Goodbye to the Hard Parts That Never Mattered(jakegoldsborough.com) |
Goodbye to the Hard Parts That Never Mattered(jakegoldsborough.com) |
> The learning loop got tighter. Ask a question. Inspect the answer. Run the code. Break it. Read the implementation. Correct the assumption. Try again.
This assumes that you have the resources (time and money/AI usage budget) and the interest/motivation to do that. If you do programming as a job rather than as a hobby, the time you can spend on things like this and the budget you can spare for it will be severely limited. Also, after you have finally coaxed the LLM into generating something that fits all the requirements without breaking in new and surprising ways, has code that's not ridiculously overengineered (that's probably the real reason behind vibe coding: as long as you don't look at the horror beneath the hood, you can tell yourself everything's fine), doesn't break some convention of the codebase and has the required test coverage and other metrics, you're not only out of time for the task at hand, but also more than ready to move on to something else.
This is my favorite part too. With my own project for example, I had the agent integrate libgit into it so I could add source control to my IDE. Yes, I could have done it myself, but it would have taken me months of trying to understand the API, testing various things, and finally implementing it all. Instead, the agent did almost all the work (I still designed the architecture and how it plugged in). It took about two weeks calendar time and most of that time I was doing something else while the agent worked.
The thing is, before AI agents, I wouldn't have even attempted the work. I wouldn't have been able to afford the time.
Before AI agents, what we would have is maybe a line in the changelog, like “git support has been added” and maybe some post if there’s a substantial Ui/Ux improvement or novelty.
Now it’s just: I did something with AI (with no description of why it has been done, just what) and it was pretty fast (compared to an exaggerated estimation).
There has been a divergence that is quite stark. I have seen it argued that this is a reflection of ability, and that as a skill multiplier those lacking in skill are feeling left behind. I'm not certain that this is accurate. I definitely see people who have made things that impressed me in the past have been more likely to embrace AI. It may just be a measure of a type of personality.
I have noticed that people who have a strong sense of possessiveness over what they create are more resistant, but those who create in order to have the world contain the thing they are making are happy to have anything that will enable them to contribute to the world.
Some people though society was paying them lots of money to type code into an editor. But that was never true - people and businesses were paying them to solve problems.
I'm not sure how you build that sense as well if you're starting out today, though.
This is a very uncharitable take; I see it mostly from devs who are all-in on AI (not saying you are one), not from devs who just use AI while programming. Basically, I see it from devs who want to never touch a keyboard again.
Maybe they never liked programming in the first place, but liked managing instead, so now they can manage instead of programming.
Maybe they never liked programming, but they did like delivering; now they can work as a product manager, delivering only.
Whatever the reason, they never really liked programming[1]. My experience in programming, since the mid-90s, is that my design is refined as I type out the programs. I gain insight into the problem as I create the program. My thoughts are refined as I read what I type (and not just with programs, with prose as well).
Possibly the all-in devs never read what they typed anyway, so the act of programming never caused them to refine their ideas. Maybe they never read what they typed as they typed it in. Maybe they did, but it never triggered introspection.
Whatever the case is, it's a very uncharitable take to trivialise those engaging in more thought, due to actually writing, as simply enjoying the act of typing.
They are not. They are enjoying the iterative process of deep thinking.
TBH, maybe this is why so many of the all-in AI vibers are trying to trivialise the typing part - if they can convince themselves that typing is so trivial and bereft of thought, then they don't have to admit to themselves that they are thinking less, and the thinking they do do, is not as deep.
----------------------------------
[1] All advances in programming, such as in data structures, algorithms, languages, theory, etc came mostly from people who liked programming. As these people can not hone their skills by working at it for money in the future, who knows if we ever get any more advances.
https://web.archive.org/web/20260606173815/https://jakegolds...
Whether or not this author wrote it themselves (I strongly suspect at least an AI editor), the specific style that was, for a time, very attractive has become the hallmark of AI-written prose; and, while the content doesn’t seem wrong from a particular view, it also leaves a taste of waste to me - waste that I spent that time reading something they didn’t write, full of -isms that aren’t theirs, when the thesis could have padded out far less paper.
The word count feels unearned, even if the content is fine.
It's one thing to have AI-written software, where the code isn't the consumable; and quite another to have AI-written essays that are intended to engage a human's critical thinking.
I am very tired of it all.
The future is exciting and breathtaking, not lamentable.
Stop crying about punch cards and hanging chads. Those are some thick rose-tinted glasses.