I feel like this is the exact opposite of the conclusion I've been coming to. In an age where anyone can vibe code stuff at the drop of a hat, I want the ability to assert guarantees/contracts at a high level, and then let AI work out the details. I want to force AI to work within the confines of an abstraction, not independently of it.
We need abstractions more than ever. The real point is we should retire unnecessary abstractions.
A good abstraction is a single edged sword which simplifies the task. A mediocre abstraction is a double-edged sword. A bad abstraction is like a single edged sword with a restrictive handle and the sharp edge is facing towards you.
This is such a poor idea. LLM Agents don't have taste (abstraction capabilities). They are kind of like water, they take the shape of the container they're in. They are so much powerful when you build your own framework they can work in.
That is not what the text is arguing for and it has many good ideas, dropping the abstractions but keeping the test harness is one of them.
I've worked on many aspect of programming language implementation through the years and I would struggle with the abstractions of parser-generators and bytecode designed by committee like an enterprise Java developer struggles with the Hibernate ORM. Dunno what people are putting in their prompts but ask LLM to roofline and it will pursue that target like a bloodhound.
Abstractions aren't a burden though, they are more like a support. A good abstraction let's you write 10 programs in the time it would have taken to write one. Also, we now have "production" languages with generics and monomorphization so abstraction is easier than ever.
Djikstra has a great article as to why English language programming will always be second to a programming language. A "formal system" (edited from regular language) is the most precise way to specify things. An "abstraction" is basically a domain specific language.
The way I like to use LLMs is to build abstractions. I like to build mathematical code and ask the LLM to use it. I think that LLMs are very mathematical thinkers and they like it when they see Semigroup<T>. And it's not hifalutin either, I'm processing calculations along DAGs with a million+ nodes in milliseconds, thanks to abstractions
Many abstractions we encounter in software industry are poor abstractions which are counter-productive. I agree that we should drop those...
But we can't drop all abstractions. It's not physically possible. The AI agent is going to break logic up into files... The AI will create/impose an abstraction for each piece of logic whether we like it or not.
So saying "drop abstractions" is actually saying "let the AI decide what the abstractions should be" and unfortunately because LLMs are trained on average code, those abstractions are often pretty poor.
Coming up with the right abstractions is actually one of the main skills which AI hasn't automated and where the human brings the most value. It affects maintainability directly; including when using AI.
I've worked on projects with poor abstractions and ones with good abstractions using AI. The ones with poor abstractions require 10x to 100x more tokens and time to solve problems and implement new features. You're constantly fighting it to prevent it from coming up with nasty hacks and workarounds. Poor abstractions create the need for hacks and workarounds.
agents are unlike humans in many ways but they have their own limitations. it's magical thinking to believe that they are able to cope with zero software abstractions, especially as a codebase increases in size and complexity
and today's agents increase codebase size and complexity much faster than humans do via traditional software engineering methods, so this is likely to hit a wall faster than if human software teams were to do this experiment.
Just yesterday, for fun, I asked Codex to generate an “online e-commerce site” in ARM assembly. To its credit, it did actually generate a program that served up a webpage over a socket, but it really did not do what I asked. The program still pulled in a large number of shared libraries (something I’ve never really thought to try in assembly). And… it was not actually a functioning e-commerce site. Iterating on that initial prompt proved painful—the program was extremely brittle and now I had to review and talk about ASSEMBLY. Maybe we can dispense with abstractions someday, but holy cow, not yet. I would much rather write an e-commerce platform in C than try to vibe-assembly my way to one.
Following this argument to its natural conclusion, we should tell LLMs to generate the binary directly. It's still an abstraction, but it's the lowest level we can go to on current consumer hardware.
I don't think raw machine code (not even assembly) is the best form to reason about program logic. It just takes too many steps to execute sophisticated processes to keep them all in your head at once - or keep them all in an LLM context at once.
I kind of angree with the anrgument that for an ideal AGI abstractions are not needed. It would write everything perfectly without even needing a programming language. It would know everything to the byte and would create perfect machine code meeting all the requirements as perfect as it’s possible.
But the reality is different. First of all LLM were trained on current data, with the abstractions we currently use.
Secondly they suffer they same problems we do, their context size is way too small to just simply ignore abstractions.
it sounds like a bad abstraction. ime llms are really bad at reasoning through efficiency and utilization problems. they fail at scheduling and bin-packing for the same reason. it's actually difficult, multivariate, and not great.
i would be shocked if a well designed framework were less efficient than "bespoke" clankercode for low-level work
This is probably true. Today. But I think we're finding the problems LLMs are solving to be both paradoxically predictable but surprising at the same time. Not saying you're wrong, but scheduling and bin-packing seem like cards that will eventually fall to a bespoke prompt.
I mean, the post was written by a Stanford CS lab, so I'm inclined to believe that they know what they're talking about and are not just bad at creating abstractions.
I know nothing about the PhD candidate and professor who co-authored this, but I don't think this statement makes sense in general.
Typically CS labs at universities contain people who have not spent a significant amount of time exposed to large industry codebases and the corresponding complexity. I think the post would have more credibility coming from e.g. the platforms team at a tech company with a monorepo.
Well, they’re also probably not scaling up in the same way as any of the commercial AI offerings, let alone the frontier labs. Faculty at Stanford probably have some decent hardware to play with, but they do not have data centers. I don’t doubt that they know their way around CUDA/PTX, but it’s not clear how relevant their message is given that their research code is very likely not being deployed in production, or at scale.
The machine learning community decided to choose Python as the main language, which costs probably tens of millions of dollars in bugs every year due to it not being type checked (astral sh is fixing this though)
The main problem with Python is that, if you’re not doing your work in highly optimized numeric libraries (which are usually written in C), then you’re paying terrible overheads to use the Python interpreter. Type checking doesn’t help with this because Python is dynamically typed but static types could in principle. Is that what you mean?
I mean yes the python interpret is not great, but I'm talking about bugs created by accidentally passing a negative number where a positive one should go, using a tensor of the wrong size etc.
I don't have any ml experience with python only numerical libraries and I always have to be careful what I'm doing otherwise an experiment can go Kaput. Types stop that.
I think it's generally fair to assume that you don't become a Stanford CS professor by being bad at writing code and creating abstractions, and that the average professor (let alone one at a prestigious university) is more knowledgable than your average HN commentor
I feel like this is the exact opposite of the conclusion I've been coming to. In an age where anyone can vibe code stuff at the drop of a hat, I want the ability to assert guarantees/contracts at a high level, and then let AI work out the details. I want to force AI to work within the confines of an abstraction, not independently of it.
Precise specs must always be given, they cannot be reasoned into existence, because they are shaped by external considerations.
No matter how intelligent the executor is, without knowledge of the relevant external context, specs will always have to be clearly defined and given.
If I want to not care about implementation details on an era of AI written code, abstraction is exactly what I need.
We need abstractions more than ever. The real point is we should retire unnecessary abstractions.
A good abstraction is a single edged sword which simplifies the task. A mediocre abstraction is a double-edged sword. A bad abstraction is like a single edged sword with a restrictive handle and the sharp edge is facing towards you.
“First, an abstraction isn't just a way to manage cognitive load; it's a shared surface where application, reuse, and review attach.”
That is some seriously convoluted writing.
Agreed. Turned into some ai slop quickly.
This is such a poor idea. LLM Agents don't have taste (abstraction capabilities). They are kind of like water, they take the shape of the container they're in. They are so much powerful when you build your own framework they can work in.
That is not what the text is arguing for and it has many good ideas, dropping the abstractions but keeping the test harness is one of them. I've worked on many aspect of programming language implementation through the years and I would struggle with the abstractions of parser-generators and bytecode designed by committee like an enterprise Java developer struggles with the Hibernate ORM. Dunno what people are putting in their prompts but ask LLM to roofline and it will pursue that target like a bloodhound.
Abstractions aren't a burden though, they are more like a support. A good abstraction let's you write 10 programs in the time it would have taken to write one. Also, we now have "production" languages with generics and monomorphization so abstraction is easier than ever.
Djikstra has a great article as to why English language programming will always be second to a programming language. A "formal system" (edited from regular language) is the most precise way to specify things. An "abstraction" is basically a domain specific language.
The way I like to use LLMs is to build abstractions. I like to build mathematical code and ask the LLM to use it. I think that LLMs are very mathematical thinkers and they like it when they see Semigroup<T>. And it's not hifalutin either, I'm processing calculations along DAGs with a million+ nodes in milliseconds, thanks to abstractions
Many abstractions we encounter in software industry are poor abstractions which are counter-productive. I agree that we should drop those...
But we can't drop all abstractions. It's not physically possible. The AI agent is going to break logic up into files... The AI will create/impose an abstraction for each piece of logic whether we like it or not.
So saying "drop abstractions" is actually saying "let the AI decide what the abstractions should be" and unfortunately because LLMs are trained on average code, those abstractions are often pretty poor.
Coming up with the right abstractions is actually one of the main skills which AI hasn't automated and where the human brings the most value. It affects maintainability directly; including when using AI.
I've worked on projects with poor abstractions and ones with good abstractions using AI. The ones with poor abstractions require 10x to 100x more tokens and time to solve problems and implement new features. You're constantly fighting it to prevent it from coming up with nasty hacks and workarounds. Poor abstractions create the need for hacks and workarounds.
agents are unlike humans in many ways but they have their own limitations. it's magical thinking to believe that they are able to cope with zero software abstractions, especially as a codebase increases in size and complexity
and today's agents increase codebase size and complexity much faster than humans do via traditional software engineering methods, so this is likely to hit a wall faster than if human software teams were to do this experiment.
Just yesterday, for fun, I asked Codex to generate an “online e-commerce site” in ARM assembly. To its credit, it did actually generate a program that served up a webpage over a socket, but it really did not do what I asked. The program still pulled in a large number of shared libraries (something I’ve never really thought to try in assembly). And… it was not actually a functioning e-commerce site. Iterating on that initial prompt proved painful—the program was extremely brittle and now I had to review and talk about ASSEMBLY. Maybe we can dispense with abstractions someday, but holy cow, not yet. I would much rather write an e-commerce platform in C than try to vibe-assembly my way to one.
Enjoy setting thousands of dollars on fire for fun!
Following this argument to its natural conclusion, we should tell LLMs to generate the binary directly. It's still an abstraction, but it's the lowest level we can go to on current consumer hardware.
I don't think raw machine code (not even assembly) is the best form to reason about program logic. It just takes too many steps to execute sophisticated processes to keep them all in your head at once - or keep them all in an LLM context at once.
In two words, accept slop.
No.
I kind of angree with the anrgument that for an ideal AGI abstractions are not needed. It would write everything perfectly without even needing a programming language. It would know everything to the byte and would create perfect machine code meeting all the requirements as perfect as it’s possible.
But the reality is different. First of all LLM were trained on current data, with the abstractions we currently use.
Secondly they suffer they same problems we do, their context size is way too small to just simply ignore abstractions.
it sounds like a bad abstraction. ime llms are really bad at reasoning through efficiency and utilization problems. they fail at scheduling and bin-packing for the same reason. it's actually difficult, multivariate, and not great.
i would be shocked if a well designed framework were less efficient than "bespoke" clankercode for low-level work
This is probably true. Today. But I think we're finding the problems LLMs are solving to be both paradoxically predictable but surprising at the same time. Not saying you're wrong, but scheduling and bin-packing seem like cards that will eventually fall to a bespoke prompt.
I mean, the post was written by a Stanford CS lab, so I'm inclined to believe that they know what they're talking about and are not just bad at creating abstractions.
I know nothing about the PhD candidate and professor who co-authored this, but I don't think this statement makes sense in general.
Typically CS labs at universities contain people who have not spent a significant amount of time exposed to large industry codebases and the corresponding complexity. I think the post would have more credibility coming from e.g. the platforms team at a tech company with a monorepo.
There is something to be argued about industry vs academic experience but this post has nothing to do with large industry codebases
Well, they’re also probably not scaling up in the same way as any of the commercial AI offerings, let alone the frontier labs. Faculty at Stanford probably have some decent hardware to play with, but they do not have data centers. I don’t doubt that they know their way around CUDA/PTX, but it’s not clear how relevant their message is given that their research code is very likely not being deployed in production, or at scale.
sounds like it might help to know something about the phd candidate and professor
The machine learning community decided to choose Python as the main language, which costs probably tens of millions of dollars in bugs every year due to it not being type checked (astral sh is fixing this though)
Hasn’t mypy been a thing for years already?
The main problem with Python is that, if you’re not doing your work in highly optimized numeric libraries (which are usually written in C), then you’re paying terrible overheads to use the Python interpreter. Type checking doesn’t help with this because Python is dynamically typed but static types could in principle. Is that what you mean?
I mean yes the python interpret is not great, but I'm talking about bugs created by accidentally passing a negative number where a positive one should go, using a tensor of the wrong size etc.
I don't have any ml experience with python only numerical libraries and I always have to be careful what I'm doing otherwise an experiment can go Kaput. Types stop that.
why?
I think it's generally fair to assume that you don't become a Stanford CS professor by being bad at writing code and creating abstractions, and that the average professor (let alone one at a prestigious university) is more knowledgable than your average HN commentor
Professors have no reputation for writing good code. Academic code is fairly notorious for being under abstracted, hard to follow, messy, etc.
If a passerby told me this story I’d have the same guess though!
When will we learn?
but what have we learned? roughly quoting the last scene in "burn after reading"
cia boss: what did we learn, palmer? cia officer: i don't know, sir. cia boss: i don't fkin know either. i guess we learned not to do it again.