Thinkings after being able to code more with AI support

Another two months have past since my writing of Practical AI usages and my understanding of its current future and I guess I already get much used to coding with AI support. The following are some of my thinkings that popped up during this time period.

After getting comfort to code with AI support

Being able to collaborate well with AI (some personal experiences will be explained later) makes me think that all existing things (websites, apps of different platforms, frameworks, libraries, etc,) can be upgraded, especially those that are relatively hard to use and have collected a lot of historical issues after being created, since AI makes refactoring code much easier than before. Tech docs can be written much better than before as well or at least having instruction docs is much easier to fulfill and the only exception might be the team doesn't have the motivation or is just lazy to do so. Even building totally new versions of existing technologies are finally very achievable and potentially profitable.

The age of giving computers back to computers

By seeing (much) more power of what AI can (help) do, I feel greedy to jump from 100% manual coding to 100% smoothly workable AI coding. Imagine recent days you are under great pressure from a company project at hand and suddenly AI can give you all the workable code you need. How do you feel? But at the same time, it's being said that programmers feel hard to get the flow state of coding (many discussions can be found online). Do you use 'Cursor' and review generated code as much as you can or already change to 'Claude' for more and more blackbox coding? The whole situation can be very confusing and struggling sometimes.

We're probably and gradually stepping into the age of giving computers back to computers instead of manually writing code to tell computers what to do like before. However, the way not to touch any code but let AI do everything is actually more a thing of human will versus AI will, meaning the mental struggle is 1000% normal since you are sacrificing and giving (some of) you/your will to AI, which might lead to the feeling of losing meanings at least from a programmer's view. Even not only about coding, this is also true for situations that cover more about designing and ideas: are you still letting your ideas support the core foundations of an app or using more and more AI for this part as well?

Again on the other side, sometimes you do not want to touch SQL or people just don't have that much energy to learn or care everything, a 100% AI blackbox of workable SQL changes might be better and tempting or you still learn and check each line of new code from scratch? If AI develops fast enough, my current efforts of knowing and collaborating well with AI itself will just be in vain and wasting time, meaning finally AI will just take control of coding and I do not need to learn and know any ins and outs of real coding :(

Finally, AI won't fell frustrated, so it'll keep trying to solve issues in different ways and then people facing these issues won't get emotional ups and downs like before but 'good' results from AI, which might support: 1) AI might 100% beat human; 2) great collaboration examples between AI and us; 3) anyways a new era of how to think old questions has already arrived.

Some personal using AI experiences

Again back to my personal experiences using 'Cursor' collected during these two months.

When trying AI coding the first time, using auto complete plus AI agent might be a better starting choice because AI agent's way of coding or helping really takes some time for you to get used to and auto complete features are much more mature than before, which means better not to give everything of your code to AI at once at the very beginning but see what AI can do and achieve gradually.

Rules (instructions) are perfect places to record how to work with AI well, for example, you always want 'External Dependencies' or 'Internal Dependencies' and similar comments above needed import sections, then rules can very easily help you with this. This is an example of your will versus AI will and also lets you record all your own understandings of how to code from your years' coding experiences. But writing rules is sometimes like a (very) higher-level thinking task that again AI itself might not help that much since your rules are the guide about how AI will do the work under certain guidances and are very related to and depend on the author's thinkings (hard but really needed).

Skills have the similar story like automatically setting up the 'Prettier + ESLint + husky + lint-staged + commitlint' code quality management workflow. So, one of the ultimate current futures of a programmer might be to have your own resources since it's not like before, and to build and own your own tech things (rules, skills or even your own libraries, and your way of using other services, etc.) must be very beneficial in the future.

Another thinking of the real future at this stage

How possible AI develops its own new languages? Or using what languages to finally fulfill certain things is not going to change that much (there's always a boundary of what AI can do based on existing tools and human plans)?

But totally I think it's already not about a complete product considering current development speed of AI but with APIs (infrastructures, languages and so on) people can build super customized their own apps of any kinds. And even in the 'very' future these apps run on the chips inside our brains.