But now, AI writes most of the code. The advantages of course are many, not the least of which is that I'm able to build and maintain multiple products at once.
But instead of coding, I've turned into a full time reviewer instead. Spending the day reading, questioning and correcting AI is tiring in it's own specific way.
Worst of all, I no longer enjoy the satisfaction of solving hard problems on my own anymore.
For those of you in the same boat, what's your feeling? Has AI given you back your time? Or has it taken something away that you loved?
But personally? I won't stop using technology and I do my best to avoid AI as much as possible.
I like exercising my mind, to develop new ways of thinking and above all, cultivate critical thinking!
If there's no critical thinking, there's no creativity, no arts, no nothing!
In my case my passion is high performance in the JavaScript language space. I wrote my own WebSocket library and perf tool. A defect in the perf tool indicated that I could send 3.25m messages in 0.002 seconds if I separated each test of an experiment into a separate call stack. That number was confirmed as valid by Claud. Two weeks later I confirmed it was a defect and after fixing it my actual performance was only 100,000 messages per second. But, now I know the cause of the performance limitations impacting WebSockets. I modified how I manage multiple pending messages and my performance went up to 800,000 messages per second. I then discovered another similar performance hack and my performance went up to 6m messages per second on a single socket, or 5m round trip. So now I have something that is 1.5 orders of magnitude faster than everybody else or 4.5 orders of magnitude faster than http. AI doesn’t have that, I have it. What do you have? There is a lot AI cannot do.
I love(d) to figure out hard code and write it, but I also loved the creative process of searching the solution space and deciding (and optimizing) what to build.
I do not in general like to review code that much, which, granted, has grown.
But I feel that for me, all the former (figure out hard code and write it) has mostly split into the other two (in lack of better word) "solution design" and reviewing.
And also I feel that the endless supply of crappy yolo code for personal project is thrilling. Professionally I watch some code like a hawk and some is yolo, based on a concrete per "job" risk assessment.
And for context, I have administered unixy systems and written software for pay since 1999.
While the new tools offer a "step up" as noted, turning developers into reviewers, and possibly empowering home users, I see some problems.
AI tends to reinvent the wheel at every turn, and whether I am a professional developer wanting reliable and thoroughly tested routines as building blocks, or the home user not wanting the fridge to launch a DDOS attack on government sites, this seems counter to good practice, and the verbosity makes review difficult, so you have to get your tools to review your tools.
And there's no learning. Our great skill is in confronting the unknown, and ma lose this when everything is a button-push (or prompt) away.
I chug away at piano, rather than letting the MIDI play stuff.