I think this article is not really correct, but I have been programming since the late 1960s, so my perspective may be different.
There have been many new tools. The simplistic DWIM AI coding assistant on my Lisp Machine seemed like magic 40 years ago and perhaps LLMs seem a little magical now.
However they are just tools to get work done faster. I love coding with assistance from Claude, Gemini, and ChatGPT.
Exactly. I didn't start coding until the '80s, but even at that point I was writing a lot of assembly code. I don't do that anymore. Each decade that I've coded I look back and see that I've moved up to higher and higher levels of complexity over the years. I was skeptical of AI coding, and I still am. But I've seen that it is capable of writing single line expressions quite well, and if the function is well scoped enough, it can even write full functions occasionally. My favorite use of AI is as a rubber duck, for me to bounce my ideas off of before I actually write any code. I expect it's capabilities will get better and better over time as everything does. Having The AI code these smaller bits allows me to focus more on gluing the bits together, just like I've done for the last 10 years with all the open source libraries that are available these days. I don't mind that I forgotten how to use the Lisa assembler for my Apple ][. I'm happy to keep forgetting stuff if I keep working on more and more complicated projects.
I think I see where this article is going. I don't ever want to be known for being good at programming Next.js. I don't ever want to look at the <div> salad that "good" React demands and feel proud of it. In my current role I'm perfectly happy writing poor code for these painful abstractions of node.
I don't even want to know where Next leaves off and React begins, or even where React leaves off and TypeScript begins, or even if I'm using TypeScript or JavaScript. Let AIs handle the miasma that is front-end development. I'll paste it in there, test it, make it as succinct as possible and get it off my plate so I can do something more useful.
This compared to the time I used AI to help with my first Ruby project, where it pointed me to some cool libraries, pushed me to look up amazing functionalities that my PHP-corrupted brain couldn't even conceive of, etc.
Using a calculator will make you a bad mathematician... yet we still use calculators. Often, on important maths, I'll even double-check with another calculator/person.
> No one in a decade will pay top dollar for a slow dev that knows their handwritten code indside out.
But they will have to pay for the dev that can debug years of code that their fleet of prompt engineers cant figure out.
> Its all about scale and speed. Now you can argue that it takes the artisanship out of development but truth is, no one cares.
I think you're correct, the MBAs making decisions won't care, and will choose scale and speed. And so will the AI driven developers, but the actual engineers will care.
> They are not coming for the tech jobs, they are already here. Just get on top of them and youll survive.
I'm not trying to survive, and neither are any of the engineers I respect. We're all well aware of the limitations of AI, and we know it's not coming for our jobs, (not yet anyways... I guess I should give it another quality revolution or so.) We're still preoccupied with trying to thrive. AI isn't a threat to us, but it is a threat to the next generation. It's near impossible to evolve, learn, grow, and adapt if you delegate all the thinking, and understanding about the multiple layers of the system you're working on to the GenAI robit your company is paying for; (instead of paying you that money)
Not really. When I'm doing something new to me, I do not let AI write that code, otherwise how can I evaluate the code that it produces?
Skills You Already Have May Atrophy
Of course they will! Just like they do when you use a library of routines. Nobody writes merge sort outside of their algorithms class. Could I sit down right now and write merge sort? No. Do I care? Not really. Would I let AI write it for me? No - see above: I don't let AI write code I'm not familiar with. I'd prefer to use a library routine and move on. Failing that, I'm going to have to re-implement merge sort. The upshot is since I've implemented before, it'll all come back to me, and I can quickly write it again.
Afterward, I might even let AI take a crack at it and compare its code to my code. I'll use whichever one is best.
You May Become Dependent on Your Own Eventual Replacement
People said that about Intellisense, too. People said that about Code Wizards. I remember when MFC was first released in the early 1990s and with one click of a mouse button it could create a Windows Hello World! program. I remember spending days going through Petzold's book learning how to write a basic Windows program. Now all I have to do is click a button!
On a side note - when is the last time you ever bothered yourself with a windows message pump? Processor caching and branch prediction optimization? How about loop unrolling optimizations? There are times when such knowledge is needed, but I'd wager 90+% of today's programmers have never implemented such things outside of the classroom: if they even did that! And they're not using AI.
Your Code Will Not Be Respected
Are you saying AI-generated code can't pass the Turing test? That's - interesting. You talk about solutions, but in my mind, coding isn't solutioning. That's what architecture and design is for. Coding is implementation. That's why coding is referred to as a craft.
Can AI get good at the craft of coding? That's a question of automation and to answer that question I'll point out there aren't too many black smiths or white smiths around today. Or weavers, for that matter. These are all trades requiring a high degree of knowledge and skill, like coding. And they're mostly gone.
But, mostly gone isn't the same thing as completely gone. People still enjoy learning these trades for fun. Imagine something like Williamsburg for Silicon Valley 100 years from now. The tour guide will explain how people used to write programs. The guests will ooh and aah that people could ever have done such sophisticated tasks. Maybe one of the guests will be intrigued and learn the coding craft.
This has been going on for millennia. The only reason you're scared now is because there are fewer and fewer crafts that can't be automated, and we've built an economic system around scarcity, especially the scarcity of labor. We know from history when such tectonic shifts happen its very disruptive to civilization. Stability is lost and wars are fought. I believe that's what you actually fear, and not AI per se. You fear the change that comes with AI.
On a positive note, you're alive at what will likely be one of the greatest inflection points in human history! How lucky is that?!
> On a positive note, you're alive at what will likely be one of the greatest inflection points in human history! How lucky is that?!
That remains to be seen. Not all inflection points are good things. If we're at a major inflection point with generative AI (and I doubt that we are), it's possible that will end up being a bad thing, not a good one.
I'm not saying that this tech will end up being good or bad on the whole, I'm saying that it's far too soon to be able to predict that.
I do agree, though, that it's not the technology itself that presents risk, it's people developing and using it that could go very very wrong.
Absolutely! Hence the fear. Technology is a double-edged sword - there are always upsides and downsides. With most technologies it seems as though we only discuss the upsides. Rarely are the downsides discussed, or even mentioned. When it comes to automation technology though, the focus always quickly shifts to the downsides. I think that reflects a deep distrust of civilization and the status quo.
I like to think we are simply at the next step of abstraction that will help explode faster development across the board. It is well overdue. Think of the step from compiled languages to scripting languages - so many engineers (myself included) know fuck-all about about garbage collection and pointers, but we can still get useful work done. For some things, you want or need to use a compiled language for greater control/speed.
Even with the rapid rise of AI, I think we will be stuck at this stage for quite a while. Yes, AI will be able to made a button for your app or make your app completely, but what people always want is greater control. Someone needs to be making business decisions about how things work and interact, or how things are architected for anticipated load. So a lot of easy and repetitive work goes away but demand shifts to more deep tech engineers and more business/design engineers.
I think this article is not really correct, but I have been programming since the late 1960s, so my perspective may be different.
There have been many new tools. The simplistic DWIM AI coding assistant on my Lisp Machine seemed like magic 40 years ago and perhaps LLMs seem a little magical now.
However they are just tools to get work done faster. I love coding with assistance from Claude, Gemini, and ChatGPT.
Exactly. I didn't start coding until the '80s, but even at that point I was writing a lot of assembly code. I don't do that anymore. Each decade that I've coded I look back and see that I've moved up to higher and higher levels of complexity over the years. I was skeptical of AI coding, and I still am. But I've seen that it is capable of writing single line expressions quite well, and if the function is well scoped enough, it can even write full functions occasionally. My favorite use of AI is as a rubber duck, for me to bounce my ideas off of before I actually write any code. I expect it's capabilities will get better and better over time as everything does. Having The AI code these smaller bits allows me to focus more on gluing the bits together, just like I've done for the last 10 years with all the open source libraries that are available these days. I don't mind that I forgotten how to use the Lisa assembler for my Apple ][. I'm happy to keep forgetting stuff if I keep working on more and more complicated projects.
Any tool that improves any work will and shall be used.
I think I see where this article is going. I don't ever want to be known for being good at programming Next.js. I don't ever want to look at the <div> salad that "good" React demands and feel proud of it. In my current role I'm perfectly happy writing poor code for these painful abstractions of node.
I don't even want to know where Next leaves off and React begins, or even where React leaves off and TypeScript begins, or even if I'm using TypeScript or JavaScript. Let AIs handle the miasma that is front-end development. I'll paste it in there, test it, make it as succinct as possible and get it off my plate so I can do something more useful.
This compared to the time I used AI to help with my first Ruby project, where it pointed me to some cool libraries, pushed me to look up amazing functionalities that my PHP-corrupted brain couldn't even conceive of, etc.
AI can offload the burden of producing more of something it’s seen a lot of. Including miasma.
AI is much less likely to help you engineer your way out of miasma.
When you automate something, you get more of it. Except attention, you get less of that.
Using a calculator will make you a bad mathematician... yet we still use calculators. Often, on important maths, I'll even double-check with another calculator/person.
I think this is a naive article trying to stop a huge wave that is already here.
I know developers that fret over younger devs that have never even seen assemby and thus do not understand the code they write and so on….
Learn how to use AI to be better and faster than everyone else.
No one in a decade will pay top dollar for a slow dev that knows their handwritten code indside out.
Its all about scale and speed. Now you can argue that it takes the artisanship out of development but truth is, no one cares.
Learn how to use AI to get faster and learn how to teach the AI to be better and more accurate.
They are not coming for the tech jobs, they are already here. Just get on top of them and youll survive.
> No one in a decade will pay top dollar for a slow dev that knows their handwritten code indside out.
But they will have to pay for the dev that can debug years of code that their fleet of prompt engineers cant figure out.
> Its all about scale and speed. Now you can argue that it takes the artisanship out of development but truth is, no one cares.
I think you're correct, the MBAs making decisions won't care, and will choose scale and speed. And so will the AI driven developers, but the actual engineers will care.
> They are not coming for the tech jobs, they are already here. Just get on top of them and youll survive.
I'm not trying to survive, and neither are any of the engineers I respect. We're all well aware of the limitations of AI, and we know it's not coming for our jobs, (not yet anyways... I guess I should give it another quality revolution or so.) We're still preoccupied with trying to thrive. AI isn't a threat to us, but it is a threat to the next generation. It's near impossible to evolve, learn, grow, and adapt if you delegate all the thinking, and understanding about the multiple layers of the system you're working on to the GenAI robit your company is paying for; (instead of paying you that money)
> I think you're correct, the MBAs making decisions won't care, and will choose scale and speed
won't that lead to aircrafts coming apart in flight, spaceships getting grounded and satellites exploding?
https://newsletter.pessimistsarchive.org/p/slide-rule-fast-a...
Same energy
Using generalizations will make you a bad writer. Yes I am showing that I am a bad writer.
You Rob Yourself of Learning Opportunities
Not really. When I'm doing something new to me, I do not let AI write that code, otherwise how can I evaluate the code that it produces?
Skills You Already Have May Atrophy
Of course they will! Just like they do when you use a library of routines. Nobody writes merge sort outside of their algorithms class. Could I sit down right now and write merge sort? No. Do I care? Not really. Would I let AI write it for me? No - see above: I don't let AI write code I'm not familiar with. I'd prefer to use a library routine and move on. Failing that, I'm going to have to re-implement merge sort. The upshot is since I've implemented before, it'll all come back to me, and I can quickly write it again.
Afterward, I might even let AI take a crack at it and compare its code to my code. I'll use whichever one is best.
You May Become Dependent on Your Own Eventual Replacement
People said that about Intellisense, too. People said that about Code Wizards. I remember when MFC was first released in the early 1990s and with one click of a mouse button it could create a Windows Hello World! program. I remember spending days going through Petzold's book learning how to write a basic Windows program. Now all I have to do is click a button!
On a side note - when is the last time you ever bothered yourself with a windows message pump? Processor caching and branch prediction optimization? How about loop unrolling optimizations? There are times when such knowledge is needed, but I'd wager 90+% of today's programmers have never implemented such things outside of the classroom: if they even did that! And they're not using AI.
Your Code Will Not Be Respected
Are you saying AI-generated code can't pass the Turing test? That's - interesting. You talk about solutions, but in my mind, coding isn't solutioning. That's what architecture and design is for. Coding is implementation. That's why coding is referred to as a craft.
Can AI get good at the craft of coding? That's a question of automation and to answer that question I'll point out there aren't too many black smiths or white smiths around today. Or weavers, for that matter. These are all trades requiring a high degree of knowledge and skill, like coding. And they're mostly gone.
But, mostly gone isn't the same thing as completely gone. People still enjoy learning these trades for fun. Imagine something like Williamsburg for Silicon Valley 100 years from now. The tour guide will explain how people used to write programs. The guests will ooh and aah that people could ever have done such sophisticated tasks. Maybe one of the guests will be intrigued and learn the coding craft.
This has been going on for millennia. The only reason you're scared now is because there are fewer and fewer crafts that can't be automated, and we've built an economic system around scarcity, especially the scarcity of labor. We know from history when such tectonic shifts happen its very disruptive to civilization. Stability is lost and wars are fought. I believe that's what you actually fear, and not AI per se. You fear the change that comes with AI.
On a positive note, you're alive at what will likely be one of the greatest inflection points in human history! How lucky is that?!
> On a positive note, you're alive at what will likely be one of the greatest inflection points in human history! How lucky is that?!
That remains to be seen. Not all inflection points are good things. If we're at a major inflection point with generative AI (and I doubt that we are), it's possible that will end up being a bad thing, not a good one.
I'm not saying that this tech will end up being good or bad on the whole, I'm saying that it's far too soon to be able to predict that.
I do agree, though, that it's not the technology itself that presents risk, it's people developing and using it that could go very very wrong.
Absolutely! Hence the fear. Technology is a double-edged sword - there are always upsides and downsides. With most technologies it seems as though we only discuss the upsides. Rarely are the downsides discussed, or even mentioned. When it comes to automation technology though, the focus always quickly shifts to the downsides. I think that reflects a deep distrust of civilization and the status quo.
I like to think we are simply at the next step of abstraction that will help explode faster development across the board. It is well overdue. Think of the step from compiled languages to scripting languages - so many engineers (myself included) know fuck-all about about garbage collection and pointers, but we can still get useful work done. For some things, you want or need to use a compiled language for greater control/speed.
Even with the rapid rise of AI, I think we will be stuck at this stage for quite a while. Yes, AI will be able to made a button for your app or make your app completely, but what people always want is greater control. Someone needs to be making business decisions about how things work and interact, or how things are architected for anticipated load. So a lot of easy and repetitive work goes away but demand shifts to more deep tech engineers and more business/design engineers.