The art of programming and why I won't use LLM

I take it you're not using a compiler to generate machine code, then?

Scratch that, I guess you're not using a modern microprocessor to generate microcode from a higher-level instruction set either?

Wait, real programm^Wartists use a magnetised needle and a steady hand.

Programming has always been about finding the next black box that is both powerful and flexible. You might be happy with the level of abstraction you have settled on, but it's just as arbitrary as any other level.

Even the Apollo spacecraft programmers at MIT had a black box: they offloaded the weaving of core rope memory to other people. Programming is not necessarily about manually doing the repetitive stuff. In some sense, I'd argue that's antithetical to programming -- even if it makes you feel artistic!

There are a handful of cases where LLMs are useful, mainly because Google is horrifically bad at bringing up useful search results, it can help in that regard... or when you can't find the right words to describe a problem.

What I would like to see out of an AI tool, is something that gobbles up the documentation to another programming tool or language, and spits it back out when it is relevant, or some context aware question and answers like "where in the code base does XYZ originate" or w/e. the difference is having a tool that assists me VS having a tool spit out a bunch of garbage code. Its the difference between using a tool, and being used by a tool

But I don't know man, I love coding with LLMs. It just opens up more things, I think on some projects I actually spend MORE time on traditional coding than I did in the past, because I used an LLM to write scripts to automate some tedious data processing required for the project. And there's also projects where the LLM gets me from 0 to 60 and then I rather quickly write the code I actually care about writing, and may or may not end up replacing all the LLM written code.

I'm sure it heavily depends on exactly what types of project interest you. The fact that LLMs and diffusion have both become fixations of mine also means I have a lot more data processing involved in lots of my projects, and LLMs are quite good at custom data processing scripts.

I suppose my suggestion to the author would be that perhaps their projects aren't amenable to LLMs in the way they want and that's fine, but don't lose hope that there are kindred spirits out there just because so many people love LLM coding; some of us are both and that may be more about what types of projects we do.

These high level abstractions are where I find the most joy from programming. Perhaps for some there is still some modicum of enjoyment from writing a for loop but for most people twenty years into a career there's nothing but the feeling of grinding out the minutia.

There's still a lot of room for better abstractions when it comes to interfacing with computing devices. I'd love to write my own operating system, CLI interface, terminal, and scripting language, etc from scratch and to my own personal preferences. I don't imagine I could ever have the time to handcraft such a vast undertaking. I do imagine that within a few decades I will be able to guide a computing assistant through the entire process and with great joy!

[0] https://github.com/williamcotton/guish

I use an LLM to generate ideas, to rubber duck, to get a lead on unknowns, and to generate boilerplate occasionally. So I do everything except replace the coding part because that's what requires the most precision, and LLMs are bad at precision. And yet, people claim massive productivity gains in specifically coding. What am I missing?

People that refuse to program with AI or intellisense or any other assistance are like carpenters who refuse to build furniture with power saws and power drills. Which is perfectly fine, but IMO that choice doesn't really affect the artistry of the final product

LLMs can do mindless drudgery just as well as I can, but in seconds instead of hours. There's nothing about remembering syntax, boilerplate code, forgetting a semicolon, googling the most common way of doing something, or combining some documentation to fill in the gaps that's even remotely "art" to me.

I never ask an LLM for what I'm artfully creating. I ask it for what I know it'll get instantly right, so I can move on to my next thought.

Like, ideally, it shouldn't really take that much code to implement a thing. I like to think of programming as writing a bunch of levers, starting with simple levers for simple jobs, incrementally ratcheting up to larger levers lifting the smaller levels. Before too long, it'll feel as though you've written a lever capable of lifting the world...or at least one that makes an otherwise wickedly difficult project reasonably manageable.

If you say that LLMs make you more productive because it allowed you to finish a project that would otherwise take forever to write, then I'm skeptical that an LLM is the best solution. I mean, it's a solution at least, but I can't help but wonder if there's a better solution.

If the problem is that you lack the understanding to take on such a project, then perhaps what we really need are better tools for understanding. I myself have found that LLMs are great for gaining a quick understanding of languages that otherwise have sparse information for beginners, but I have to wonder if perhaps there's a better way.

If, on the other hand, the problem is that writing that much code would take forever, then I have to wonder if the real solution is that we need a better way to turn programming languages into patterns (levers) and turn said patterns into larger patterns (larger levers)

A partial solution works, but only partially well, and occasionally has consequences one has to reckon with

* They are great for overcoming procrastination. As soon as I don't feel like doing something or a task feels tedious I can just delegate it to an LLM. If it doesn't solve it outright it at least makes me overcome the initial feeling of dread for the task.

* They give me better solutions than I initially had in mind. LLMs have no problem adding laborious safeguards against edge-cases that I either didn't think of or that I assessed wouldn't be worth it if I did it manually. E.g. something that is unlikely and would normally go to the backlog instead. I've found that my post-LLM code is much more robust from the get go.

* They let me try out different approaches easily. They have no problem rewriting the whole solution using another paradigm, again and again. They are tireless.

* They let me focus on the creative parts that I enjoy. This surprised me since I've always thought of myself as someone who loves programming but it turns out that it is only a small subset of programming I love. The rest I'm happy to delegate away.

We had an overdemand for devs during late ZIRP early COVID leading to bootcamps and self taught pulling a lot of untrained into the industry. Many of them have left the industry.

Add to that the whole data science bubble and it’s bursting where we had tons of degrees and job openings for sort-of-devs. Lot of those jobs are gone now too.

Don’t forget the pull of “product management” and its demise outside big tech.

Now we have hiring freezes and juniors leaning on LLMs instead of actually spending an hour trying to solve problems.

Interesting times.

With ai I need to review the output but not because there may be some issues I didn't noticed, but because that may be issues the tool itself didn't noticed, so it's less of an "apply this specific change" but more of an "apply some change"

For work stuff? I couldn’t care less if the code comes from me, my colleagues or an LLM. As long as it works and it’s secure, we’ll ship it.

Folks, career != job.

In an ideal world, our abstractions would be so perfect that there would be no mundane boiler-platey parts of a program; you'd use the abstractions to construct software from a high level and leave details be. But our abstractions are very far from perfect: there's all kinds of boring code you just have to write because, well, your program has to work. And generally that code is, if you look, most of your code. This because making good abstractions is really hard and constructing fresh ones is often more work than just typing out the different cases. If you think this is mistaken, I'd gently suggest you take a fresh look at your own code.

Anyway, that's where LLMs come in. They help write the boring code. They're pretty good at it in some cases, and very bad at it in others. When they're good at it, it's because what the code should do is sort of overspecified; it's clear from context what, say, this function has to do to be correct, and the LLM is able to see and understand that context, and thus generate the right code to implement it. This code is boring because it is in some vague sense unnecessary; if it couldn't be otherwise, why do you have to write it at all? Well you do, and the LLM has taken care of it for you.

You can call this work the LLM is displacing "art", but I wouldn't. It's more the detritus of art performed in a specific way, the manual process required to physically make the art given the tools available.

You could object that the LLMs will get better in the sense that not only that they will make fewer mistakes, but they will be able to take on increased scope, pushing closer to what I'd consider the "real" decisions of a program. If this happens -- and I hope it does -- then we should reevaluate our lofty opinions of ourselves as artists, or at least artists whose artistry is genuinely valuable.

Author needs to get into Bret Victor. Has no idea how much more fun he could be having.

Programming is a step on the way to access to the state space of information. When we get to that stage, programming will seem like a maze of syntax, that has its own idiosyncrasies that force you into corners or regions in the state space, just like any DAW plugin or 3D tool, or any tool at all that exists.

I feel like you have to drop this kind of thinking to get anywhere past intermediate, not to mention you become a nightmare to anyone who has even a touch of pragmatism about them.

Sad to say, LLMs have made me lazy coder for the past two years or so. But I do deliver/finish work much faster, so my incentives to keep using LLMs as coding co-pilots are overshadowing my incentives to write code the "old way".

And for what it is worth, sometimes these posts read like modern luddite confessions - the rants just sound too personal.

Like what great ideological purpose are you serving if you delegate writing tailwind boilerplate to a LLM, or a basic axios get request?

Does your code become impure if instead of copying code from the documentation you get an LLM to do it?

But this has led me to wonder if there will be gradual pressure to build on top of LLMs, which, in turn, will really only be useful with the tried and true. Like, we’re going to be heading towards an era where innovation means “we can ask the LLM about it”. Given the high capital costs required to train, I wouldn’t be shocked to see LLMs ignoring new unique approaches and biasing to whatever the big corps want you to do. For “accuracy”.

I just sense were about to hit an era of software causing massive problems and costs, because LLMs are rapidly accelerating the pace of accidental complexity, and nobody knows really how to make money off them yet.

As we approach this point, programming (program creation) will become a largely non-technical skill but the few who have that skill will be needed to solve the problems which LLMs create and which its operators don't have the know-how to fix.

I use LLMs as an alternative to asking a question on stackoverflow. And I'm not confident LLMs can currently save time on large projects: take a couple of hours to write the code or spend a few seconds to generate the code and then a couple of hours understanding and debugging it.

I don't see it as a "judgemental piece to anyone that i am alluring to" (I think you mean alluding) but I do think it's an honest assesment of those who are attempting to rely heavily on AI.

I use LLMs cause it takes care of the syntax and leaves me more time to think about the logic and how everything is connected together. LLMs still lack in the latter category.

For one-off tasks, it's truly a time saver. In the future, I see people trying to do more complex stuff with chains of LLM calls (already happening but it's still a WIP). Break a complex task into sub-tasks, then solve them sequentially to achieve something non-trivial in the end. I'm a bit skeptical though - not so much because of my skepticism of LLMs in general but more about humans not being good at maintaining code bases they didn't write (often even with code that wrote). You cannot trust LLMs to also maintain the codebase. There's too much context about the life of the code, all the dependencies, and even business logic, that is missing.

I wrote a whole Slack clone (not feature complete of course) in just 1 day. I love to code, and I do have to at times but this feels like the end of what was the joy of my youth and adulthood, my favorite pastime is gone. I will still develop applications and enjoy my creativity but with probably 1/50th the effort it took before.

https://news.ycombinator.com/item?id=38840626

For me it feels like it enables me to express myself even better. I use 1 for basically a super charged intellisense or clever auto complete. Finishing lines as I am typing them. Exactly the way I was going to write them.

It saves time on the more boiler plate glue code and allows me to maintain a better flow and momentum on more expressive areas.

I don't see it taking away anything but only enabling me to do more, faster, and better. I'm not telling it to write full apps because it can't.

It seems to me that there is a middle ground between writing programs myself and downloading code from a package manager. LLMs fill up some space within that gap.

I think refusing to use an LLM because of "reasons" is the same as refusing to use packaged libraries for "reasons". That is, reasons for both definitely exist but I consider that kind of stubborn intransigence to be a sign of mental disorder.

I recommend simply adjusting the criteria you use to decide when and when not to use packaged libraries to help determine when and when not to use LLMs.

I like to code in the middle of the night, solving very complex problems and building cool applications.

Programming is just a mean to an end, please stop this cringe romantic rhetoric. If you really love programming you don’t care about the medium, the fact that a program is represented as text is just a transient phase in the history due to the current tech we have available in this specific moment in time. Programming (today) is expressed as text, LLMs auto complete boring text, so that us “the artists” can write more of it faster, end of story.

I wrote a blog post about it yesterday:

https://henrikwarne.com/2024/08/25/programming-with-chatgpt/

Programming was mostly a hobby in the days of 8-bit PCs. It was a profession for some decades. Maybe it will be a hobby again in 5 years. Like gardening, sailing, fishing - professions at one time, now hobbies.

On the other hand, the arrival of futuristic capabilities like computers speaking human languages is what drew me to technology in the first place. Luckily, you can choose to look forward and backward. You don't have to pick only one.

I would never judge anyone who finds your non-capitalized, grammatical-error ridden essays seductive.

Not helped by the choice of using a monospaced font. I get that it is often an aesthetic choice, but given that a blog post is written with the idea to be read, one I don't think is a particularly good one. Although the last time I made a remark about that on HN it became clear to me that a lot of people don't see the issue. Even if there are decades worth (at this point) of research that makes it clear that a sans serif font (or even a serif font on modern displays) works better for readability. ¯\_(ツ)_/¯

In this case though, the combination of the monospaced font, everything being in lowercase and the run-on sentences I really am scratching my head here.

Are you trying to get a message out there? Or are you mostly going for aesthetics?

Of course I understand that the "sweet spots" of logic, design, and platform architecture shifts over time. I can work at a higher level of abstraction now than I did when I started out writing PHP in 1999. I'm also willing to believe low-code/no-code solutions or turnkey solutions will get better over time. There's a reason I don't try to find restaurant clients to build a glorified menu website for anymore.

But the hoopla around LLMs has only served to reinforce the disconnect between the people who express artistry in code and the people who see programming as nothing more than a "feature-implementation-pipeline" that's ripe for rote automation. I'd expect as much from folks in marketing, or CEOs, but to see so many CTOs make public claims they must privately know is simply bullshit…well, that's shaken my faith in this industry, I'll tell you that much.

I'm feeling optimistic though. The tides are turning against generative AI and when the hype cycle is finally over, we'll have a handful of narrowly-purposeful, ethically-engineered models which provide some specific yet useful productivity gains…and beyond that, it'll just be the slow-but-steady progress of the discipline of hand-writing code.

But to the point of the post, yeah, I agree. The people using LLMs are just using it to compensate for a horrible toolkit. It doesn't help them think or break down problems that nobody has seen before. If it makes them quicker it's because they were using caveman-level tools before. I haven't typed every character in the code I produce for at least a decade at this point. Most of my time and effort is in thinking. But for braindead code monkeys, sure, anything will make you quicker.

That said, in my experience, trying to get ChatGPT to spit out sensible code is sometimes like arguing with a conspiracy theorist. I ask a question, receive an answer, ask a follow-up question and scrutinise one of the details, and the interviewee just falls apart and grasps for other straws.

My formal education was in music composition and film scoring specifically. I never once, even then considered myself "an artist" or thought that I was pursuing art. I did consider myself a craftsman: I was learning a style of music composition where I was expected to realize someone else's vision for their purposes. One could argue that the filmmaker was an artist and the film a work of art, but as the composer my job was to produce a work product that conformed to a specification as directed by someone else. I cared about creativity, quality, and having my product fit for purpose. I do not see the craftsman as inferior to the artist, but the goals are different and in a medium such as film you must have both artists and craftsmen (and sure sometimes a single person is both, but that's not a requirement).

When I am a programmer I see myself in exactly the same way. Professionally, I'm paid to satisfy someone else's need for automation/data wrangling. The solution I deliver is a work of craftsmanship, not of art. I will be creative in exercising my craft: often I'm just given a statement of the problem that needs solving and the solution design and development are left to me. I may find unique capabilities or processes which are better than those anyone envisioned... but at the end of the day I am creating an expression of my craft, not trying to achieve a deeper artistic goal. Trying to achieve an artistic goal with programming, at least in most professional programming, would likely lead to some sort of malpractice: the craft of the product diminished in time, cost, or features to gain an unasked for message.

Another issue I take with this article is that it confuses the tools with the result. Today, artists that paint use a large variety of brushes, spatulas, and other tools and paints which have all the benefits of being formulated with the knowledge of modern chemistry for stability, texture, and durability over time. Would the artist of old be justified in suggesting that using the new tools make you not an artist since you don't have to control for the limitations of the past? As a composer, I used sample players, sequencers, notation programs, and DAWs, even when writing for a conventional orchestra. True Beethoven didn't have these tools, but does this mean a modern composer that might use the same tools as me don't produce art... even if the final result is played by a traditional orchestra? I find LLMs to be in this category. When producing art (or craft as I discussed), the final product is ultimately what is to be judged... unless you find programming as being more about performance art where the journey is more the product that the final result.

Anyway, as a programmer I absolutely use LLMs. Sometimes they help, sometimes not. Even at my most creative, I don't find developing a good regex or a shell script something I care to be creative in approaching. How I apply the LLM, my judgement of the quality of what it produces, or whether I think the kind of "lowest common denominator" approach an LLM will produce is appropriate or not in a given situation still puts me and my creativity plainly at the center of producing a craft-full, and perhaps artful, product.

BTW what's the aider vs cursor opinion?

Unlike art where you can tell if the end product is AI generated even, software is not. The end users do not care if you used AI to generate your react front-end or back-end, your employers dont care who wrote the code as long as its bug free and works like in the scope document.

Ironically, I don't see software developers writing manifestos or complaining that AI generated software isn't software or that they are stealing from them.