About this episode
Daniel Jalkut returns to the show so we can both vent about MacOS 26 Tahoe.
Listen to the original episode
Episode summary
Daniel, good to see you. We’re both buried in snow, the city’s gone quiet, and I kind of love when everything slows down.
Same here—topped off the gas for the snowblower and took some photos. It’s surreal out, but the pause is nice.
All right, are you ready for my Tahoe gripes? I worry people are tired of me railing on macOS 26.
It’s wild how intense the backlash is—icons, visual changes—like it shipped last week.
I used to critique the WWDC betas, then pulled back after Apple folks said it was too early; by fall the takes felt stale and I missed my window to complain.
But that’s the point of betas—you need public feedback. Big outcry reversed those weird Safari tabs a few years ago; sometimes you need a collective push.
Liquid Glass clearly started on iPhone and matured by 26.1/26.2 there, but on Mac we assumed they’d fix rough edges before release—and then they didn’t.
It’s death by paper cuts. The app switcher label often isn’t under the selected app; that’s the kind of thing that should be fixed instantly.
Long app names like Google Chrome overlap the next icon, and the selection highlight is so faint—on iPad it’s basically invisible. A transient UI needs a bold selection, not a squint test.
I’m calling Tahoe a mess of annoyances, not catastrophic, but even the “it’s fine” crowd admits it’s not Apple-level. Norbert Heger’s post on window-resize targets landed because everyone feels it.
I missed that bug until I read it, which says a lot. I’d call Tahoe a mitigated disaster—I obsess over shipping polish to users more than demanding it for myself.
Retired Apple folks tell me to keep pushing. Mid-2010s Mac UIs still look current; today’s declines aren’t just taste—they’re real regressions.
Seeing mis-indented help text and mismatched controls all over is embarrassing. Liquid Glass isn’t even consistently adopted across Apple’s own apps, which makes the OS feel incoherent.
There was a time when “to the pixel” meant exactly that; one-pixel drifts got fixed before ship.
Fewer but better only works if the finish is immaculate. I’d trade features for that level of craft any day.
Design is deciding for the user. Apple Silicon’s hardware triumph masked a slow UI slide, and Tahoe ripped the cover off.
We rebounded from the keyboard saga; maybe this is the software valley before the climb. Leadership changes give me hope.
With Alan Dye gone and Steve LeMay in, optimism is warranted. The real test is whether pride blocks fixes, like restoring visible, rectangular menu highlights that read as menus again.
A new design head can “refine” without confessing sins. Unify the system, make the hard calls, and move forward.
If Jon Ternus is the heir, it fits: hardware is Apple’s strongest pillar and software engineering is stable—my Mac’s been up for over seventy days without a hiccup.
Back in the day, the PowerBook control strip was a rogue hardware-led interface win before it spread platform-wide.
Phil Schiller prototyped it on a flight; Jobs’s taste-centered structure killed fiefdom spats and enforced coherence.
At Apple, taste was hired, not standardized. Interviewing was decentralized, but craft was in the water.
Indie Mac shops lived for nitpicks; a one-pixel bug could cost sleep, and that zeal defined “Mac-like.”
QA trained me to file every flaw. I stop filing only when a developer signals, “who cares,” which tells me to walk away.
Critique the work, not the person. I’m grateful when readers flag ambiguous sentences or typos—it’s a gift, not an attack.
On icons in menus: plastering symbols everywhere erases meaning. The HIG used to warn against this, and even Apple’s apps disagree on which icon fits which command.
Safari slaps symbols on almost everything, even generic globes for sites without favicons. It adds visual noise without clarity.
Icons help when they truly disambiguate—window tiling or photo rotation—but if every line gets a symbol, nothing stands out, and tiny magnifying-glass glyphs are unreadable.
Think restaurant menus: if every dish has an icon, the vegetarian mark gets lost in the clutter.
You said AI let you ship a MarsEdit feature you’d never have tackled otherwise. What did you do?
WordPress’s XML-RPC muddles auth and permission errors across dozens of languages, so I used Claude and GPT to extract all localized error strings, map them back to English, and detect when to prompt for a password. AI wrote the scripts, I compressed the data from a megabyte to kilobytes, and what would’ve taken days took a couple of hours.
I had a mini-version of that: asked for a bookmarklet to number paragraphs in Safari and add a click rebase. It worked on the first try.
The trick is giving it a chance because sometimes it really is that easy. Debates about energy and hype aside, the productivity boost is real.
Like spreadsheets in the eighties, it’s overhyped yet still transformative; once you change one thing and the whole sheet updates, you never go back.
Shout-out to Brent Simmons and Paul Kafasis for shipping Tahoe builds without the menu icon clutter and documenting how to do it.
And about Core Intuition “ending”—you’re a lovable liar. You versioned the show and pop in now and then.
Guilty. We said goodbye but kept the feed alive, so the occasional episode still sneaks out.
That’s it from me. If you want more, check out Core Intuition; in whatever app you’re using for this show, search for Core Intuition and you’ll see the new episode on sharing numbers that Daniel brought up. Before we wrap, where should folks follow you?
That pretty much covers it, including my nostalgia rabbit holes. I post first on micro.blog thanks to Manton Reece, and it mirrors to Bluesky. I also share on Mastodon and Threads. I stepped away from Threads for a bit over Meta’s philosophy, but I’m back because different friends have picked different homes.
When Twitter worked, one place felt ideal, but once it went sideways there was no return. Now there are several Twitter-like spots. The good news is resilience if one turns bad; the tradeoff is hopping around because everyone’s scattered. It’s probably healthier overall. People can find you across those networks, and your last name is easy to search, so they’ll catch you there.