Question for the Software Engineers, with a rant.

[sparky08]

The golden age was when programmer analysts handled the entire lifecycle themselves. When the first business analyst showed up, followed by dedicated testers, I knew the future was getting less bright, and I was right.
Then came non-technical managers who didn’t really understand the work.

Wow, you nailed it. I was just about to type the same thing.

I've been gone 4.5 years, and nothing has changed.

 

[JoeWras]

Sorry for the side track.

I was a chip designer. I taught a graduate hardware design class once. I told the students don't worry about bugs because everyone will blame the software. I was joking, but it is somewhat true.

I just saw this thread, and sorry to bring up an old post. I literally retired over this issue!

In my last project, the hardware people swore it was my fault. I wrote very precise software that had precision timing requirements. They were convinced my timing was wrong. I stayed on a few extra months to help fix "my" bug and do some other stuff. The time came I had to leave without a fix, and in the last week, the hardware people were starting to say "wait a minute." I knew I might get cleared eventually.

A few months later I ran into a co-worker and he gave the gory details. I won't get into them except to say that the hardware team's tools let them down. It was not a logic problem, but a physical chip construction problem that allowed intermittent leakage. The simulation tool/tester said this should not happen! But it did, creating a once every 500 chip problem.

Ultimately, since the hardware team relied on a simulation tool, which was written in software, it was a software problem. Ha ha. Just not mine! I was right!

 

[Markola]

I wonder if we’ll soon long for the “Good Old Days of Centralized Corporate Enshitification” soon if the SAAS model is truly going the way of the Do Do Bird, once we can all vibe code our own apps on different AI models? If we think we’re dependent on a bewildering jungle now, just wait until it disentigrates.

 

[pjigar]

What most people don't realize is that this is the nature of the problem. When you have n inter-related features, you have potentially up to 2^n (2 to power of n) possible bugs. When you add one more related feature, the possible number of bugs go to 2^(n+1) i.e. twice as many. And marketing loves to add new features. The only solution is to revamp the testing plan every time you add a new feature which is costly.