In the Vietnam War, soldiers called M16A1 "Mattel 16" because of its plastic parts and it being notoriously unreliable.

Though, Eugene Stoner didn't design a bad weapon. M16A1 passed the test phase perfectly, but it was tested by experienced marine soldiers who knew what they were doing. Eugene and Armalite didn't realize that even though the weapon worked reliably for marines didn't mean it would still be reliable in the hands of inexperienced privates.

This is why you should always account for proficiency and experience of your users.

Basically just assume every user is a complete imbecile (which is in most cases true).

If the persona your building for has a brain, and it's not a potato, you've set the bar far too high.

UX101

This is why monkey testing exists

Should I have a capuchin monkey as a tester?

> "plastic parts and it being notoriously unreliable."

And the ammo with the extreme humidity didn't help. I was told on some days, you couldn't empty a magazine without the ammo clogging up the works.

On the range, the M16 is/was deadly accurate at a ridiculous distance. That's why the military picked it. IIRC, by the time they figured out the fix, the new AR variants were already adopted.

I had the opportunity to shoot a M16 once and wow. Very little recoil and with no practice, I held a tight group (fired as fast as I could pull the trigger).

The instructor said smiling "And that's why the insurgents use remote triggered IEDs. Image what a trained Marine can hit at 1,000 yards."

Also let’s not forget the test units had a chrome plated firing chamber. The production ones had bare steel… you can imagine bare steel and humidity didn’t get along great. I guess this is the equivalent of building an app that works great but then running it on a potato server.

@jeeper this is the equivalent of building an architecture based on Hindley Milner types and testing it in Lisp, but then actually implementing it in PHP.