I'm pretty much the same as Cornstalks. I've made my own DX9 engine from scratch but anymore I'm more concerned with getting the game done then endlessly fiddling with all the other bits. I think its kind of a natural progression of programmers. When I first started out I wanted to reinvent everything as a means to get experience but the older I've gotten the less interested I've become in how everything works. Just that it works is good enough for me. Once you get a job programming all day it gets harder to want to do anything outside of work (at least for me) so the more tools I can use to let me get a game done faster is great. Which is why I'm really loving Unity at the moment.

As a kind of a tangent, I would say that anybody who is truly interested in doing game programming should create at least one game engine on their own to see how everything fits together. Whether that is specifically a game engine or core code collected through several projects is up to the person. In retrospect, my game engine wasn't very good but it does give me an understanding why other game engines do some of the things they do.

As they say there is a time and place for everything. Reinventing the wheel is a great way to learn and as Cornstalks and stupid_programmer have said are great for learning how game engines work. However if you're more interested in making a playable game in a reasonable amount of time especially if that amount of time is limited due to work/college/school/etc.

Definitely when I get something working I feel a great sense of achievement even if I did just "reinvent the wheel". For example I rewrote a couple of STL headers like "string.h" and I had great fun and gained a lot of C++ programming knowledge by doing then, even if my friend told me I was completely bonkers.

There is of course a disadvantage to reinventing the wheel in that your solution may not be the most efficient method of completing the task and of course it is more often not the case that existing libraries and code that has been tested through fire and brimstone and used by a large number of people, therefore it is less likely to have any critical bugs or faults. This is a major factor when considering whether or not to use an external library.

My personal policy on "reinventing the wheel" doing so intentionally is best confined to side projects which have the goal of understanding the details of a particular concept. In projects in which actual work is to be accomplished, "reinventing the wheel" is permissible if it is the best way to solve the problem/meet the specifications for the software. Otherwise, it is best avoided; one should always try to see if somebody else has already solved the problem, and to determine whether their solution is applicable to your problem. Note well that this does not mean that one should seek the solution before the problem is fully understood; taking existing solutions to problems and adapting your problem to the solution is exactly backwards; one should adapt solutions to problems, not vice versa.

I'm guilty of trying to "reinvent the wheel".

I'm doing some hobby general a.i. programming, and I know that people have coded the techniques and features I'm coding.

It's hard to continue. People value originality more than rehashes, even if those rehashes are done well.

A lot of people have discouraged me from pursuing my goals.

I want to learn, and if I can I want to make software better. It's easy to forget in the discussion of originality vs. reinventing the wheel that the first wheel isn't always the best wheel. There is always room for innovation. It is hubris for me to think that a novice programmer like myself can do things better, but knowing that doesn't dampen my desire to try.

I don't expect any help. If I do need help, I ask technical questions that have factual answers. I also research what people have done (which anyone who is trying to learn should do anyway) and try to find useful alternative or better ways to do it.

I think everyone needs to reinvent the wheels, up to a certain degree. You need to know how the wheel works before you can use it confidently.

The only time I would reinvent is when I don't need all the extra garbage that others have. For example, I need a linked-list, but find out that the existing linked-list implementations are creating a lot of dummy objects internally. Or I need a simple Python interpreter, but I don't need the full-blown features, just enough to parse basic Python scripts.

I'd reuse others when I know how to make it, but too lazy to do it myself.