I hear you.  

That really sounds like fool's luck [it's just a local saying, I hope you don't take it personally ]. Besides, ATMegas aren't that sensitive, at least not the ones on Arduino look like boards, I often swap +V and -V, no damages yet..

I don't necessarily have disastrous things (everyone makes design mistakes, then start again) or a specific order, but among the things I often do wrong are the enclosure sizing, spaces that are too tight that it is difficult for me to assemble the components. It says you shouldn't waste material, but for the future I need a balance, material used vs. practical assembly.
I didn't really use connectors, or even jumper wires, but I improved the situation by using male+female headers. Besides that, I also found more okay ways to use switches and sockets. I was soldering quite a lot of wires, and it's not always necessary.I don't know how, but I managed to make disassembly harder than assembly because I didn't think too much about subsequent interventions.
And one more thing, more related to software than hardware, I need to create documentation for using various microcontroller boards, I have some that don't work directly through "plug and play", I either need to use another programmer or some configurations are needed in the IDE.
Among many others.

Pretty amazing the variety of "lessons learned" and that we're all human.  We could almost write a book.  What would you call yours?

Lessons from the school of hard shocks.

My worst mistake during a project

One of the worst mistakes I made happened while I was testing a baby monitor with a screen.

Like many people who work with electronics, I’ve accumulated a drawer full of power adapters over the years. Many of them look similar, and nowadays with USB-C everywhere it’s easy to assume that if the connector fits, it will work.

During testing I grabbed a different adapter instead of the original one. The connector fit perfectly, so I didn’t think twice about using it.

Unfortunately, that adapter supplied a higher voltage than the device expected. In my case the adapter delivered a higher voltage than the device was designed for, and since there was no input protection or overvoltage protection, the internal circuitry failed immediately. Being a low-cost device, it didn’t include any protection stage at the power input.

The result was simple: the monitor died and I had to buy another one.

Since then I always check voltage and polarity before powering any device, and I keep chargers organized with the equipment they belong to.

It was a small mistake, but it reminded me how important basic protection circuits are in electronics design, even in low-cost devices.

1500 $, in india you can get a full Ai pc at that cost, Almost a new workbench max upgrade

I paid US$15 to Mistral AI to code a simple software driver for an air quality monitor.

At first, it simply copied some codes from Github and even did not correct their bugs.

Then, when I requested more and more details about my software driver, Mistral AI sent me even more complicated rubbish codes.

I at once cancelled my subscription and stopped paying again.

As to the relevant software driver published on github, there was actually only one mistake on it.

I spent about half hour to find it and correct just by changing a character.  Done!  

Later, I tested my own software driver on the actual hardware and it worked smoothly.

My story teaches us.  AI cannot understand our real world.  GAI is a financial hype. 

This was back when FPGAs with 25,000 Logic elements were quite expensive.  They've come down in price since then.