If you've used personal computers long enough, you learn to recognize high level trends. You also learn how to recognize/predict when you're on the wrong side of a very strong trend. Being on the wrong side can be costly. For example, I wrote a large DOS app in 1994 - whoops!

Other examples:
Montezuma's Return, a video game made in the mid/late 90's, was developed primarily under DOS. We had to scramble to port it to Win 95. It was stubbornly written in pure C, which contributed to the code being impenetrable.

On a later game I worked on around 1999-2000, I learned my lesson and developed/tested it under Windows 2000 (while the main devteam used Win98). I realized that the Win95/98 kernel was a dead end and the NT kernel was the future. (Good call!)

On this same game I wrote the software renderer. In 2000 publishers demanded software rendering as a backup to Direct3D, but writing it consumed a lot of our development time. This was a malinvestment, but we had to do it. This was the last time I wrote a commercial SW renderer.

I sense a high-level trend away from C++ and towards memory safe languages for high performance application development. I strongly suspect making a large investment into C++ now is possibly a long term mistake.

Another example of an incredibly strong trend was the rise of the internet in the US in the early/mid 90's. In 1990 BBS's were king. (I remember - I used them!) By 2000 they were almost completely dead.