▲ 72 points
back
64 comments
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.
Yes. Compare also https://www.khoury.northeastern.edu/home/cmartens/Courses/74... where the abstraction itself mechanically enforces certain guarantees and contracts.
The more semantically rich a programming language, the more it demands mathematical veracity.
The more a language is mathematically sound, the more its linguistic constructs converge to enforce same.
Programming languages which follow this path ultimately support similar capabilities; Applicatives, Functors, Monads, Monoids, and often meta-programming via higher-kinded types and/or intrinsic AST code generation.
> 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,
Exactly right!
More details in my comment here - https://news.ycombinator.com/item?id=49269323
See also, Software Abstractions: Logic, Language, and Analysis by Daniel Jackson - https://mitpress.mit.edu/9780262528900/software-abstractions...
I have the same thought on this. Having some abstraction where we have total control and we can make clear judgments is the perfect place where AI should. Removing this abstraction will just make things hard for us and just pray that all the guards around are sufficient.
> I want to force AI to work within the confines of an abstraction, not independently of it.
Yes, this is the way. We need programming languages that make this easier.
Just a few more months, and the industry will consider OOP the next big thing again. Interfaces as contracts between modules? Obviously superior!
formal verification? maybe theorem prover or hoare logic can come into play here.
“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.
Claude seems to love the word "surface". I'm almost surprised to not find the verb "buy" to mean "get" in this generation.
Agreed. Turned into some ai slop quickly.
"The question is whether letting it sit higher is a difference in degree or in kind, and we don't think it's in kind"
I was scratching my head reading it. The cognitive load was high with cryptic sentences like the above. AI-slop or not, there is no excuse for dense prose particularly if it is intended for a human reader.
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.
The post seemingly tries to get at this with its discussion of 'oracles', but the quality of its writing and argumentation does it no favors. I actually hope it was written by AI, because if not, the authors could stand to benefit from a Claude detox. It's rife with the signs of brainrot from LLM over-reliance.
> Precise specs must always be given
This is the paramount need with AI.
Earlier when we were writing code by hand, since both specification and coding were done by us we could easily iterate over them in our head and thus keep them in sync.
Now the only input is the specification and it must be precise (however incomplete) so that we can be sure that there is no ambiguity when AI interprets it. Iterate and refine the spec as needed and you are assured that the code does what you want it to do. Inform AI to enforce the invariants from the spec directly in code as asserts/contracts and you have guaranteed traceability through code. This is the "Correctness-by-Construction" (CbC) approach to programming.
It would be worthwhile for folks to study the classic specification languages like Z-notation/B-method/Alloy/etc. and learn to use those productively with AI.
Some resources:
VibeContract: The Missing Quality Assurance Piece in Vibe Coding - https://arxiv.org/abs/2603.15691
Specification-Driven Development as the Foundation of AI-Native Enterprise Software Engineering - https://arxiv.org/abs/2607.16680
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.
I'd say: A good abstraction is a formal and abstract statement about the nature of the underlying (sub)system. It is self consistent and allows an easier way to reason and interact with greater complexity underneath.
And the more you see an abstraction as a "wad of API calls and functionality", the more leaky the story gets.
Some systems are notoriously difficult to get this statement right because they are volatile and prone to changes. But I think as in live or in academia, our aspirations should be to strive for a better abstract understanding and keep building better abstractions.
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.
Abstractions are ultimately a tool for humans to manage complexity in a particular class of problem, and to reduce cognitive load. Good abstractions come from imaginative human brains with taste and lived experience. Abstractions are also mind constructs that often have analogies in the real world. Even at the level of the CPU circuitry, abstractions serve that role. An LLM can only spit out what was already invented by humans and stitch it together in a tasteless, erroneous manner.
Bruce Lee was ages ahead of its time with the Water phrase.
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.
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.
Software abstractions are a reification of problem domain understanding. They embody a solution within the mathematical constraints of a programming language such that the collaborations, when done properly, appropriately reflect said understanding.
Discarding abstractions is nothing more than rediscovery of a Big Ball of Mud[0].
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
If I want to not care about implementation details on an era of AI written code, abstraction is exactly what I need.
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!
Yes exactly.
"Why not bomb every square inch?", "Why would you ever let go of the machine gun trigger?", etc.
I assume there must have been moments in the history of war when such bad and desperate ideas were seriously proposed.
The fact that we're still in this profoundly ignorant era of LLMs is embarrassing.
> The job of abstraction as a cognitive offloader is starting to retire. Agents are taking that job.
Call me when the agents can take my high level directives and write assembly.
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.
Personally I think we will just become more ambitious with our abstractions since the LLMs can generate more complex stuff at the lower level. To take the article's examples, LLMs will write kernels that fuse 3+ layers of a model, so the abstractions can enable more complex forms of kernel fusion.
This holds half-true.
What we decided to do is essentially just hand-rolling PTX with the agents under a harness. If there is lacking some function, we write/take a CuteDSL kernel, export PTX on compilation, transform it with the harness and then feed it into the agents.
We beat cuBLAS, and we beat all of abstractions - Triton, Helion, TLX, Mojo, ThunderKittens, cuteDSL, etc on Hopper and Blackwell for all of our workloads over 3-4 weeks.
vLLM dropped torch.compile support because they realized that they programmers were good enough to just generate the Triton kernels directly for all the passes efficiently.
If you work with this for prod the writing is on the wall sadly. The abstraction layer is just really much lower if you want full perf.
But you end up writing your own abstractions in the process?
Many times I wonder if it's possible to collapse abstractions? The idea of making the whole stack simpler. I'm not overly optimistic of LLMs achieving this, but it they do, I will be paying more attention.
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.
Even with an ideal AGI you would need abstractions because in order to talk in specifics you would need an abstraction.
Reading this, a lot of the sentence construction felt like LLM slop and it seems like Pangram agrees. https://www.pangram.com/history/35257829-ad18-4128-9d61-fa03...
Adding slop text makes even the core idea feel like LLM slop, even if the idea was original thought by the authors.
I would also prefer if they add a heading to this article mentioning that they used LLM’s writing it rather than just not mentioning it anywhere.
The tone of the article reminds me of the OMeta effort, where a system would be described by concise mathematical rules and OMeta libraries would generate the concrete code (many times larger).
This is one of the better posts I've seen on HN. The comments that are putting it down are not from readers that engaged it a whole lot.
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.
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.
In two words, accept slop.
No.