When the mind wanders during routine tasks, it doesn’t wander randomly. It goes looking for something. I spent a stretch of late June doing the unglamorous housekeeping that comes before anything gets to call itself finished — a license file, a corrected README, a landing page that describes what a tool actually does rather than what it once did. The kind of work that asks almost nothing of you, which is exactly why the mind slips its leash while your hands are still busy.

Mine went looking for competence, and it found it in two very different places at once.

The first was an argument. WWDC26 had just happened, and something about it — all the AI, none of the follow-through — kept pulling at a thread until the thread turned out to be thirty years long. Objective-C++ solved C++ interoperability cleanly in the 1990s; Swift, eight years after promising the same thing, still can’t. Cocoa gave developers a working gallery view for free; SwiftUI shipped without one for two years and still asks you to build it yourself. A researcher’s on-device photo classifier proved, in 2017, that private semantic AI could work at real scale — and nobody ever carried it anywhere else in the system. None of this is a story about missing technology. It’s a story about a company that kept arriving at the right answer and then declining to keep it. I wrote all of it down, evidenced, sourced, addressed to people who could actually do something about it. Compliance over competence, stated plainly, three words holding up a very long letter.

The second was a memory, and it turned out to be the same argument told from the other side.

Twenty years ago I ran a microscopy core, and the job was never really about my own science — it was making sure the researcher in front of me got the most out of whatever confocal time they had. Somewhere in that work I built a 3D reconstruction for a colleague who didn’t even want the movie submitted; a journal ended up putting frames from it on the cover anyway, compressed and slightly worse than it should have been, because nobody warned me in advance it needed to survive print. I built shading and contours out of nothing but an edge-detection filter and an inverted projection along one axis — tricks so convincing that people didn’t believe you could drag a threshold slider live and watch a structure breathe under your own hand. A four-dimensional reconstruction of a developing zebrafish pituitary never made it into print at all, because the PI didn’t want movies in his papers, and it still sits in an archive today as a kind of ghost — technically excellent, publicly invisible. And sometimes none of the trickery was needed: a researcher brought me cells colonizing a surgical sponge, and the raw geometry of the fibers composed the image entirely on its own.

I don’t think these two things — the letter and the memory — happened to arrive in the same week by accident. The letter exists because I know, with total certainty, what real competence looks like when an organization commits to it: a bug report answered so generously the fix exceeded what was asked; a piece of software good enough that people refused to believe it was live. I only recognized Apple’s failure as clearly as I did because I’d already lived the alternative, on the other side of the same kind of work, in a room with a confocal microscope and a researcher who needed the data more than I needed the credit.

That’s the part worth keeping, more than the report of what got built and what got argued. The past isn’t a place you visit when the present gets frustrating. It’s the only reliable instrument you have for measuring the present at all. I know Apple is capable of more because I’ve built things, with far fewer resources, that were more generous than what a trillion-dollar company now considers finished. And the same standard runs the other direction too — elaritysystems.com exists, in its small and unglamorous way, because I still believe a tool is only really done when someone who’s never met you can pick it up and understand it without asking. That’s not a new idea. It’s the same one that made a compressed movie frame worth putting on a cover, and a live threshold slider worth not believing.

You don’t get to build a good future by forgetting what good used to feel like. You get there by refusing to.