Current thoughts on AI
Like any wannabe intellectual worth their salt, I also need to give my current two shits about AI. Of course, this is not my first post on AI, and it won't be the last. Or maybe it will, only fools think that they can predict the future.
Anyways, I came across this very excellent article by Glyph the other day. I recommend that you read it. It's well-written, entertaining and includes some great observations.
In the past few weeks, I also have been playing around with Claude Code a bit for programming. I have friends that say that they are 5x more productive with Cursor or Claude, but I just can't replicate this. I am not saying that they are lying, but I think they are caught in the hype. And as philosophers have been arguing for thousands of years: reality can be very subjective.
Needless to say, my own usage of LLMs for coding has been mixed. I see the value for quickly writing a stand-alone script using an API or language that you are not familar with. It's much quicker than looking up all the docs and syntax yourself. And it does not matter if you understand exactly what it does or why it uses those method calls, if it's a use-once-and-throw-away script.
However, working on larger codebases that I am familar with, in a language that I know very well, I just don't see the value. Sure, it can autocomplete and save you some time. But then again, if the majority of your job as a software engineer is writing boilerplate code, you might need to take a step back and think about WTF you are doing with 8 hours of your day.
I don't think LLMs in programming will go away entirely. But I think that the hype cycle will die down. I don't think the "progress" will continue linearly as it did in the past 3 years, not unless there is some other break-through in AI which introduces a step change shift. From an economic point of view, I think there will be a big crash soon. None of the AI companies are profitable. I doubt that if they would charge the real cost that as many customers as now would keep using their autocomplete on steroids service.
But then again, at the beginning of this post, I said that only fools think they can predict the future. And my dear reader, I am no damn fool! No, no. This is just my gut feeling as humble observer.
Anyways, I came across this very excellent article by Glyph the other day. I recommend that you read it. It's well-written, entertaining and includes some great observations.
In the past few weeks, I also have been playing around with Claude Code a bit for programming. I have friends that say that they are 5x more productive with Cursor or Claude, but I just can't replicate this. I am not saying that they are lying, but I think they are caught in the hype. And as philosophers have been arguing for thousands of years: reality can be very subjective.
Needless to say, my own usage of LLMs for coding has been mixed. I see the value for quickly writing a stand-alone script using an API or language that you are not familar with. It's much quicker than looking up all the docs and syntax yourself. And it does not matter if you understand exactly what it does or why it uses those method calls, if it's a use-once-and-throw-away script.
However, working on larger codebases that I am familar with, in a language that I know very well, I just don't see the value. Sure, it can autocomplete and save you some time. But then again, if the majority of your job as a software engineer is writing boilerplate code, you might need to take a step back and think about WTF you are doing with 8 hours of your day.
I don't think LLMs in programming will go away entirely. But I think that the hype cycle will die down. I don't think the "progress" will continue linearly as it did in the past 3 years, not unless there is some other break-through in AI which introduces a step change shift. From an economic point of view, I think there will be a big crash soon. None of the AI companies are profitable. I doubt that if they would charge the real cost that as many customers as now would keep using their autocomplete on steroids service.
But then again, at the beginning of this post, I said that only fools think they can predict the future. And my dear reader, I am no damn fool! No, no. This is just my gut feeling as humble observer.