I've been using Samsung's new Galaxy Fold8 for the last week (More Birchtree members can read my initial impressions here). The reason I got this phone should be no surprise – rumors are very strong that Apple will be releasing a folding phone next month that has very similar dimensions. I wanted to get a feel for it, as well as having more insightful takes than most about how Apple does or does not do things with folding phones better than the competition.

Here's the deal, though: this thing is small. Like really small.

I've put a physical size comparison at the top of this post to show how the screen size compares to Apple's iPhone 17 Pro Max and 13 mini. As you can see pretty plainly, the Fold8 is even shorter than the tiny (by modern standards) iPhone 13 mini. Sure, it's a little wider, but the vast majority of the stuff we look at on our phones are vertically scrolling lists, and that content rarely benefits from a wider screen.

This means that when you use this device as a phone, it is the smallest phone you've probably used in a decade or more. Obviously it functions, and there are certainly some people who like that sized screen, but I think it's fair to say based on sales numbers, that most people prefer a larger display for their phone. For me, what this has meant is that while I can use the outside screen on this device, I don't particularly love it, and I find the experience to be less pleasant than using any normal-sized smartphone from the last ten years.

But the beauty of folding phones is that they can unfold and give you a larger screen when you want it. Unfolding this thing into a nearly 8-inch tablet that has a screen of a quality the iPad mini could only dream of is really cool. However, at that point, you're not using a phone anymore, you're using a tablet, and that's a different experience. Suddenly, it's a two-handed device, for example. This isn't a case where you have a small phone that can turn into a big phone…you have a small phone that can turn into a small tablet.

I'll reiterate now that I do think this is cool. However, I also think that there will be people who find this experience a bit lacking. They'll want a phone experience that feels more like what they're used to, and this passport-style device doesn't really deliver that. Samsung solves for this by simply having a larger version of this phone in their lineup as well. The Fold8 Ultra is a bigger device with a 6.5" outside display that sits in between the sizes of the current iPhone 17 Pro and 17 Pro Max. All the while, it manages to be 0.1mm thicker and 7% lighter than the iPhone 17 Pro Max.

As for my prediction, I think that using this passport-sized phone made it immediately clear to me that this is not a one-size fits all device category, and that a larger model is surely coming in the next couple years from Apple. I believe there will be Apple pundits in the first week of owning this device saying Apple completely nailed it and this is the perfect size for a device like this, and I think they will all eat their words. You may instinctually be reading this and thinking this can't be the case, but I'll point you to just a few Apple products with 2 size options: iPhone Pro, iPad Air, iPad Pro, MacBook Air MacBook Pro, Apple Watch…if Apple's got a hit product, they're gonna make 2 sizes of it.

As a bonus prediction, there seems to be a lot of momentum around the name iPhone Ultra, but I’m a little skeptical of that. I know it’s dangerous to make any predictions about Apple naming things, but I really just don’t think this one makes sense. This phone will be ultra in price, but it will not have the best specs or features in quite a few other ways, and when you add on top of that the expectation that there will be a second larger model in the next year or two, what’s that one gonna be called? The iPhone Ultra Max? If they do, I’m prepping my headline now: iPhones used to be a walled garden, now they’re an ultra max prison.

Quick Subtitles 3 is out now for Mac, iPad, and iPhone. While this is not the biggest update ever, there's some really good stuff here. More than any of my other apps, this is an app for professionals who just want things to work reliably and quickly. This update aims to improve things on both fronts. It is a free update for all paid users.

AI features without a subscription

Since version two, Quick Subtitles has had a cleanup and analyze feature, both of which rely on GPT or Gemini models in the cloud. You are able to use these features by either bringing your own API key and paying for your own usage, or you could pay for a cheap in-app subscription.

Starting in version three, all paid users get to use the cleanup and analyze features at no additional cost. This is powered by GPT 5.6 Luna. There is a cap. You are only allowed to perform 20 AI actions per month, but for most YouTubers and podcasters, that should be sufficient. The limit is there mostly to avoid any sort of abuse situations.

If you'd prefer to use a different model or to have unlimited use, you can still bring your own API keys, and that all works just as well as it always has.

I'm not able to support this yet, but I think that once the OS 27 updates are released, I'll be able to adopt the Apple Foundation model in the cloud. That will allow me to provide unlimited usage as long as that model proves to be good enough.

iCloud sync

This is the summer of Matt adding iCloud Sync to all of his apps, apparently. Now, when you transcribe something on your Mac, your iPhone and iPad will immediately get that transcription. The AI features sync as well.

Another minor thing in this update is that you can drag to reorder the document list on the left, and of course those will sync over iCloud as well.

Mac-assed windows

Like all of my apps, Quick Subtitles is built on SwiftUI, and SwiftUI has interesting ideas on how setting screens should work. They're very clearly built for iPhones and iPads, and you need to go out of your way to make them work even remotely well on the Mac.

As of version three, the Mac app has a completely separate settings window that behaves exactly how you expect a Mac settings window to work. It will never be sized awkwardly or have weird inputs. Cards on the table. I don't think it's a great settings screen yet. I think it's totally cromulent, but hey, you gotta have something for the V4. 😉

The Whisper model starts up quick

Depending on how you use the app, you may have run into situations where you drag in a new file to transcribe with the Whisper model, and it would take a minute or two for the model to get ready to actually start the job. This was caused by a quirk of how macOS stores the model data, and I have since revamped this in a way that should make the Whisper model load in a second or two every time.

This is the low-key hero feature in this update.

A little bit of polish

Listen, I'm not going to do an update without tweaking the UI at least a bit. It's not the main point of this update, but you gotta have a little fun, right?

The drop zone and model picker in the sidebar are now made of a glass material, and your list will elegantly slide beneath it (without killing legibility). All of the empty states have been updated to be a little nicer. The AI features now have more animations that play while they're loading. I've toned down the yellow a bit. Yellow is still the accent color of the app, but it's much more of an accent now rather than dripping over everything.

Check it out

Quick Subtitles is a free update for all existing users. If you haven't purchased the app already, it is a one-time $24.99 purchase to get it for life. The price will be increasing on September 1st, 2026, so if you are looking to get this app, now's the cheapest time to do so.

I joined Mastodon way back in 2016, but I dropped off and didn't come back until late 2022. When I came back, I made a new account and started fresh, building up a new followers list. That number went up quick for a while, but it eventually leveled off. Today I sit at 4,177 followers. I'm not a celebrity or anything, but that's decent – certainly enough of an audience to build a business on.

But how real is that 4,177 number? We all know that subscriber counts are a bit misleading. Obviously, Mastodon doesn't have an algorithmic feed that decides who should and shouldn't see your posts, but it's not immune from the idea of "subscriber rot". Subscriber rot is the idea that while you may have been interesting enough for someone to follow you at one point, they stop being a real set of eyeballs who might see your stuff when you post something. Maybe they muted you, maybe they don't read every single thing and your stuff falls through the cracks, or maybe they just don't use the service at all anymore.

Back to that 4,177 number on Mastodon.

I was curious, so I started digging. I have 4,000 followers, but it doesn't feel like I do. Engagement on my posts is fine, but it seems focused on a smaller number of people. Interestingly, Mastodon has a page in settings where you can see all the accounts you follow and that follow you, and you can filter for a few things. There's a "dormant" account status that seems to trigger when someone does not use Mastodon for over 30 days. I enabled that filter and it returned 2,914 dormant accounts. Of those, 2,244 haven't been active in over a year.

Assuming Mastodon's own reporting is correct, I actually have 1,263 "real" followers, meaning 70% of the people who follow me, don't even use Mastodon anymore. About half of them haven't used it in more than a year.

This example focuses on Mastodon, not because I'm trying to pick on my favorite social network, but because it's where I can get the best data about this. It sounds kind of bad that 70% of my "audience" isn't actually real, so you could see why corporate social companies might not want to so easily raise this sort of data to their users. An open question I cannot answer is whether the 70% number I got here is typical for Mastodon users. Is this just how it goes, or are my follows more likely to be people who migrated off Twitter a few years ago and ended up not liking Mastodon? Also, how does this line up with other services? If someone has 4,000 followers on X, how many of those are probably active accounts? What about YouTube subscribers? Threads users? I don't know.

Where traffic to Birchtree comes from

I run analytics on this blog, and I get pretty darn good data on referral traffic, aka traffic from when my links appear on other websites. I pulled all the referral traffic data for the last 90 days, and charted it out. Here's what I found:

Google Zero may be coming for us all, but it's not here for me yet. About half of my traffic is direct, coming mostly from people who subscribe via RSS. Of the other half of my visitors, 46% of them come from Google, and 42% come from Hacker News. That leaves just 12% of all referral traffic coming from elsewhere. In fact, let's actually remove those 2 from the chart so we can actually see what's left.

A couple things are interesting to me here. First off is the scale of views coming from alternate search engines. Kagi, DuckDuckGo, and Brave Search are all around the same, and combined deliver about 7% of the traffic as Google. Bing, meanwhile, is basically not in the game at all.

Then we get to Mastodon, which is sending the most traffic, which I would hope to be the case. It's the only social network I'm remotely active on, and it's where I put the most effort into promoting my work. Although it must be said that 600 referrals in 90 days works out to about six referrals per day, which ain't much. More notably, it's nearly identical to the traffic Bluesky is sending me, and I'm not active on Bluesky. I have what amounts to a bot account over there that does share links when I post them, but I'm not scrolling the feed or interacting with folks over there. The Facebook number is more perplexing. I guess shout out to anyone using Facebook sharing Birchtree links?

All this said, I just think it's important to remember that subscriber rot is a real thing, and that one should take these numbers with a grain of salt. This is the case on the big corporate social networks, and it's even the case on the indie, zero-algorithm social web as well. Page views, purchases, and actual engagement are the things that truly matter, and I think it's wise to index on those things more than a subscriber number.

A traditional calculator takes numbers and operators as input, does some math, and outputs an answer. LLMs are essentially the same thing, except the input and output are words. You input some words, it does some math on them, and it gives you an answer. They’re tools that do a job. We can debate the implications of this sort of calculator existing, but I think it’s a really useful way to describe them.

I bring this up because the definition of writing, and what constitutes a "perversion of writing", has been…let's just say…a topic of conversation this week. This debate revolves around watermarking text that AI generates. The complaint here is that of the enormous amount of math being done to generate the text models deliver already, adding a specific bit more is an affront to everything.

Quick aside, this video does a good job explaining how watermarking systems work in LLMs if you want to actually understand what's being discussed.

As far as what I think about watermarking text so that software can determine whether it was generated by AI, fuck yeah. I think anything generated with AI, whether it’s images, video, code, or text, should be able to be checked by the user if they want to know.

Arguments have been made about how text is somehow different from images or video, and I think that's incredibly misguided, and shows a lack of respect for other mediums. Sure, writing can be sacred, but nothing these language models generate is sacred. Like number calculators, these word calculators are hugely useful, but no one is going to type something complex into a TI-83 Plus, gasp, and go, "hang it in the fucking Louvre".

The general population has clearly indicated that they want to know when something is AI-generated. I think it is a particularly nihilistic view of the world to suggest that people do not deserve to know this if they want to find out.

At the same time, a lot of people who think this way also love the nanny state and will fight tooth and nail to protect it. Prominent tech pundits will call people "morons" if they think that people should be trusted to install apps outside of the App Store. They love the nanny state when it's corporations doing it, and they hate it when a government does it. Except when it's AI…then it's good when the government does it, bad when corporations do it…it's a lot to keep track of.

Look, I understand there are risks when you let people install software from outside an app store's walled garden. I would actually be okay with Apple requiring App Store distribution as long as they stopped requiring their own payment method and would only block apps for safety and security reasons, not business or content reasons. But these anti-nanny state warriors are tricky to pin down. They want absolute control over how adults use their computers in some cases but not others.

I don't know. How about we tell them that the EU mandates that Apple approve all apps on iPhones? Then suddenly they'll think it's the worst idea in the world.

My industry checks suggest that Apple has indeed scaled back its hardware shipment plans this year due to memory shortages.

Ten years ago, if you wanted to buy the new iPhone and get it on release day, you had to have Apple.com open in three different browsers. You needed to have the Apple Store app open on your phone, and you needed to be frantically refreshing these every couple seconds at the stroke of midnight Pacific time to have any hope of ordering one quick enough. Mercifully, in recent years, Apple has improved their availability significantly, to the point that even though new phones sell significantly more than they did ten years ago, they are readily available on launch day, even if it takes you a few minutes, hours, or in some cases, even days to get your order in.

Well, strap in because I am a bit worried that 2026 iPhone pre-orders will feel a lot more like 2016 iPhone pre-orders.

For all intents and purposes, we believe DeepMind is no longer a frontier lab.

I think this is being a bit overdramatic. It is true that there's been a good amount of executive shakeup and that the company has failed to release any semblance of a frontier model in all of 2026. That's an eon in this industry right now, but Google could come out with something extremely impressive any day now and change the narrative. How often has the narrative shifted over the past couple years of who was winning, who was losing, who was dead?

I'm not arguing that all of the cynicism around Google's delivery of Gemini models this year is wrong. Just that until they throw in the towel like Apple did earlier this year, ironically, by going all in on Gemini, I wouldn't count anyone out. It was barely a month ago, people were talking about how XAI was just going to sell its server capacity going forward, and now they've released Grok 4.6, which appears to be pretty darn good (obviously, Fuck Elon Musk).

Quick Reviews 3 is out now and is an update about giving the people what they've wanted.

iCloud sync

This was going to be in v2, but I never really got it working (hangover from my bad data model choices in the v1 app), so I'm really happy to have it here now. Using the magic of iCloud, your reviews will appear across all your devices, complete with all the color and styling choices you make along the way.

The new cinema layout

I've been sharing this one on my ongoing movie thread on Mastodon for almost a month, and I really like it myself. I've been using a variant of the original design for years now, and this is the first new idea I've had that actually gets me everything I want.

Refined review flow

The review flow is something I think will just keep evolving over time, but I think it's better than ever in v3. I've adopted a new write, style, share model where each section should be very easy to use.

Premium subs get better auto-complete than ever, with live results as you type (movies, shows, and games), and the rest of the flow is a bit easier to use. You can also get to any step in the flow from the review list with a new custom menu.

Review apps

People have been asking me for the ability to review apps for ages, and I'm happy to say that in v3, you finally can! They also get a nice, custom layout that lets app icons sit naturally in the frame.

Premium users also get auto-complete here as well.

Tons of little things

• Unlimited history available for everyone (used to be paid)
• Enhanced year in review screens
• Faster loading with big libraries
• UI flourishes here and there
• More reliable Letterboxd syncing

Quick Reviews is out now and is a free update for existing users. If you'd like better search, Letterboxd sync, and the 2026 "wrapped" feature, consider the reasonably priced premium sub.

iCloud sync

iCloud sync is the headliner, in my opinion, and it's changed how I use my own app. I tend to load episodes into the app on my Mac, and before iCloud Sync, that's where I'd have to do the work. However, now I have the freedom to import the file on my Mac and then listen to the episode on my iPhone as I do a workout and add chapters where they need to go.

From there, when I get back to my Mac later, the chapters are already there, and all of this works in every direction with every feature in the app.

Enhance Audio (Mac-only)

This one's kind of huge for me. Podcaster who records at exactly the right level and edits to exactly the right audio level. But not everybody does that, including myself.

And the Mac version of Chapterize now has a one-click enhance audio button that performs two actions. First, it normalizes the audio to the expected volume for podcasts, and it reduces noise. It's not doing anything crazy with AI here. It's just using traditional tooling to get this done. If your podcast already sounds perfect, you don't need this. But if you struggle a bit, this might be a one-click fix similar to previous tools like The Levelator.

Listen to your shows

The new shows tab in version 2.0 lets you follow your shows, and get information about them. Now you can also listen to them. This is absolutely not a full podcast app in the vein of Overcast or Apple Podcasts, but if you need to listen back to part of a previous episode, this will get the job done.

Better on iPhone

Once I had iCloud Sync working, it made me work a bit more from the iPhone, and this made me realize how tough the experience was there. One of the things I've tried to do with this app is have a consistent experience across all platforms, and I think the ideal layout for working on the desktop (including iPad) is not the same as the iPhone.

As such, I've given the editing view on the iPhone a bespoke layout that works a lot better. No features are removed, but they're all more well considered for the iPhone's relatively tiny screen.

And the rest

• Give any saved show its own accent color (in the "info" button on the Shows page)
• On the Mac, Settings is now a proper preferences window
• Scrolling a long chapter list is noticeably smoother
• Fixed a bug that could make chapters disappear from an episode, and a sync issue that could leave an episode with its chapters and transcript but no audio

If you make a podcast and you're not using Chapterize yet, this is a great time to start. It's free to try on Mac, iPad, and iPhone.