#DailyBloggingChallenge (31/50)
I have been using #chatgpt for a couple months as a developers' tool. Although originally against the concept of such tools, I quickly realized that it will become an integral part of developing in the future.
The good thing about the tool is that it saves a lot of internetting though only if you already know what you are looking for. (As a beginner, video tutorials are still the better option.)
I have built a static website that was 80% generated from the tool, or compute data via #R in no time, or write #bash as if I was a pro.
Though on the flip side, I didn't learn much. I made one key mistake and that is that I #copypasta the code. If you are an expert in a field, then this is fine, but being a novice, this removes a concrete learning route. (This also falls inline with the reasoning behind this [challenge](https://qoto.org/@barefootstache/110743952822270245)).
Another downside is that the tool is many times quite dumb. You ask it to change one thing and all it does is change the intro but spit out the same result again.
Another factor is how up-to-date is the tool. While using ChatGPTv3 one quickly learns its barriers when asking about the latest versions of libraries, which don't exist in its dataset.
In the end, AI is here to stay, either you can learn how to use the technology or not.
#DailyBloggingChallenge (35/50)
In #programming there is this consensus of don't touch a working process. That means that any alterations done to the process were deemed necessary because something of the process was not quite right, so it was broken.
Now this could be obviously things like not getting the expected result or the time needed to get the result is just taking to long or less obvious things like the code is ugly or it is developer unfriendly for debugging.
These broken changes is what makes oneself a better programmer in the long run and thus I am glad that I started my #bash programming experience initially without #chatgpt.
For example, initially JSON files were built line by line instead of via jq.
#DailyBloggingChallenge (36/50)
#TIL that parsing #RegEx via `sed` and pattern matching two different syntax variants are in use. The former requires a lot of escape character where as the latter doesn't.
Take for example
```
\([0-9]\{2\}\-\)\+\([0-9]\{2\}\)
```
which is the correct syntax when using `sed`, where as it won't work with pattern matching over `=~`. In such a case it would be
```
([0-9]{2}-)+([0-9]{2})
```
This cost me so much debugging fun.
#bash