Broken in a browser I never open
6 September 2026
At 01:40 last night someone on a Mac opened the pegboard generator in Firefox, clicked the 3D tab, got nothing, and did the one thing I hope every visitor does. They hit Report a problem and pasted the error message into the box. No email, no name, eleven words of browser error: the specifier "three" was a bare specifier and was not remapped to anything.
I read that over morning coffee and felt the specific chill of a bug that is about to get much bigger than its report.
Wrong in a bigger way, again
A year of writing these generators has produced exactly one law of bug reports, and the vents story was the first proof: the report is always the small end of the problem. That one said "vents aren't being added" and turned out to mean the vents had never worked at all. This one said pegboard, Firefox, 3D tab. It meant every 3D tab, in every one of the 36 generators, in Firefox, since the day each of them shipped.
The mechanism was a two-line difference between browsers. My pages load the 3D library lazily, only when you actually click the 3D tab, and the piece of configuration that tells the browser where "three" lives was being added at that same late moment. Chrome accepts the late registration. Firefox, strictly following the spec, ignores it, and then the import fails with exactly the eleven words from the report. I develop in Chrome. Every test I had ever run was in Chrome. The fix was to declare that mapping statically in the page head, where both browsers respect it, and it went out to all 36 generators the same evening.
Why nobody said anything for weeks
The part that bothers me is not the bug. It is the silence. Firefox users have been on the site the whole time, and the logs show them previewing and downloading like everyone else, because everything except the 3D tab worked perfectly. The 2D previews, the sliders, the files. Someone clicks 3D, sees an empty panel with a terse "failed" note, shrugs, and goes back to the working parts. From the inside, that failure class is invisible: no server error, no crash in my logs, nothing. The page fails politely, in the visitor's browser, in a browser I never open.
One person out of all of them decided the empty panel was worth eleven pasted words. That is the entire difference between a bug that gets fixed overnight and a bug that quietly outlives the site.
The testing hole, named
Verifying the fix had its own comedy. I drive a headless browser through every UI after changes like this, so I pointed a headless Firefox at the patched page, and the 3D tab failed again. Different error though: it now loaded the library fine and died creating a WebGL context, because headless Firefox on my test box has no GPU path at all. Chrome ships a software renderer for this, Firefox does not. So the proof of the fix is the error moving, from "cannot find the library" to "cannot find a graphics card", and the second one only exists on my test rig.
Which names the hole precisely. My automated checks can confirm Firefox loads the code, but the full render path still gets verified by one human in one browser. The next slot of tinkering time goes to wiring the Firefox pass into the same check loop the Chrome one runs in, WebGL or not.
If you are the person with the pegboard hook and the Mac: the 3D view works now, and your eleven words fixed thirty-six pages. The report button does not ask who you are. It just needs the error.