty directly most of the critiques I made here, about which I have a few more thoughts.

First is Krieger on working backwards to the technology:

MK: We talk consistently about this being, and it’s going to sound funny, hopefully the last company we found in our lives, and not because we don’t want to do it again, but because it can be a platform for realizing all of what we want do in terms of taking what’s at the forefront of technology — I think you can break this down to what Kevin and I do — we’re not researchers, we’re not the people trying to build things that are twenty years out from the technology and you have to squint, “Ah, maybe that’s going to work”. We take what’s at the forefront, on the cusp, and try to bring it to bear with what I think is strong product sense and engineering chops, and practicality.

…

I think it involves the taste to understand, all right, what about this is begging for consumer products to be built and how do we do that really well? Not like, “Yeah, this technology exists. Let’s go solve a problem with it.” But based on everything that we know, what can we build that will solve a problem? And the cool thing about that is that that’s a template or almost like a recipe that I think we can follow for a long time, and do it with the right teams and do it together and evolve over time, and that’s the core of what we’re trying to do. In general I see this as a platform where this is what we’re working on right now, but the approach doesn’t change. It’s just the manifestation that might over time.

I think the founders genuinely believe Artifact has, in fact, started with the product but that’s exactly the problem, treating news as just another product. This is going to veer into the philosophical quickly, but it’s actually important in a deeply social sense to understand how the news is different from content or even media. Of course, that misunderstanding has been so embedded in Silicon Valley for so long it’s practically a truism.

Systrom then addresses my other big question:

KS: I think people always want to know what’s going on in the world around them, and the way they learn about that is through various mediums throughout time. It can be through the printed newspaper, it can be through a magazine, it can be through a radio show, it can be through a podcast, it can be through a video on TikTok, it can be on cable television. You name it, they can learn about the world around them. But that job — we use jobs-to-be-done theory a lot in our work — that job-to-be-done about learning about the world around you and going deep on your interests will never go away and we just happen to do it with a specific medium, which is easy to do asynchronously when you’re sitting there. Whether you’re in bed or you’re waiting for a meeting, you don’t have to listen to it. You don’t have to have it on audio, you don’t have to have headphones in.

I suppose I’m slightly more optimistic they at least understand the purpose, or maybe a purpose of Artifact, but they seem more hung up on how their chosen medium differentiates them.

More revealing are the topics that didn’t come up: does Artifact plan to hire a public editor? What tools will they offer other media organizations? If they succeed in backing into a Twitter-like social network, how will they manage the user generated content — more AI? How will they take responsibility for their synthetic copy once they start, for example, auto-generating summaries? How will Artifact deal with the avalanche of disreputable, synthetic content that is surely just around the corner?

It would be impossible to answer these questions at launch, but the will to answer them needs to be there. We’ll see if Systrom and Krieger are up for the actual work now that the app is out there.

When the iPhone was first unveiled in 2007, it included a “pretty sweet solution” for developers who wanted to build their own apps for the new mobile platform: web apps. Anyone could build a web app and even save it to the iPhone’s home screen, just like the built-in apps¹. When the Apple app store was unveiled a year later, it ushered in a wave of new development, not to mention a revenue stream that would qualify as an excellent business if it weren’t a line item on the balance sheet of the world’s most valuable corporation.

Web apps remain a viable option for app-like experiences on Apple devices, and the web community has evolved best practices for creating single page apps (SPA) that work more like apps than traditional web pages.

I’ve always considered Apple’s demotion of² web apps to be a balance of needing to optimize the experience for the majority of their users with responding to the incentives created by their app store. As a web partisan myself, I acknowledge SPAs can be a suboptimal experience if the developer hasn’t put in the work to optimize for the installed-state; and of course Apple clearly wants to encourage as much development as possible to stay within the ecosystem it controls.

Still, it’s been 16 years. An entire generation of apps-that-should-have-been web pages has sprung up, often shoehorned into supposedly cross-platform frameworks that create a subpar user experience sludge. Nowhere is this more true than for media — how many apps from newspapers or magazines have you installed, solely for a very specific purpose like receiving breaking news alerts? How many of those apps are just wrappers around web views? How many of those apps should have been web pages?³

So, it’s incredibly encouraging to see the latest iOS and iPadOS betas (16.4, beta 1) bring some much-needed functionality not just to Safari but to single-page apps (Apple calls them “Home Screen web apps” [oddball capitalization all theirs]). Here’s Brady Edison and Jen Simmons writing for the official WebKit blog:

Now with iOS and iPadOS 16.4 beta 1, we are adding support for Web Push to Home Screen web apps. Web Push makes it possible for web developers to send push notifications to their users through the use of Push API, Notifications API, and Service Workers all working together.

A web app that has been added to the Home Screen can request permission to receive push notifications as long as that request is in response to direct user interaction — such as tapping on a ‘subscribe’ button provided by the web app. iOS or iPadOS will then prompt the user to give the web app permission to send notifications. Once allowed, the user can manage those permissions per web app in Notifications Settings — just like any other app on iPhone and iPad.

The notifications from web apps work exactly like notifications from other apps. They show on the Lock Screen, in Notification Center, and on a paired Apple Watch.

This is the same W3C standards-based Web Push that was added in Safari 16.1 for macOS Ventura last fall. If you’ve implemented standards-based Web Push for your web app with industry best practices — such as using feature detection instead of browser detection — it will automatically work on iPhone and iPad.

I have to admit, I read through this post several times because it was disappointing to learn that new push APIs only apply to web pages saved to the home screen and not, as on desktop Safari, just normal tabs the browser. I understand the choice both from a user experience and system resource management perspective, yet, if you want to enable push notifications for a web site, a user first has to save the page to their home screen.

Web developer extraordinaire Jeremy Keith is similarly disappointed and lays out the issue precisely:

I think that would be fair enough, if it weren’t for the fact that adding a website to the home screen remains such a hidden feature that even power users would be forgiven for not knowing about it.

…

As long as this remains the case, we can expect usage of web push on iOS to be vanishingly low. Hardly anyone is going to add a website to their home screen when their web browser makes it so hard.

It’s harder to describe to a non-developer how to save a web page to their home screen than it is to simply point to an app in the app store, especially since Apple has a system in place for easily linking from the web to their store with a single line of metadata.

Obviously, I’d like to see that home page installation hurdle easier to overcome for web developers. One way it could work is when someone chooses to enable notifications from a web page, the built-in OS-level alert that requests notification permissions could include something along the lines of “To enable alerts for this website, you need to save it to your home screen⁴. Add this page and enable alerts?” And, naturally, if you remove the saved page, the OS should alert you that you’ll also be disabling notifications. As a concession, I would accept some mechanism for web developers to be able to proactively ask people to add a page to their homescreen — something like the aforementioned app banner could work, or maybe a JavaScript-accesible browser API (and proposed cross-platform standard, of course).

I’m excited about these new features and better web apps on Apple devices. I have high hopes for the next beta!

UPDATE
February 24, 2022

In response to Jeremy Keith’s complaint, Mozilla’s Michael Verdi kind of responded that people do add web pages to their home screen, actually:

As I was reading about this news I saw that the commentary around it repeated an often heard assumption that says, as Jeremy Keith puts it, it’s a “fact that adding a website to the home screen remains such a hidden feature that even power users would be forgiven for not knowing about it.” No one ever seems to cite a study that shows this. I always see this written as if it is indeed a statement of fact. But it just so happens that recently we were testing some prototypes on iOS (unrelated to web apps) and we needed participants to add them their home screens. Of the ten people we talked to, four were familiar with this flow and had saved various things this way. When I mentioned this to others on the UX team a few shared similar stories.

So four of ten people in a user test – what does that tell us? It tells us that it’s something that at least some regular people do and that it’s not a hidden power user feature. More than that, it’s a good reminder to check your assumptions.

Four out of ten! Well ok then!

More seriously, open web advocate Alex Russell wrote a detailed look at the Safari 16.4 preview as an admission that Apple has largely failed compared to their peers. Russell provides ample evidence that Safari has truly been behind for years and it appears the threat of regulation is forcing them to do some catching up.

I’ve had the pleasure of sharing Marcin Wichary’s company on a few occasions, he’s the sort of person it’s impossible to be in a foul mood around. Just a delight.

Since 2016, he’s been working on what he’s referred to as a book, but now looks like something even more wonderful, about keyboards. Yes, the thing you type on, which most of us take completely for granted, Marcin has written over a thousand pages about (two volumes!) with stories, photographs, and history.

This is the only book that connects the world of typewriters to the universe of computers. Whether you’re into vintage typewriters, classic clicky IBM keyboards, or modern mechanical wonders, it will have something for you. None of the above? Get ready to become a keyboard nerd anyway, and look at an everyday boring QWERTY slab with newfound respect

It’s the absolute best kind of obsession. The Kickstarter is coming in February and the book should be published later this year.

UPDATE February 24, 2023

After giving it an earnest go for several more weeks after I wrote this, I’m no longer using Reader as my daily feed reader and switched back to NetNewsWire. It was honestly felt like trading in some interesting but ill-fitting pants for a pair of well-worn jeans. The main issue for me is how Reader pre-fetches the linked content of an item in a feed, rather than displaying the contents. This breaks most link blogs, for example, but also just creates what I find to be a disjointed reading experience. I understand the idea, it might be something you’re looking for if RSS never quite stuck for you.

I’ve been an avid and enthusiastic user of the Readwise app for a while now. The idea behind Readwise is it’s an app built for a (relatively) small but very enthusiastic audience of active readers. It saves highlights from the web, e-books, and paper books (via a very clever and incredibly well implemented photo and OCR feature). The best part about Readwise, though, is that it has a daily review feature that resurfaces your highlights and helps you actually learn and remember what you’ve read.

For a few weeks, I’ve been alpha-testing¹ their new Reader app (now in beta). Reader aims to be a power app for, well, power readers — it offers read-it-later functionality, like Instapaper or Pocket, as well as an RSS reader, PDF support, email forwarding for all those newsletters, and will even show YouTube transcripts inline with videos you save to watch later. Everything you highlight in Reader gets saved to Readwise and there’s (of course) an AI integration for summarizing and helping to clarify that actually adds little bits of delight.

Reader, like Readwise, is aiming for a niche audience of avid readers, researchers, note takers, and deep divers. As with many people over the past few months, I’ve been seriously reconsidering my media environment — I’ve shuttered my twitter account and find Mastodon interesting, but I’m ultimately not all that interested in trying to rebuild social media. Reader has definite potential to be a primary window through which I see the web.

What’s particularly compelling to me about Reader is it’s built entirely on the open internet — web pages, RSS feeds, email. These are all protocols not platforms that Reader sits on top of and adds tremendous value to, value that I happily pay for.

I like this thought experiment from Matt Webb positing that spreadsheets could (and should!) be an alternate interface to apps — a power UI that gives someone a data level view of their apps that are abstracted away by so much conventional interface.

How about an app with a spreadsheet under the hood?

Like, the experience would be this: you’re using your photos app or Zoom or expense filing SaaS tool, then you go to Settings and scroll aaaaall the way to the bottom, and tap a power user button that says “Open Spreadsheet.”

Then, magically, Google Sheets opens with all your data in it, and you can sort and query it in all the ways you couldn’t before, and change the titles of your expense claims with spreadsheet functions and that all gets reflected back into the app, or build a callout to an AI to describe all your photos and add natural language tags, do it yourself or ask a friend who knows Excel functions, or whatever really. Anything the app developers didn’t add because they’re building for the 80% use case, and you’re building just for you

This is a bit of an inversion of what apps like Airtable or Notion or even Retool offer and, as Webb notes, of the way APIs work generally. With something like a Google sheets interface you could remix several apps and even your friends data. I like it!

Just to complete the ouroboros, it seems like it would actually be possible to build this using those various APIs. Like Zappier but with rows and cells. Or even simpler, just a bunch of open source functions in a GitHub repo.