2026-08-10 22:39:53
TL;DR: CaptionSnap is a $9.99 one-time menu bar app that saves your Teams, Zoom, and Google Meet transcripts by reading the live captions the meeting app already displays. No audio recording, no bot in the call, no cloud, no model download, and Accessibility is the only permission it asks for. Speaker names come from the platform, so attribution is a lookup instead of a guess. You have to enable captions yourself, transcripts are only as good as the platform's captioning, and it requires macOS 26 on Apple Silicon.
Here is a list of the many things I don't like about transcription apps.
Well, I found an app that doesn't do any of those things and still delivers better-than-average accuracy while using a tiny percentage of my CPU. It's called CaptionSnap.
Teams, Zoom, and Google Meet already transcribe your meeting. That's what live captions are. The text is right there on screen, generated by the platform, attached to the correct speaker, and then thrown away the second the call ends.
Every other tool in this category ignores that and starts over from scratch. Record the audio, run it through a model, guess at who said what, hope the diarization holds up. CaptionSnap just reads the captions the meeting app is already displaying and writes them to a Markdown file with timestamps and the speaker names the platform supplied.
That one decision is the premise for the whole app. It's why there's no model download, no audio pipeline, no microphone prompt, and no bot in your call. There's nothing to overcharge you for.
Real speaker names, not guesses. The names come from the meeting platform, so "who said that" is a lookup, not an inference. Whisper-based tools get diarization wrong constantly on four-person calls with crosstalk. This sidesteps the problem instead of solving it.
One permission. Accessibility. That's it. No Screen Recording, no microphone, no camera (CaptionSnap FAQ). If you've ever watched a meeting app collect four permissions on the way in, you know why that matters.
Minimal CPU. Not a tuning achievement. There's no neural network running, so there's nothing to burn cycles on. Your fans stay quiet and your battery survives a day of calls.
Plain Markdown on your disk. No account, no sync, no server, no export step. The developer claims the app makes no network connections at all (CaptionSnap homepage). Optional summaries run through Apple Intelligence on-device.
It works behind other windows. The meeting can sit in the background while you do actual work, and the transcript keeps filling in.
CaptionSnap has to be running in your menu bar when the meeting starts. It does not turn on captions in Teams, Google Meet, or Zoom for you, so that's your job too. CaptionSnap provides instructions for each platform. As long as you don't disable captions later, the setting should hold between restarts of each app, but verify it yourself. The worst version of this is finding out after the meeting that nothing was captured.
This is for you if you take meeting notes on a Mac, you don't want a bot announcing itself in a client call, and you'd rather have a plain text file than a web dashboard. If you aren't fond of explaining to IT why you need an app with Screen Recording permission, CaptionSnap doesn't put you in that conversation.
This is not for you if you need the audio and not just the text, or if you work with people who never enable captions. Don't get it if you're on Intel or an older macOS, or if you want AI-generated action items and CRM sync. For that last one, this is deliberately the wrong tool and the subscription products below are the right ones.
macOS Live Captions. Free, built in. Vanishes when you dismiss it, no speaker names, no timestamps, nothing saved. Useful in the moment, useless a day later.
MacWhisper. Free tier, or €64 one-time for Pro (macwhisper.com). Transcribes local audio with Whisper and Parakeet, records Zoom, Teams, Webex, Skype, and Discord without a bot, and Pro includes automatic meeting start and end detection plus speaker recognition. Much larger tool, much larger job. It does the model download and the CPU work that CaptionSnap avoids, and in exchange it also handles interviews, lectures, podcasts, and system-wide dictation. If you need transcription generally and not just meetings, this is the one. I wrote about it in 2024: MacWhisper - Transcription Powerhouse.
Granola. Free Basic tier with limited history, $14 per user per month for Business, $35 for Enterprise (granola.ai/pricing). Also avoids joining as a bot, but it processes in the cloud and is built around AI-enhanced notes and team integrations rather than a raw transcript you own.
Otter.ai. Free tier at 300 minutes per month, Pro at $16.99 per user per month billed monthly or $8.33 billed annually, Business at $30 monthly or $19.99 annually (otter.ai/pricing). This is the thing CaptionSnap is reacting to: it joins your Zoom, Teams, and Meet calls as a notetaker, and everything lives on Otter's servers. I've used Otter for dictation, but not for meetings.
Unless otherwise noted, the alternate apps mentioned above are based on research and not individual testing.
Fourteen-day trial, no credit card, then $9.99 once for up to three Macs, with a 30-day refund policy (CaptionSnap pricing). Ten dollars, paid one time, against $14 to $17 a month for the subscription notetakers.
This is Wes Stockfisch's first shipped app. I give developers the benefit of the doubt, but be aware, there's no maintenance track record here, and the macOS 26 plus Apple Silicon requirement means you need a current M-series machine. But the idea is right. The captions already exist. Reading them instead of recreating them is the kind of obvious move nobody made because, well, they mnight have been overthinking it on this..
One caveat before the verdict: I haven't been able to test this extensively. Retirees don't have a lot of Zoom calls. But in the testing I was able to do, CaptionSnap held up.
In my opinion, ten dollars to stop explaining a bot to clients is not a hard decision.
2026-08-09 17:04:48
TL;DR: AutoShelf ($19.99 one-time) ships an embedded MCP server that works, but do you actually need that? An AI model can write a launchd LaunchAgent and a shell script once and it runs forever, no app, no daemon, no purchase. The MCP's real value only shows up once you're managing enough rules to want conversational audit and editing instead of hand-maintained scripts, and even there, my hands-on testing found the rule engine underneath thin: no dynamic subfolder sorting, limited management by file size, and an action list a fraction of Hazel's. AutoShelf delivers less using AI than its marketing promises and loses the file-organizer argument outright.
I've had a Hazel rule watching my Downloads folder since before I can accurately date it. It reads PDF invoices and files them by vendor, decompresses ZIPs and moves the app inside to Applications, converts HEICs to JPG, and hands ebooks to Calibre and audiobooks to Audiobookshelf without me touching any of it. That's the bar. A new license for Hazel is $42, which is steep for some and I've written before about cheaper options. AutoShelf wants to be added to that list, and it's making its case with a feature none of those alternatives had: an MCP server for use with AI agents plus a CLI.
Contents attribute has done on-the-fly text recognition on images and text-layer-less PDFs for years, it's the mechanism behind the vendor-filing rule in the intro. AutoShelf catching up to a feature Hazel already had is parity, not an edge.Does anyone even need a layer between a prompt for file management and their file system? For a well-defined one-off job, not really. A capable AI model can write a launchd LaunchAgent with a WatchPaths key, native FSEvents, no polling loop, and a small shell script that switches on extension and moves the file. Write it once and it runs forever: no app, no background daemon, no $19.99. That's possible without Hazel or AutoShelf.
An MCP connection can do something a bespoke script can't: with it running, AutoShelf can immediately edit any rule with natural language, without you having to track down the launchd job and script you created six months ago. I created eight AutoShelf rules and asked Claude to check on them through the MCP connection; it returned current rules, current paths, and current enabled/disabled flags. A launchd script has no equivalent self-report; ask it "what are you doing" six months from now and the honest answer is "go read the shell and simulate it," and any edit needs a launchctl unload/load before it's live. MCP matters once you're managing enough rules that you'd rather audit, pause, or edit them by asking than by maintaining scripts by hand, or you want the GUI, activity log, and undo the app provides that a bare launchd job doesn't have unless you build it yourself.
Against Hazel specifically, AutoShelf is still ahead on this one front. Hazel doesn't have an MCP server and doesn't claim to; anything an AI does to a Hazel rule today runs through a general-purpose file-editing layer poking at Hazel's XML underneath, which is fragile and unsupported. So among app-based rule engines with a GUI and an activity log, AutoShelf's MCP integration is real. On my Mac, it was easy to set up and worked as advertised on the first try.
This is where my hands-on testing exposed the weaknesses in AutoShelf. It can't sort a folder of mixed files into subfolders by extension unless you create every destination subfolder yourself first and then write a separate rule for each file type. I had to build five nearly identical rules, "Sort PNG files," "Sort JPG files," "Sort JPEG files," "Sort GIF files," "Sort MD files," each one hand-built with its own hardcoded destination path. Hazel does the same job with a single "Sort into subfolder" action and a pattern; AutoShelf makes you write and maintain N rules for N extensions.
Two more gaps from my testing: size conditions only go down to whole megabytes, no decimal places, so you can't write a rule for anything under 1 MB or between whole-MB steps. More importantly, AutoShelf's action list (Move, Copy, Trash, Rename, Add Tag, Archive to ZIP, Optimize, Convert Image/Video/Audio, Photos import) is a fraction of Hazel's, which adds sync, upload, tags, color labels, comments, alias creation, three Music/Photos/TV import targets, and five different scripting hooks (Shortcuts, AppleScript, JavaScript, Automator, shell) on top of pause, notify, and rule-chaining controls.
AutoShelf's App Store listing says it will "watch any folder you choose, including nested subfolders," and the 1.5 changelog calls it "Nested Folder Watching." (App Store listing) But on its own Product Hunt launch, the developer told a commenter directly that watched folders are top-level only, files dropped into a subfolder are not picked up, and true recursion is a tracked feature request, not a shipped one. (Product Hunt) Read charitably, "nested folder watching" means you can manually add specific subfolders you already know about as additional watch targets, one at a time. That is not the same as what Hazel does, which is recursive: add one "Run rules on folder contents" rule and Hazel will walk into subfolders of subfolders it has never seen, at any depth, applying your existing rules the whole way down. (Noodlesoft manual) That's a real discrepancy between AutoShelf's marketing copy and its confirmed current behavior, and now my own rule list shows the practical cost of it: I'm hand-maintaining what Hazel would do with one rule.
If your files land flat, one folder, no meaningful variety in file type, and you are considering a $19.99 one-time tool with a working AI hook, AutoShelf's free tier is worth an afternoon. If your needs are more complex, stick with Hazel. My own eight-rule test folder is the case study: if you have more than three or four file types, or anywhere the folder structure has depth, AutoShelf turns into more manual rule-building than the automation it's supposed to replace.
2026-08-08 19:32:00
My parents were young when I was born, only 17, but despite the relatively small time gap, we are from distinctly different generations. They've been divorced since the 70s, so their needs and circumstances, while similar, are still unique. Last year, my dad and his wife moved to an assisted living facility about a 20-minute drive away. It takes about three hours to get to my mom's. That's doable, but not as easy.
For most of my life, each has been independent. They are both part of the generation that almost understands cell phones, but not quite. My career was in IT, so I have long been their go-to for tech support and purchase advice. Other than that, they never needed much help. Until they did. In both cases, the transition happened suddenly.
Last year, my dad fell late one night, breaking his shoulder and hitting his head - hard. My stepmother has advanced Alzheimer's. She was asleep, but heard him calling for her. Luckily, Dad was able to coach her through retrieving his Life Alert device to summon help. When the paramedics arrived, they made note of a half-empty bottle of bourbon sitting on the counter.
He's never recovered from that night. The damage to his shoulder is irreparable. The concussion and his relationship with alcohol have impaired his short-term memory and decision-making skills.
My mother, an independent widow, has been remarkably fit for years. Since turning 70, she's walked across England and Spain - and I mean all the way across. She lives by herself on an 11-acre tract of land she's long maintained without outside help. Then, suddenly this spring, a hidden back injury surfaced and nearly incapacitated her overnight. She went through weeks of intense pain, lost the ability to drive, and couldn't stand long enough to prepare a meal. Her situation has improved somewhat, but some of the physical impairments might be permanent. Mentally, she is sharp as ever, highly opinionated with some endearing quirks, but she's all there and then some.
I wasn't prepared for either parent's transition. Whenever I've known someone who was providing aid for their mom or dad, my thought pattern was just "That's tough." I never really considered getting mentally ready to step up. The thing is, the universe doesn't care if you are ready or not. When the time comes, unless you are a terrible human, you just start doing what it takes.
Sure, there are books and websites that purport to help you get ready for your new responsibilities, but they aren't clairvoyant, so they don't know what's going to be the event that flips the switch from independence to "needs help." You won't know either.
In my case, the energy it takes to do what needs to be done is disconnected from my feelings. My relationship with each of my parents is and has been very different.
I am much closer to my mom. She's been my supportive parent forever, a wonderful grandmother and great-grandmother, generous with her time and energy. She's smart and interesting and genuine.
My dad is difficult. He's also smart, but he's been absent from my life in most meaningful ways. He leans towards being judgmental, and his one overriding filter for examining everything in the world is how it affects him. He's not interested if it doesn't.
Again, the universe doesn't care. Their needs are unrelated to their degree of likability. They need what they need from you, and the truth from where I'm standing is simple. You do what needs to be done and you use whatever you have to for motivation. My mental gymnastics work like this:
I realize that this instruction manual is nothing of the sort, but it's my real, lived experience. There's no manual. There's just the drive to Dad's, and the drive to Mom's, and the fact that I make both of them.
Enjoyed it? Please upvote 👇
2026-08-07 19:49:53
TL;DR: Part two of the Apple Account series. Backup apps that pull your iCloud data down to storage you control, email providers that speak IMAP/CalDAV/CardDAV so Apple's own apps keep working without iCloud, ways to buy and track software outside the App Store, and a password manager that isn't Keychain. My own five picks are at the end, along with why you should be testing a restore every quarter.
Yesterday I laid out what happens when Apple closes your account: your identity if it's an @icloud.com address, your photos and messages, your purchases, your hardware in every practical sense, and a 2023 federal ruling that says suspicion is the only standard Apple has to meet. That post was the bad news.
This is the other half. These are the apps and services that move your data out from under a single Apple Account, what each one actually protects, and where each one falls short. There are more options here than anyone needs, and I have not run most of them. At the end I'll tell you the five I actually use and why I stopped looking after those.
You can keep using Apple's Mail, Calendar and Contacts without tying any of the data to your Apple Account, by pointing them at a provider that supports IMAP, CardDAV and CalDAV. Reminders is the exception; Apple moved it to a private format, though Things 3 can pull to-dos out of a designated Reminders list into its own inbox.
Five providers with full IMAP/CardDAV/CalDAV support on macOS:
All five work natively with macOS Calendar and Contacts. Fastmail has the smoothest CalDAV experience of the ones I've touched. Posteo is the cheapest. Mailfence leans hardest on open standards.
Whenever you have the choice, download applications directly from the developer rather than from the App Store, so your access survives your Apple Account. You give up automatic updates, Family Sharing, and in a few cases unlimited installs.
Updaters that close the first gap:
Then track the licenses somewhere you can export from:
Keep your passwords, recovery codes, server credentials and export encryption passwords outside iCloud Keychain. Then protect the new vault's recovery material somewhere else again, or you've just relocated your single point of failure.
Unless I've said otherwise above, the apps in this post come from research rather than daily use. The five below are the exception.
Five, and the list has been stable long enough that I've stopped shopping.
BackiGo pulls my iCloud Photos and Drive down to local storage on a schedule. It's the only one on this list that gets irreplaceable data out of Apple's custody.
Fastmail holds my mail, calendars and contacts on a domain I own, and Apple's own Mail, Calendar and Contacts talk to it over IMAP, CalDAV and CardDAV without knowing the difference. Two years in. If Apple closed my account tomorrow, my address would still work, which was the whole point.
Updatest keeps the direct-download side of my Mac current. Buying from developers instead of the App Store costs you automatic updates, and this is what I bought to pay that cost back.
License Store holds the keys and invoices for everything I didn't buy through Apple. The JSON export goes out on a schedule and lands with my other backups, per the caveat above.
Bitwarden keeps my passwords and recovery codes out of iCloud Keychain. The free tier does everything I need and I pay for Premium anyway, because a project I depend on this much should get some money from me.
None of this counts until you've restored from it. A backup you have never pulled a file out of is not a backup; it's a hypothesis with a progress bar. Once a quarter, pick something you'd hate to lose, a photo from ten years ago, a document with a deadline attached to it, and restore it to a machine that isn't signed in to the account you're insuring against. Open it. Confirm it's the full-resolution original and not a thumbnail, that the metadata came along, that the encryption password you set eighteen months ago still unlocks the archive. The failures I've seen were never missing backups. They were backups that ran green for two years and turned out to be empty, stale, or locked with a password nobody wrote down.
2026-08-07 02:22:06
TL;DR: Mac power users vet apps carefully and then hand their entire digital life to Apple without a second thought. Apple can close your account on suspicion alone, keep your unspent balance, and never tell you why; a federal judge confirmed all of that in 2023 and threw the case out. An Apple Account is simultaneously an authentication system, cloud-storage account, software license repository, device-activation credential, communications identity, and payment account. A full lockout can therefore affect data, purchases, services, and hardware management at once.This post covers how accounts get closed, what you lose when it happens, why getting back in is so hard, and eleven things to do about it now.
When the conversation turns to abandoning third party apps in favor of using Apple's default stack, I keep waiting for someone to mention the huge red flag that should instantly discourage making that move without taking some intense protective measures.
The steps I take before adding an app to my workflow are second nature. but detailed. They are probably the same steps most of you take. First I evaluate the feature set to see if it meets my needs. I look at the privacy policy, the cost of buying the app, and the cost of keeping it updated. Any app that takes a significant investment of money or time gets some research about the developer or the company behind it. Sometimes a one-person development team means the app dies when the dev loses interest. Finally I look at what third-party services the app depends on. Dropping $90 on an app Google can kill with an abitrary decsion isn't a deal breaker every time. Just most of the time.
Until last December, I did not apply the same scrutiny to the infrastructure underneath all of it, the company behind much of my digital life. I'm talking about Apple. An entire career in Mac support didn't give me a full understanding of what losing access to an Apple Account would do to someone personally and professionally.
Then Paris Buttfield-Addison, who has written more than twenty books on Apple development and helps put on the longest-running Apple developer conference Apple doesn't run itself, found his account closed for reasons Apple would not specify. After some of the biggest names in Mac journalism publicized the case, his account was restored, but Apple has refused to go on the record with any details about why it happened or, more importantly, what you and I could do to keep it from happening to us.
It's more common than you'd think.
None of this requires Apple to be right. In 2023 a federal judge dismissed a class action over exactly this question, and the reasoning matters more than the outcome. Matthew Price ran multiple chargebacks on in-app purchases that didn't work, some of them after Apple's own support told him to take it up with his bank. Apple terminated his account, kept the $7.63 balance sitting in it, and cut him off from what he says was $24,590.05 in purchased content. Judge Haywood Gilliam dismissed the case without leave to amend: "It is simply immaterial whether Plaintiff's multiple chargebacks were actually fraudulent or manipulative because even as alleged, Apple suspected that they were." Suspicion is the standard, and you agreed to it when you accepted the terms. The same terms say you "shall not sue or recover any damages from Apple" over a decision to terminate your access.
Apple won't tell you why your account was closed, reportedly out of fear that its fraud prevention methods will be reverse engineered. Frontline support staff don't have access to the reason either, so they can't find out on your behalf and they can't escalate your issue. Making a new account on hardware already linked to suspected fraud can earn you another ban for attempting to circumvent fraud prevention. I could find no documented path to export data from a closed account.
Tomorrow I'll post the apps and methods that let you keep using Mail, Calendar, Photos, Contacts and file storage without putting all of it inside one account Apple can close on suspicion.
2026-08-04 18:47:51
"Radar Love" by Golden Earring
What song makes you want to take a midnight drive? - Radar Love by Golden Earring
This has got to be the classic answer for a reason: the bassline mimics tires eating up highway miles, and the lyrics are literally about driving through the night to a lover,
