It seems like this is largely a matter of what you’re asking for. If you wanted to do more research into the state of the field, an AI might be pretty good at answering your questions.
Yes, telling it that you want parsimonious solutions that reuse existing libraries makes a big difference. Why would it even try to do that sort of stuff if you didn’t ask?
The original discussion about the project (https://news.ycombinator.com/item?id=49746163) is very weird. Lots of call-outs about how the author is some sort of celebrity and random accounts vouching for him, with little discussion on the substance.
I posted a sharp critique in the original discussion, aiming to be civil while critiquing the project. I may have been a bit terse, and would probably rephrase some of it now to avoid confusion, but I don't think I was ever outwardly disrespectful.
I received several very emotionally charged responses centered in the personal credentials of the author. They felt very out of place and did not engage substantively with any of the things I said. It was indeed very weird
The author, who I hadn't heard of before yesterday, actually seems like a cool dude. He was quite responsive, normal, and engaged with my feedback, which makes other random accounts being offended on his behalf all the more uncanny
Yes I think fundamentally the effort required to understand, write, and verify a formal specification is just way higher than is reasonable in most situations.
There are some cases where it is pleasingly simple - usually low level algorithms like compression, sorting, search etc. Basically things you'd find in leetcode questions.
Most software isn't like that.
I think the actual answer is just that the very latest models (e.g. Astra) are actually quite good at writing normal tests, and you can just skim them to make sure they're doing something sane.
Narrowly, I think you're spot on. The effort required to understand the machinery around formal verification is a function of the surface area of the thing being formally verified. Specifically, formally verifying the surface area of general purpose programming languages is difficult.
My approach with Hale (shameless self plug) is that the programming language itself first offers another strata of structure to program within, a type of graph. Once the structure of the program is expressed as a graph, understanding how formal verification works is a clean encapsulation of graph activities.
If you have to write LAWS.bend which is pure code describing the laws then it isn't basically like those old days of writing unit tests and that too tests first hence the TDD?
So what is the unique idea here except a vibe coded compiler that generates C and everything else is handled by clang+llvm?
From README.md:
The compiler (not kernel) is 99% AI-written and has not been fully audited yet.
Also, why the compiler is not written against and with LAWS.md so that no audit is required at all?
This is probably not necessarily true. “f: list[1 A] -> list[A] pure, worst-case time n log n, such that for all x and 0 <= i <= j < len(x), f(x)[i] <= f(x)[j]” is probably good enough for nearly everyone unless the program synthesiser is actively adversarial; probably 99.999% of the list-sorting in the world is done via standard library functions anyway, which suggests that people don’t much care exactly how it happens.
It will never tell you that what you’re building already mostly exists as work that you can build on.
It will if you remember to ask it. I've got into the habit of starting any new project with a session where I ask a search-enabled LLM to help me figure out what the prior art for a problem is. It's saved me quite a bit of time.
@LiamPowell the author is clearly aware of formal verification, they've written several implementations of dependently typed languages, and ... despite the presentation of their work, which has some obvious flaws (as can be judged by reception) ... their many comments indicate that they know what they are talking about.
Your post is setting up a strawman between automatic formal verification and formal verification using interactive theorem provers ... obviously there is a spectrum, and Ada/SPARK are navigating the space to try and automate much of the work required to automatically dispatch with obligations to prove (computable) properties about programs.
Bend2 is a QTT -- it's dependently typed, and comes from the lineage of systems which are focused on being expressive enough to formalize mathematics.
Of course you need to build a somewhat significant "standard library" of theorems, tactics (as metaprograms), etc ... to approach what is built into the compiler in Ada. These are different approaches with different trade offs.
Your post isn't clear, you don't go into any of these details ... why did you post this? Do you think this is clear writing?
the author is clearly aware of formal verification, they've written several implementations of dependently typed languages
I'm not familiar with the author, I just saw the language posted the other day. I'll add a note to the top.
These are different approaches with different trade offs.
Why would we want the tradeoff where the LLM has to write significantly more code and where the specification needs to be more complicated? If the author is aware of the state of the art then I think they made a poor choice, but that's not the point.
Your post isn't clear, you don't go into any of these details
Bend just serves as a useful example, my general point is about how people will vibe-code a solution without an understanding of the field, leading to worse results than if they spent a little while understanding the field and then vibe-coded their thing.
If you're going to insinuate that the author of Bend2 doesn't understand PLs and formal verification, you should do so with some proof and not a hot take dunk.
I think it's fine to critique the language and the approach without criticizing the author and I hate that this site has become Tech Drama News, like the worst parts of Twitter.
Yes, I didn't realise how much it comes off as a critique of the author personally when I wrote it. I have added a note to that effect to the top of the article.
Perhaps because their concerns are different? Similar reasoning: "why invest in dependently typed languages over SMT solvers? SMT solvers are automatic, and it is way easier to add annotations to my programs"
The answer is that these technologies are not universal in trade off space.
Of course, if you can have an LLM prove something by dispatch to an SMT solver ... you want to do that (and, indeed, LLMs do sometimes do this even when working on Lean proofs)
But the space of statements that you can prove using verification with SMT is smaller than dependently typed systems.
Also, Ada/SPARK are very good and robust -- they've been around for ~50 years ... but there's been a significant amount of work in formal verification since then, right? I would be more clear that Ada/SPARK are state-of-the-art for "mission critical software" (and have been for many years), but there's been a lot of research work in formal verification of low level programming since then (Low*, for instance)
The author of Bend has completely missed that this is the current standard in the field of formal verification, if they even know that this field exists at all. They have instead come up with this whole system requiring verbose specifications and even more verbose proofs. A little research before vibe-coding an entire language and compiler could have substantially improved the result because the author would have known what to ask for.
This example matters beyond Bend, vibe-coding makes it makes it far too easy to implement a design that’s horribly broken or decades behind the current state of the art because you can immediately get a result without ever having to do any research. If you ask a LLM for a language where it’s possible to prove that a function is formally correct by building up a proof from basic principles then it will happily do so, it will never stop to suggest to you that computers can already build complex proofs without the need for a LLM and eliminate 99% of the work. It will never tell you that what you’re building already mostly exists as work that you can build on.
---
That's why all your LLM requests to build something substantial should start with "run prior work research first". Of course, at some point everything converges (if we share our outputs open-source) and then we may have solid standard patterns and libraries and do not need to waste trillions of tokens globally to rebuild the same minor, fundamental things, each one in their silent little silo.
IF we share, it will be of course to the monetary detriment of LLM providers who will have less income overall, and of course now they can't repackage anymore all our collective input, thoughts, human 'thinking traces' that they collect in their meta-data, as their new 'innovations' any more to inflate IPOs / stock prices.
It is trying to please you, and it always determines that the way to please you is to fulfill the original, core request. Any caveats or first steps will always be secondary to the ultimate goal of “this person wants to do X, so I will do X.”
The only first step I have found somewhat consistently useful, because as we know LLMs do not behave consistently, is when doing tech troubleshooting I will go “look at documentation for X before answering” so that it will search manuals and such. Helps avoid speculation. But even then, it’s still not full proof.
Sidebar: this is one of the core problems of LLM’s currently. You are basically arguing with them to get them to behave a certain way all the time and it’s not always clear if they’re doing what they’re being told to do. Then add the compounding layer that the longer the conversation goes on, the more likely it is to misunderstand or just ignore things as it descends into context-length-induced madness
That's why all your LLM requests to build something substantial should start with "run prior work research first".
Yes, but I think there are incentives to not do this for many LLM providers. Doing prior-work research is slow (web searches aren't fast, LLMs are rate-limited or blocked from plenty of pages, etc.), and sometimes contradictory which annoys LLM users, many of whom like faster gratification cycles from the agent slot machine handle.
Also, writing a bunch of bespoke code instead of leveraging prior art makes a lot of users feel like they own something novel/big/important, and also poses a larger maintenance surface for the LLM to make future changes (which costs tokens).
I don't think there's, like, a conspiracy at LLM providers to set up system prompts/RAG/etc. to discourage research-and-use-prior-art-by-default approaches. Rather, OpenAI/Anthropic/Google/etc. are optimizing for real but sometimes misleading success metrics which often lead away from a research-first approach.
Pro: it can be an excellent way to learn if you realize good problems don't come easy, many such cases where I abandon something having learned from it and that's life
Con: the machine will tell you you have easily found a good problem, and engineered the perfect and necessary solution, if you let it
I think they know the standards of formal verification. They just surf the AI hype, whip up a verbose Python-like language that is worse than any existing prover language and have 20k bots star it.
The field in question is formal verification. It’s notable that those two words appear nowhere on Bend’s webpage or in its codebase. The developer has built an entire language around a field seemingly without realising that said field exists.
I checked the developer's X account, they have written numerous posts about formal verification, so this specific claim ("without realising that said field exists") seems to be false.
Some noteworthy lines from the README.md[0]:
[0]. https://github.com/bendlang/bend/tree/main
It seems like this is largely a matter of what you’re asking for. If you wanted to do more research into the state of the field, an AI might be pretty good at answering your questions.
Yes, telling it that you want parsimonious solutions that reuse existing libraries makes a big difference. Why would it even try to do that sort of stuff if you didn’t ask?
Because it should be smart. I mean, why would a human coder ever observe standard best practices unless the client specifically asks them to?
But it won't tell you if you don't ask. It won't tell you, "This approach is stupid, Ada SPARK exists".
The original discussion about the project (https://news.ycombinator.com/item?id=49746163) is very weird. Lots of call-outs about how the author is some sort of celebrity and random accounts vouching for him, with little discussion on the substance.
Not even the demo on that release works well.
The whole original conversation was very smelly from the very start, 20k stars included on the GitHub page with lost history.
We know what kind of turf that is.
I posted a sharp critique in the original discussion, aiming to be civil while critiquing the project. I may have been a bit terse, and would probably rephrase some of it now to avoid confusion, but I don't think I was ever outwardly disrespectful.
I received several very emotionally charged responses centered in the personal credentials of the author. They felt very out of place and did not engage substantively with any of the things I said. It was indeed very weird
The author, who I hadn't heard of before yesterday, actually seems like a cool dude. He was quite responsive, normal, and engaged with my feedback, which makes other random accounts being offended on his behalf all the more uncanny
Yes I think fundamentally the effort required to understand, write, and verify a formal specification is just way higher than is reasonable in most situations.
There are some cases where it is pleasingly simple - usually low level algorithms like compression, sorting, search etc. Basically things you'd find in leetcode questions.
Most software isn't like that.
I think the actual answer is just that the very latest models (e.g. Astra) are actually quite good at writing normal tests, and you can just skim them to make sure they're doing something sane.
Narrowly, I think you're spot on. The effort required to understand the machinery around formal verification is a function of the surface area of the thing being formally verified. Specifically, formally verifying the surface area of general purpose programming languages is difficult.
My approach with Hale (shameless self plug) is that the programming language itself first offers another strata of structure to program within, a type of graph. Once the structure of the program is expressed as a graph, understanding how formal verification works is a clean encapsulation of graph activities.
If you have to write LAWS.bend which is pure code describing the laws then it isn't basically like those old days of writing unit tests and that too tests first hence the TDD?
So what is the unique idea here except a vibe coded compiler that generates C and everything else is handled by clang+llvm?
From README.md:
Also, why the compiler is not written against and with LAWS.md so that no audit is required at all?
"Know what to ask for" is what will keep me with a job for a while longer, I guess.
The code itself is the most compact representation of the rules you want applied.
This is probably not necessarily true. “f: list[1 A] -> list[A] pure, worst-case time n log n, such that for all x and 0 <= i <= j < len(x), f(x)[i] <= f(x)[j]” is probably good enough for nearly everyone unless the program synthesiser is actively adversarial; probably 99.999% of the list-sorting in the world is done via standard library functions anyway, which suggests that people don’t much care exactly how it happens.
It will if you remember to ask it. I've got into the habit of starting any new project with a session where I ask a search-enabled LLM to help me figure out what the prior art for a problem is. It's saved me quite a bit of time.
Think you could argue that’s more LLM-assisted engineering than it is vibe coding.
It seems implicit in the article that the author should have remembered to ask, as part of the prior research.
@LiamPowell the author is clearly aware of formal verification, they've written several implementations of dependently typed languages, and ... despite the presentation of their work, which has some obvious flaws (as can be judged by reception) ... their many comments indicate that they know what they are talking about.
Your post is setting up a strawman between automatic formal verification and formal verification using interactive theorem provers ... obviously there is a spectrum, and Ada/SPARK are navigating the space to try and automate much of the work required to automatically dispatch with obligations to prove (computable) properties about programs.
Bend2 is a QTT -- it's dependently typed, and comes from the lineage of systems which are focused on being expressive enough to formalize mathematics.
Of course you need to build a somewhat significant "standard library" of theorems, tactics (as metaprograms), etc ... to approach what is built into the compiler in Ada. These are different approaches with different trade offs.
Your post isn't clear, you don't go into any of these details ... why did you post this? Do you think this is clear writing?
I'm not familiar with the author, I just saw the language posted the other day. I'll add a note to the top.
Why would we want the tradeoff where the LLM has to write significantly more code and where the specification needs to be more complicated? If the author is aware of the state of the art then I think they made a poor choice, but that's not the point.
Bend just serves as a useful example, my general point is about how people will vibe-code a solution without an understanding of the field, leading to worse results than if they spent a little while understanding the field and then vibe-coded their thing.
If you're going to insinuate that the author of Bend2 doesn't understand PLs and formal verification, you should do so with some proof and not a hot take dunk.
I think it's fine to critique the language and the approach without criticizing the author and I hate that this site has become Tech Drama News, like the worst parts of Twitter.
Yes, I didn't realise how much it comes off as a critique of the author personally when I wrote it. I have added a note to that effect to the top of the article.
Thanks. Read your update and I think this is an interesting conversation to have in the open.
Perhaps because their concerns are different? Similar reasoning: "why invest in dependently typed languages over SMT solvers? SMT solvers are automatic, and it is way easier to add annotations to my programs"
The answer is that these technologies are not universal in trade off space.
Of course, if you can have an LLM prove something by dispatch to an SMT solver ... you want to do that (and, indeed, LLMs do sometimes do this even when working on Lean proofs)
But the space of statements that you can prove using verification with SMT is smaller than dependently typed systems.
Also, Ada/SPARK are very good and robust -- they've been around for ~50 years ... but there's been a significant amount of work in formal verification since then, right? I would be more clear that Ada/SPARK are state-of-the-art for "mission critical software" (and have been for many years), but there's been a lot of research work in formal verification of low level programming since then (Low*, for instance)
---
That's why all your LLM requests to build something substantial should start with "run prior work research first". Of course, at some point everything converges (if we share our outputs open-source) and then we may have solid standard patterns and libraries and do not need to waste trillions of tokens globally to rebuild the same minor, fundamental things, each one in their silent little silo.
IF we share, it will be of course to the monetary detriment of LLM providers who will have less income overall, and of course now they can't repackage anymore all our collective input, thoughts, human 'thinking traces' that they collect in their meta-data, as their new 'innovations' any more to inflate IPOs / stock prices.
Or maybe it should be a flag to indicate that there might be more to think about before handing the task of to an LLM.
As effective as “make no mistakes.”
It is trying to please you, and it always determines that the way to please you is to fulfill the original, core request. Any caveats or first steps will always be secondary to the ultimate goal of “this person wants to do X, so I will do X.”
The only first step I have found somewhat consistently useful, because as we know LLMs do not behave consistently, is when doing tech troubleshooting I will go “look at documentation for X before answering” so that it will search manuals and such. Helps avoid speculation. But even then, it’s still not full proof.
Sidebar: this is one of the core problems of LLM’s currently. You are basically arguing with them to get them to behave a certain way all the time and it’s not always clear if they’re doing what they’re being told to do. Then add the compounding layer that the longer the conversation goes on, the more likely it is to misunderstand or just ignore things as it descends into context-length-induced madness
Humans dont let LLMs raw dog internet web searches.
Yes, but I think there are incentives to not do this for many LLM providers. Doing prior-work research is slow (web searches aren't fast, LLMs are rate-limited or blocked from plenty of pages, etc.), and sometimes contradictory which annoys LLM users, many of whom like faster gratification cycles from the agent slot machine handle.
Also, writing a bunch of bespoke code instead of leveraging prior art makes a lot of users feel like they own something novel/big/important, and also poses a larger maintenance surface for the LLM to make future changes (which costs tokens).
I don't think there's, like, a conspiracy at LLM providers to set up system prompts/RAG/etc. to discourage research-and-use-prior-art-by-default approaches. Rather, OpenAI/Anthropic/Google/etc. are optimizing for real but sometimes misleading success metrics which often lead away from a research-first approach.
Pro: it can be an excellent way to learn if you realize good problems don't come easy, many such cases where I abandon something having learned from it and that's life
Con: the machine will tell you you have easily found a good problem, and engineered the perfect and necessary solution, if you let it
I haven't seen a language launch this controversial on HN since V-lang in 2019.
Glad we finally have some new drama to follow, definitely more entertaining than AI news.
I think they know the standards of formal verification. They just surf the AI hype, whip up a verbose Python-like language that is worse than any existing prover language and have 20k bots star it.
This is the way to succeed these days.
I checked the developer's X account, they have written numerous posts about formal verification, so this specific claim ("without realising that said field exists") seems to be false.
Back in 2018 they were working on Formality, an Ethereum formal verification project. They are the Victor in this video about it: https://slideslive.com/38911748/introducing-formality
3 bend posts on the front page of HN. Astroturf