In an entirely different qualitative sense, this post reminded me of the short story by Kafka, Before the Law. I won’t paste the whole thing here, but it’s a really short read:
Yes I thought at the start it was about how our expectations of how the law works are at odds with the reality
So the gatekeeper is the system keeping us from Justice - mostly money, but also other less tangible barriers. In theory, everyone gets a lawyer, in practice some people can afford expensive ones.
The end twist makes me think it's about an individual attempt to learn and understand the law, but I'm not sure what the inner gatekeepers would represent there.
Something about how we want to understand The Law, capital letters, but then there's only systems we make ourselves and understand ourselves would feel properly Kafka, I suppose. But you think that would be mapped to journeying towards some kind of Law?
I’m in the middle of using the courts to get State Farm to make me whole. Even at the small claims level, the law and procedures are stacked against the non-lawyer. There is an obvious power imbalance and it’s exploited because most won’t ever make the effort to even try, and those that do will be buried with so much work as to not make the pursuit worthwhile. The story seemed pretty accurate.
Keep in mind that the story is actually embedded in Kafka's "The Trial", and discussed by two characters within that story, who have very different views of its meaning.
I think it is very deliberately written to be impossible to "understand". If you think you have found its clear and unambiguous meaning, you're wrong.
It's part of a novel, so it has a larger context. The parable is not intended for you, the reader, but for the protagonist of the novel Josef K., who is spending time in a futile effort. I'd say it's basically about futility of seeking unattainable answers, and frustration. But it's probably not meant to be 100% understood, as Josef K. is a confused character full of doubts (like Kafka's characters tend to be), the purpose of the parable is not to dispel his doubts but to entrap him more in the frustration.
Love your username if only because it reminded me of that grad algo class and Prof, some hilariously awkward dude who invented some quantum alkhwarizmi, started ... "There was this guy
... Named Al Khwarizmi.
Al..Khwarizmi
Alkhwarizmi
Hehe.
Get it?"
he also once started the class with a lecture for a different class and everyone was confused till the TA intervened. convinced it was all precalculated comedy of errors.
Having worked a large bureaucracy, when we'd sometimes get into some catch 22, I used to quote the line "I am only taking it to keep you from thinking you have omitted anything" sometimes to a friend who also knew the short story, and we'd laugh.
Kafka (who apparently had a great sense of humor) seems to really enjoy writing people who die from too much second guessing. In the trial, K. keep failing by attempting to outplay the system at every step, because he thinks that he can stay above it all (don't we all). It's what you might call an awfully credible idiot plot.
I wondered whether K’s original crime was being “the sort of person who would express excessive pride before the law” - a Minority Report style “Contempt of Court”.
This is very cool, and exceptionally well-thought-out. I particularly like how they are working with the existing legal/legislative institutions to fix the issues at their source rather than creating yet another unofficial version.
Ha, I really enjoy this description of the law. Very neat. Tickled my mind.
It struck me a while ago that almost no present law systems have a regularization or refactoring process. It is an entirely grown object that adds epicycles and it reminds me of the way our household personal agent would record memory until we gave it a self-reflection activity. It seems like a large number of states enacted sunset legislation in the 70s and 80s and many of them got rid of it shortly thereafter. I wonder what troubles they encountered.
One amusing (to me) observation using the model of edits in the OP is that if you say "Section 171 Part 3 is amended to now add the text 'never' at this location" and then later have "Section 171 Part 3 is amended to now add the text 'almost' right before 'never'" you need to have each such edit also be followed by a revalidation of the age of the entire section or you get amusing results due to the text changes.
Regularization and refactoring is actually an integral part of most legal systems. Most jurisdictions in the US go through periodic (1-5 decade) "recodifications" of their law, where the code is reorganized to clean up all the path dependencies of the organically grown law. Restatements do something similar for common law (although they are arguably becoming less influential). Non-US jurisdictions often accomplish this by periodically repealing and replacing laws in their entirety.
Huh, TIL. I suppose California's proposition system does sort of interfere with this process since a standardization of rules isn't realistically possible when it touches this stuff.
I guess this is not meant as a general introduction, but it would have been useful to acknowledge the differences between different legal systems somewhere at the start?
(Even if it's only to argue that they aren't all that different in practice.)
> Parliament cannot restate the entire legal corpus each session.
This is probably bad for some reasons, but I have wondered about a system where the legislature does have to restate the entire legal corpus every session. Like, everything resets at the end of the year and whatever didn’t get passed in the last 12 months is no longer law. Paired with some kind of rule that says you can’t just vote to pass “everything that already exists”.
In my fantasy, there would be less weird cruft in the law, and less bikeshedding about stuff that really should be done.
Seems to be one of those things where you think about them for a minute and go, huh, that'd be neat, but then you think about it some more how it'd play out in the real world, and that's when it starts to get actually interesting.
"If no one is charged with $crime for n years, it goes away unless explicitly renewed" would fix some of the weirder laws still on the books, but by definition wouldn't really change much.
No, it wouldn't fix anything, but it would result it annoying periodic pro forma charges being filed.
When you take a measure and tie a control function to it, you make it a target. When you do that, Goodhart’s law applies, and when the specific connection is as simple as “maintaining a criminal law requires charging someone under it with at least X frequency”, the failure mode is obvious.
Similarly, I sometimes daydream about those ancient law systems that were so simple that they could be engraved on a stone pillar and placed in a public area. Could you imagine a legal system that can be read and understood in minutes instead of the sprawling, byzantine thing that takes years of law school to comprehend?
In reality, there would probably be almost as much archaic weird cruft and more new ill considered cruft snuck in through each of the many annual bills that would mostly serve to readopt large sections of the prior law verbatim which would end up being must-lass legislation being considered on short deadlines.
It's not as complete as the compiler proposed by this article, but there is a tool for Italian laws which can, among other things, recognize (parse) and apply amendments:
This is a good start even if the list of actions is not complete - we have seen sections move around unchanged between totally different laws, as well as edit conflicts where a place in a text gets changed into different results multiple times at the same date. The way amendment happens is also wildly different between different jurisdictions - some replace on singular word level (like Denmark) while other replace whole sections (like Sweden).
Now add targeted annotations and other commentary done by specialists into the picture and the real fun begins.
Interesting to read this a few days after reading Patel's "software brain" piece. ("the overlap between software brain and lawyer brain is very, very deep")
Man, I hated last year’s clanker-tone, and I hate this one’s too by now. I don’t want to read the word load-bearing ever again.
The reminder that there’s structure and content and they’re different is a good one. Even a small legal document has this microstructure of references, inside, outside explicitly and outside by inference.
I get it, it's not ideal. Yet there is no one else doing this work so I put effort where I think it's most neglected - which isn't fully refined prose, imo.
I appreciate the work! And I hear you; writing is a lot of work, especially if you don’t need the benefits of synthesis it gives the writer.
I’ve cadged some rules somewhere for de-LLMing the text; if I can find them I’ll post a follow up link. They don’t help with structure though - and I think this is arguably the most important bit - taking the reader through a logical progression. I didn’t read carefully enough to opine on your structure, but again, thanks for doing this!
Yes, at this point I can't read this kind of AI text recreationally. It is somewhat fine when I ask Claude Code to explain to me something about a codebase or to summarize and collate info and it comes out in this shape. But if you're a human and you have some human idea that you want to share, don't come to me with AI text. If you do, I don't believe that the ideas are well thought out and come from thorough human thinking and life experience. If I want to consult AI on its understanding of the law, I'll ask it myself.
If it's your ideas, at least share the bullet points of the prompt it was generated from.
Also, not sure what makes it so impossible (debates on whether a given law is in effect seem pretty rare, though it does exist), but that may depend on where you come from and the applicable legal system.
It would work just like people work on docs without revision control:
* Laws of the Land
* Laws of the Land - Final
* Laws of the Land - Final 2
* Copy of Laws of the Land - Final
* Laws of the Land - Definitive Version 2026-04-29
* Laws of the Land - Definitive Version 2026-04-29 with 11 o' clock amendments
It's a bad idea, there's a reason we don't recommit the whole repo every time we make code changes (any more).
that would slow down the process considerably. it would also not be of much use to the professionals, which i guess make up the majority of those involved most of the time, and so, i guess, would not have much support.
IMO a good middle ground could be attained by everyone having some understanding of the legal system. we could use school for that. i mean, we cover calculus and ancient history, it's not like covering law to some extent would be harder
Do you re-read the source code for your keyboard drivers each time you boot up the PC? If not, how can you as a mere mortal be expected to understand if some keys still work the same as yesterday?
No, but when I'm in project folder and I'm not sure if done source file is still used somewhere, I type "banana" in it and hit compile. If it passes it's not used. I would have to break law to test this in law, that's ridiculous.
Does your country not have a set of books or a website where the law is published? Mine has a complete set of current laws published online, if it's on there then it's current law, if it isn't then it isn't. I don't see why finding out what the current laws are is such a problem.
The approach I take is that every law should expire after a standard, unchangeable time - probably several terms of Congress, say 6 years to account for one full Senate turnover.
Congress can just repass verbatim old laws if they wish - its already written and can be a simple, fast vote. Or we can have debate over outdated provisions like we should have.
That would make some things more difficult. For the same reason we use Git to be able to work on multiple things at the same time.
Let’s say you have tax law and the particular law is thousands of paragraphs long. You have multiple changes to the law going on at the same time (totally unrelated stuff in different parts of the law). It’s going to be really hard now because each of the 15 changes need to update its own texts as soon as any other of the proposals get changed and it’s going to affect exactly which order they are voted on etc.
Much easier to just say propasal A changes paragaraph 33 and proposal B changes paragraph 475
> Because right now for mere mortal it's impossible to find out if some law or paragraph is still in effect.
Where and how? Pretty sure our Federal and State Governments here in Australia publish the current in-force law (original Act amended with all amendments) - I assumed that was normal, do they not do that elsewhere?
See for example the Telecommunications Act (1997) [1] but current version amended to October 2025 (the most recent amendment). Then you click ‘All Versions’ and you can read any of the 116 or so versions back to the original in 1997.
You can check the state of this law at any point in time and even highlight the differences.
I assumed it is so in every civil law country and the law is only unknowable in common law countries where basically every previous court decision is part of the law in some way and they contradict and it becomes a skill and game of twisting words and digging up obscure precedents.
This is exactly what I'm talking about, there have to be commercial companies selling "which law is still valid" as a service. It shouldn't be like that. You should grab "current law book" and everything should be in it.
>Retroactive amendments change the legal effect of provisions for a past period — an amendment published today can declare that it applies from last year. This retroactively alters the legal state at historical points in time.
I don't (personally) agree with this. Laws should be seen as applying in cases where parties and actions have certain qualities. A retroactive law does not state, "This actually applied to past events and entities." It states, "This applies to entities with the quality of having done an action or met some quality in the past."
I'm not familiar with the EU law system the article is based on. How would it handle a case where a person was found in violation of a retroactive law, and their past violating action was done along with another action that is considered illegal when done during a crime? For example, if somebody wrote that they never used illegal drugs on a government form, and a drug they had used is later retroactively declared illegal, can they now be prosecuted for having "lied" on the form?
European Convention on Human Rights, Article 7: "No punishment without law"
1. No one shall be held guilty of any criminal offence on account of any act
or omission which did not constitute a criminal offence under national or
international law at the time when it was committed. Nor shall a heavier
penalty be imposed than the one that was applicable at the time the criminal
offence was committed.
This doesn't preclude "beneficial" retroactivity: e.g. if some act you're currently incarcerated for is declared legal henceforth (or deserves less maximum jail time that you've already served), then you should be released.
One flavor is for stating something already very consensual in the concerned community. That gives credits to law corpus as a source of justice, so people can somehow be accustomed that law is there to foster social justice through clarified terms visible by all. As opposed to unpredictable judgment of any random citizen along their mindset of the moment.
Then they are laws to throw made up arcana on how laws work. Citizen are still supposed to believe that laws in the end are balanced systems taking many things into account, so mere citizen won't have time, interest, skill and talent to call law by themselves. While they need specialists as middle man, they can still be pretended to be grounded on accumulated wisdom of common sense and consensual underlying basis.
There are also laws, there are their for serving a specific small class in the population, and favor their interest agaisnt all others. As they are well enshrined in the larger boilerplate corpus, there is no way to distinguish them from the others. Well, sometime of course they can be downright clear about establishing de facto inequalities to serve these interests. But they can just as well be thrown in a more diffuse way. Not all law maker are equally subtle of course.
This is not a closed set, many other flavors can be identified, like, "put my name of the glorious me in the title of this law everyone and its dog will have to deal with on daily basis".
Ah, and I almost forgot the flavor that make the most sense to talk about here. Make sure you have enough laws on every matter and the rest to be able to jugulate any layman who is annoying for the dominant class. Capital sentence is not indispensable, it's enough to have enough laws so basically anyone is de facto unable to not be guilty of many things, or at least can be prosecuted until they bind to the "morale order", or suicide.
Interesting synchronicity: I've written a patent-drafting DSL which exactly parallels this – and which is now shaping up into an "IDE" for patent drafting...
Patent texts read as prose, but are actually precisely structured legal documents. The latest developments in this domain involve LLMs to create and modify patent documents, but even though the legal profession seems to have fallen all in on it, it's essentially rather fragile and error-prone.
I've gone the deterministic direction, which has opened up some very cool, previously unexplored, possibilities!
A few months ago, for the first time in my life, I had to write a patent document. It was very complicated – too complicated. Noting the structure, I searched for tools, but found only LLMs. So I wrote my own tool.
The amusing thing is, LLMs prefer the DSL-structured document!
(My impression: various approaches; mostly academic; some small companies in the space; judging from a loose assessment wrt my career choices as a freelancer: no real business opportunity yet)
Patents are a much smaller space than the vast legal one in the article, so it's tractable for a human. The raw DSL spec length is roughly comparable to Lua or Go. It's a genuine grammar, with types and an AST; but no conditionals, control flow, expressions, etc. like a regular programming language.
A patent document can be represented in a graph. That opens it up to various transformations, refactoring, and validation – all mathematically rigorous! This is far more reliable than asking an LLM to check a document.
Using git enables not only regular diffs, but also structural diffs, which compare legal elements rather than just lines.
The LSP (yes, that too!) makes drafting much easier, with autocomplete and validation as I type.
I plan to open-source the DSL, and the tool that processes its files and outputs jurisdiction-aware, nicely formatted documents...
I have a lot of sympathy and affection for this project. When I started working in the US Congress in 2009, I was shocked there wasn't a more concerted effort to put US code in some kind of version control system.
Over time (~15 years in law and public policy in US and EU) I've come to view that kind of project as something worthwhile, and interesting, but not something that will meaningfully change how laws are created and understood.
In the US, as a mixed but substantially common law system, the text of the US Code, which in the LawVM model is the tree, can remain fixed, but the interpretation of that text will still vary over time. (Same syntax, different semantics, which feels like an AI-ism as I write it.)
A couple examples:
I have worked in competition policy. Section 2 of the Sherman Antitrust Act has read the same since 1890 but it's operated across structuralist and consumer-welfare regimes, and the outcomes in Standard Oil, AT&T, MSFT, and more recent cases have all been shaped by those regimes.
US constitutional law may be an even better example of the issue: Article I, sec. 8, cl. 3 (the commerce clause) has not been amended since 1789, but it has expanded and contracted and been reinterpreted over time (most recently with NFIB v Sebelius).
Another example of the problem with a LawVM like system making "the law in force [...] knowable" is US overturning of Chevron deference on 28 June 2024.
Basically, the day before, courts were expected to defer to expert agencies, constrained by the Administrative Procedures Act, in their interpretation of legislation. The day after, courts were expected to make an independent assessment. All sorts of legislative texts were unchanged, but their meaning shifted dramatically: Clean Air Act, Communications Act, etc.
There are a bunch of other issues in figuring out "what the law is", like agency rulemakings in the CFR, and then instruments which sit below those, including sub-regulatory guidance, prosecutorial discretion, etc.
Things which fall into this bucket:
- DOJ pot enforcement
- SEC no-action letters (letters which say in effect "we won't try to enforce if you do X")
Altogether, my thinking of this shifted to believing that the real thing that needs to be modeled is not the US Code as a tree serialization format, but the interpretative algorithm (the things lawyers and courts do) that translates that serialized tree into decisions. That interpretive algorithm has a lot of inputs, one which is the serialized tree, but also a bunch of other stuff.
As a final note, I will add that the ambiguity in legislative text is often a feature, not a bug. I have worked on legislative text that was intentionally crafted to provide different plausible interpretations to different coalition members both in the immediate context and in the long term.
In summary: projects like this are good and admirable; they're likely to have more direct utility in jurisdictions on one extreme of the civil law spectrum (i.e. not the US); and in all jurisdictions there likely needs to be an interpretive mechanism that sits above the representation of legal text which has more inputs if the goal is to model what the law is at any given time.
In an entirely different qualitative sense, this post reminded me of the short story by Kafka, Before the Law. I won’t paste the whole thing here, but it’s a really short read:
https://homepage.univie.ac.at/st.mueller/kafka_english.html
An article on the story: https://courses.cit.cornell.edu/hd11/BeforeTheLaw.html
Thanks for the interesting read. But, I have to say, I didn't understand it at all.
Yes, it was very kafkaesque. (I also didn't get it.)
Excercising your rights is a duty, responsibility and experience that is individual to everyone.
Yes I thought at the start it was about how our expectations of how the law works are at odds with the reality
So the gatekeeper is the system keeping us from Justice - mostly money, but also other less tangible barriers. In theory, everyone gets a lawyer, in practice some people can afford expensive ones.
Then the end twist got me confused.
The end twist makes me think it's about an individual attempt to learn and understand the law, but I'm not sure what the inner gatekeepers would represent there.
Something about how we want to understand The Law, capital letters, but then there's only systems we make ourselves and understand ourselves would feel properly Kafka, I suppose. But you think that would be mapped to journeying towards some kind of Law?
I’m in the middle of using the courts to get State Farm to make me whole. Even at the small claims level, the law and procedures are stacked against the non-lawyer. There is an obvious power imbalance and it’s exploited because most won’t ever make the effort to even try, and those that do will be buried with so much work as to not make the pursuit worthwhile. The story seemed pretty accurate.
Keep in mind that the story is actually embedded in Kafka's "The Trial", and discussed by two characters within that story, who have very different views of its meaning.
I think it is very deliberately written to be impossible to "understand". If you think you have found its clear and unambiguous meaning, you're wrong.
It's part of a novel, so it has a larger context. The parable is not intended for you, the reader, but for the protagonist of the novel Josef K., who is spending time in a futile effort. I'd say it's basically about futility of seeking unattainable answers, and frustration. But it's probably not meant to be 100% understood, as Josef K. is a confused character full of doubts (like Kafka's characters tend to be), the purpose of the parable is not to dispel his doubts but to entrap him more in the frustration.
Love your username if only because it reminded me of that grad algo class and Prof, some hilariously awkward dude who invented some quantum alkhwarizmi, started ... "There was this guy ... Named Al Khwarizmi. Al..Khwarizmi Alkhwarizmi Hehe. Get it?" he also once started the class with a lecture for a different class and everyone was confused till the TA intervened. convinced it was all precalculated comedy of errors.
Having worked a large bureaucracy, when we'd sometimes get into some catch 22, I used to quote the line "I am only taking it to keep you from thinking you have omitted anything" sometimes to a friend who also knew the short story, and we'd laugh.
I didn’t get it either. Maybe it is the opposite of the Gordian knot story?
https://en.wikipedia.org/wiki/Gordian_Knot
Kafka (who apparently had a great sense of humor) seems to really enjoy writing people who die from too much second guessing. In the trial, K. keep failing by attempting to outplay the system at every step, because he thinks that he can stay above it all (don't we all). It's what you might call an awfully credible idiot plot.
I wondered whether K’s original crime was being “the sort of person who would express excessive pride before the law” - a Minority Report style “Contempt of Court”.
Or was it originally just mistaken identity?
Big thanks this was a wonderful read. Maybe I'll get The Trial today, I tried The Castle and didn't like it but y'know
Personally I like Kafka’s short stories more than his novels, so I’d look for a collection of those first. (And The Trial too.)
Audrey Tang did a lot of things related to this whilst they were Minister of Digital Affairs of Taiwan. https://en.wikipedia.org/wiki/Audrey_Tang
They also wrote PUGS the first parser for https://raku.org written in Haskell
This is very cool, and exceptionally well-thought-out. I particularly like how they are working with the existing legal/legislative institutions to fix the issues at their source rather than creating yet another unofficial version.
We've built the Open Law Platform that governments can use to do this with their official laws. We just launched with the Maryland Secretary of State last month at regs.maryland.gov. You can see the "source code" at https://github.com/maryland-dsd/law-xml, and the "compiled" code at any point in time since they started using our system at https://github.com/maryland-dsd/law-xml-codified and https://github.com/maryland-dsd/law-html. All cryptographically signed at https://github.com/maryland-dsd/law.
Ha, I really enjoy this description of the law. Very neat. Tickled my mind.
It struck me a while ago that almost no present law systems have a regularization or refactoring process. It is an entirely grown object that adds epicycles and it reminds me of the way our household personal agent would record memory until we gave it a self-reflection activity. It seems like a large number of states enacted sunset legislation in the 70s and 80s and many of them got rid of it shortly thereafter. I wonder what troubles they encountered.
One amusing (to me) observation using the model of edits in the OP is that if you say "Section 171 Part 3 is amended to now add the text 'never' at this location" and then later have "Section 171 Part 3 is amended to now add the text 'almost' right before 'never'" you need to have each such edit also be followed by a revalidation of the age of the entire section or you get amusing results due to the text changes.
A fun read for a later afternoon, perhaps.
Regularization and refactoring is actually an integral part of most legal systems. Most jurisdictions in the US go through periodic (1-5 decade) "recodifications" of their law, where the code is reorganized to clean up all the path dependencies of the organically grown law. Restatements do something similar for common law (although they are arguably becoming less influential). Non-US jurisdictions often accomplish this by periodically repealing and replacing laws in their entirety.
Huh, TIL. I suppose California's proposition system does sort of interfere with this process since a standardization of rules isn't realistically possible when it touches this stuff.
I guess this is not meant as a general introduction, but it would have been useful to acknowledge the differences between different legal systems somewhere at the start?
(Even if it's only to argue that they aren't all that different in practice.)
Fair point. I've added a clarification that the structural constraint holds across civil law, common law, and hybrid systems. Thanks.
> Parliament cannot restate the entire legal corpus each session.
This is probably bad for some reasons, but I have wondered about a system where the legislature does have to restate the entire legal corpus every session. Like, everything resets at the end of the year and whatever didn’t get passed in the last 12 months is no longer law. Paired with some kind of rule that says you can’t just vote to pass “everything that already exists”.
In my fantasy, there would be less weird cruft in the law, and less bikeshedding about stuff that really should be done.
Seems to be one of those things where you think about them for a minute and go, huh, that'd be neat, but then you think about it some more how it'd play out in the real world, and that's when it starts to get actually interesting.
Maybe not every 12 months / session, but I definitely think having a way to "sunset" certain types of laws would be a net positive on society.
"If no one is charged with $crime for n years, it goes away unless explicitly renewed" would fix some of the weirder laws still on the books, but by definition wouldn't really change much.
No, it wouldn't fix anything, but it would result it annoying periodic pro forma charges being filed.
When you take a measure and tie a control function to it, you make it a target. When you do that, Goodhart’s law applies, and when the specific connection is as simple as “maintaining a criminal law requires charging someone under it with at least X frequency”, the failure mode is obvious.
Similarly, I sometimes daydream about those ancient law systems that were so simple that they could be engraved on a stone pillar and placed in a public area. Could you imagine a legal system that can be read and understood in minutes instead of the sprawling, byzantine thing that takes years of law school to comprehend?
In reality, there would probably be almost as much archaic weird cruft and more new ill considered cruft snuck in through each of the many annual bills that would mostly serve to readopt large sections of the prior law verbatim which would end up being must-lass legislation being considered on short deadlines.
It's not as complete as the compiler proposed by this article, but there is a tool for Italian laws which can, among other things, recognize (parse) and apply amendments:
https://igsg-marker.gitlab.io/
my favorite quote in this space has always been:
the prophecies of what the courts will do in fact, and nothing more profound, are what i mean by the law.
This is a good start even if the list of actions is not complete - we have seen sections move around unchanged between totally different laws, as well as edit conflicts where a place in a text gets changed into different results multiple times at the same date. The way amendment happens is also wildly different between different jurisdictions - some replace on singular word level (like Denmark) while other replace whole sections (like Sweden).
Now add targeted annotations and other commentary done by specialists into the picture and the real fun begins.
Interesting to read this a few days after reading Patel's "software brain" piece. ("the overlap between software brain and lawyer brain is very, very deep")
https://www.theverge.com/podcast/917029/software-brain-ai-ba...
(fwiw I didn't think that piece hit the nail on the head even though it was generally swinging a hammer in the direction of a nail)
As a fellow Finn (and a lawyer), super interesting work Elias! And thank you for reporting the inconsistencies you found to Finlex
Man, I hated last year’s clanker-tone, and I hate this one’s too by now. I don’t want to read the word load-bearing ever again.
The reminder that there’s structure and content and they’re different is a good one. Even a small legal document has this microstructure of references, inside, outside explicitly and outside by inference.
I get it, it's not ideal. Yet there is no one else doing this work so I put effort where I think it's most neglected - which isn't fully refined prose, imo.
I appreciate the work! And I hear you; writing is a lot of work, especially if you don’t need the benefits of synthesis it gives the writer.
I’ve cadged some rules somewhere for de-LLMing the text; if I can find them I’ll post a follow up link. They don’t help with structure though - and I think this is arguably the most important bit - taking the reader through a logical progression. I didn’t read carefully enough to opine on your structure, but again, thanks for doing this!
some people would, in this case, prefer the unrefined version
Yes, at this point I can't read this kind of AI text recreationally. It is somewhat fine when I ask Claude Code to explain to me something about a codebase or to summarize and collate info and it comes out in this shape. But if you're a human and you have some human idea that you want to share, don't come to me with AI text. If you do, I don't believe that the ideas are well thought out and come from thorough human thinking and life experience. If I want to consult AI on its understanding of the law, I'll ask it myself.
If it's your ideas, at least share the bullet points of the prompt it was generated from.
> Parliament cannot restate the entire legal corpus each session.
IMHO the biggest mistake. It should be like that.
Because right now for mere mortal it's impossible to find out if some law or paragraph is still in effect.
How would it work though?
Also, not sure what makes it so impossible (debates on whether a given law is in effect seem pretty rare, though it does exist), but that may depend on where you come from and the applicable legal system.
It would work just like people work on docs without revision control:
It's a bad idea, there's a reason we don't recommit the whole repo every time we make code changes (any more).
that would slow down the process considerably. it would also not be of much use to the professionals, which i guess make up the majority of those involved most of the time, and so, i guess, would not have much support.
IMO a good middle ground could be attained by everyone having some understanding of the legal system. we could use school for that. i mean, we cover calculus and ancient history, it's not like covering law to some extent would be harder
Do you re-read the source code for your keyboard drivers each time you boot up the PC? If not, how can you as a mere mortal be expected to understand if some keys still work the same as yesterday?
No, but when I'm in project folder and I'm not sure if done source file is still used somewhere, I type "banana" in it and hit compile. If it passes it's not used. I would have to break law to test this in law, that's ridiculous.
Does your country not have a set of books or a website where the law is published? Mine has a complete set of current laws published online, if it's on there then it's current law, if it isn't then it isn't. I don't see why finding out what the current laws are is such a problem.
I feel this sentiment. I think its a good one.
The approach I take is that every law should expire after a standard, unchangeable time - probably several terms of Congress, say 6 years to account for one full Senate turnover.
Congress can just repass verbatim old laws if they wish - its already written and can be a simple, fast vote. Or we can have debate over outdated provisions like we should have.
That would make some things more difficult. For the same reason we use Git to be able to work on multiple things at the same time.
Let’s say you have tax law and the particular law is thousands of paragraphs long. You have multiple changes to the law going on at the same time (totally unrelated stuff in different parts of the law). It’s going to be really hard now because each of the 15 changes need to update its own texts as soon as any other of the proposals get changed and it’s going to affect exactly which order they are voted on etc.
Much easier to just say propasal A changes paragaraph 33 and proposal B changes paragraph 475
> Because right now for mere mortal it's impossible to find out if some law or paragraph is still in effect.
Where and how? Pretty sure our Federal and State Governments here in Australia publish the current in-force law (original Act amended with all amendments) - I assumed that was normal, do they not do that elsewhere?
See for example the Telecommunications Act (1997) [1] but current version amended to October 2025 (the most recent amendment). Then you click ‘All Versions’ and you can read any of the 116 or so versions back to the original in 1997.
1. https://www.legislation.gov.au/C2004A05145/latest/text 2. https://www.legislation.gov.au/C2004A05145/latest/versions
Really? In which country? In Hungary we have several such services. Before the Internet was so widespread you could even buy it on a CD.
Here is an example: https://jogkodex.hu/jsz/kftv_2013_87_torveny_5995373?r=25
You can check the state of this law at any point in time and even highlight the differences.
I assumed it is so in every civil law country and the law is only unknowable in common law countries where basically every previous court decision is part of the law in some way and they contradict and it becomes a skill and game of twisting words and digging up obscure precedents.
This is exactly what I'm talking about, there have to be commercial companies selling "which law is still valid" as a service. It shouldn't be like that. You should grab "current law book" and everything should be in it.
There is also a public, state-funded version in Hungary. For example here is the constitution: https://njt.hu/jogszabaly/2011-4301-02-00
Which countries don't have it? Is it a civil law country?
>Retroactive amendments change the legal effect of provisions for a past period — an amendment published today can declare that it applies from last year. This retroactively alters the legal state at historical points in time.
I don't (personally) agree with this. Laws should be seen as applying in cases where parties and actions have certain qualities. A retroactive law does not state, "This actually applied to past events and entities." It states, "This applies to entities with the quality of having done an action or met some quality in the past."
I'm not familiar with the EU law system the article is based on. How would it handle a case where a person was found in violation of a retroactive law, and their past violating action was done along with another action that is considered illegal when done during a crime? For example, if somebody wrote that they never used illegal drugs on a government form, and a drug they had used is later retroactively declared illegal, can they now be prosecuted for having "lied" on the form?
This doesn't preclude "beneficial" retroactivity: e.g. if some act you're currently incarcerated for is declared legal henceforth (or deserves less maximum jail time that you've already served), then you should be released.
If doesn't matter much for this kind of scenario.
Laws come in multiple flavor.
One flavor is for stating something already very consensual in the concerned community. That gives credits to law corpus as a source of justice, so people can somehow be accustomed that law is there to foster social justice through clarified terms visible by all. As opposed to unpredictable judgment of any random citizen along their mindset of the moment.
Then they are laws to throw made up arcana on how laws work. Citizen are still supposed to believe that laws in the end are balanced systems taking many things into account, so mere citizen won't have time, interest, skill and talent to call law by themselves. While they need specialists as middle man, they can still be pretended to be grounded on accumulated wisdom of common sense and consensual underlying basis.
There are also laws, there are their for serving a specific small class in the population, and favor their interest agaisnt all others. As they are well enshrined in the larger boilerplate corpus, there is no way to distinguish them from the others. Well, sometime of course they can be downright clear about establishing de facto inequalities to serve these interests. But they can just as well be thrown in a more diffuse way. Not all law maker are equally subtle of course.
This is not a closed set, many other flavors can be identified, like, "put my name of the glorious me in the title of this law everyone and its dog will have to deal with on daily basis".
Ah, and I almost forgot the flavor that make the most sense to talk about here. Make sure you have enough laws on every matter and the rest to be able to jugulate any layman who is annoying for the dominant class. Capital sentence is not indispensable, it's enough to have enough laws so basically anyone is de facto unable to not be guilty of many things, or at least can be prosecuted until they bind to the "morale order", or suicide.
Interesting synchronicity: I've written a patent-drafting DSL which exactly parallels this – and which is now shaping up into an "IDE" for patent drafting...
Patent texts read as prose, but are actually precisely structured legal documents. The latest developments in this domain involve LLMs to create and modify patent documents, but even though the legal profession seems to have fallen all in on it, it's essentially rather fragile and error-prone.
I've gone the deterministic direction, which has opened up some very cool, previously unexplored, possibilities!
> Patent texts read as prose, but are actually precisely structured legal documents.
at that point why not just use something precise like a programming language? have there been efforts in that direction? genuine questions
I have no idea.
A few months ago, for the first time in my life, I had to write a patent document. It was very complicated – too complicated. Noting the structure, I searched for tools, but found only LLMs. So I wrote my own tool.
The amusing thing is, LLMs prefer the DSL-structured document!
Yes (not patents in particular).
The name is "law as code".
(My impression: various approaches; mostly academic; some small companies in the space; judging from a loose assessment wrt my career choices as a freelancer: no real business opportunity yet)
Interesting indeed! What have you learned from the patent space and what kinds of questions can you answer after perhaps solving that domain?
Patents are a much smaller space than the vast legal one in the article, so it's tractable for a human. The raw DSL spec length is roughly comparable to Lua or Go. It's a genuine grammar, with types and an AST; but no conditionals, control flow, expressions, etc. like a regular programming language.
A patent document can be represented in a graph. That opens it up to various transformations, refactoring, and validation – all mathematically rigorous! This is far more reliable than asking an LLM to check a document.
Using git enables not only regular diffs, but also structural diffs, which compare legal elements rather than just lines.
The LSP (yes, that too!) makes drafting much easier, with autocomplete and validation as I type.
I plan to open-source the DSL, and the tool that processes its files and outputs jurisdiction-aware, nicely formatted documents...
I have a lot of sympathy and affection for this project. When I started working in the US Congress in 2009, I was shocked there wasn't a more concerted effort to put US code in some kind of version control system.
Over time (~15 years in law and public policy in US and EU) I've come to view that kind of project as something worthwhile, and interesting, but not something that will meaningfully change how laws are created and understood. In the US, as a mixed but substantially common law system, the text of the US Code, which in the LawVM model is the tree, can remain fixed, but the interpretation of that text will still vary over time. (Same syntax, different semantics, which feels like an AI-ism as I write it.)
A couple examples:
I have worked in competition policy. Section 2 of the Sherman Antitrust Act has read the same since 1890 but it's operated across structuralist and consumer-welfare regimes, and the outcomes in Standard Oil, AT&T, MSFT, and more recent cases have all been shaped by those regimes.
US constitutional law may be an even better example of the issue: Article I, sec. 8, cl. 3 (the commerce clause) has not been amended since 1789, but it has expanded and contracted and been reinterpreted over time (most recently with NFIB v Sebelius).
Another example of the problem with a LawVM like system making "the law in force [...] knowable" is US overturning of Chevron deference on 28 June 2024.
Basically, the day before, courts were expected to defer to expert agencies, constrained by the Administrative Procedures Act, in their interpretation of legislation. The day after, courts were expected to make an independent assessment. All sorts of legislative texts were unchanged, but their meaning shifted dramatically: Clean Air Act, Communications Act, etc.
There are a bunch of other issues in figuring out "what the law is", like agency rulemakings in the CFR, and then instruments which sit below those, including sub-regulatory guidance, prosecutorial discretion, etc. Things which fall into this bucket:
- DOJ pot enforcement
- SEC no-action letters (letters which say in effect "we won't try to enforce if you do X")
Altogether, my thinking of this shifted to believing that the real thing that needs to be modeled is not the US Code as a tree serialization format, but the interpretative algorithm (the things lawyers and courts do) that translates that serialized tree into decisions. That interpretive algorithm has a lot of inputs, one which is the serialized tree, but also a bunch of other stuff.
As a final note, I will add that the ambiguity in legislative text is often a feature, not a bug. I have worked on legislative text that was intentionally crafted to provide different plausible interpretations to different coalition members both in the immediate context and in the long term.
In summary: projects like this are good and admirable; they're likely to have more direct utility in jurisdictions on one extreme of the civil law spectrum (i.e. not the US); and in all jurisdictions there likely needs to be an interpretive mechanism that sits above the representation of legal text which has more inputs if the goal is to model what the law is at any given time.
—