Glonce

Lexifina describes a Git-style version history for the US Code

toolsThe Glonce editorsHacker News🔥59Glonce rating59Impact35Novelty40Relevance55Surprise35Players10Credibility60

Lexifina has published an essay arguing that lawyers should have what programmers have: the ability to point at any line and see who wrote it, when, and what the file looked like on any past day. Today, the post says, a lawyer reading a federal statute usually cannot do that without an afternoon in the notes at the foot of the section. The authors add that the gap matters more now because so much first-pass legal research runs through language models, which are often unable to use the proper lineage of a law.

The post explains why the US Code is hard to treat as a versioned record. Outside the titles Congress has enacted as positive law, the code is only prima facie evidence of the law, and the Statutes at Large govern where the two differ. The Office of the Law Revision Counsel publishes it as a series of release points, each current through a named public law, and keeps reorganising it; a bill to enact title 41 as positive law was introduced in September 2025. Which public law wrote a sentence is recorded in amendment notes, in prose. When a change took effect is often in a separate effective-date note, and where Congress said nothing, a law takes effect on the day it is enacted.

The worked example is the definition of a small business debtor in 11 USC § 1182(1). It was created in 2019 with the existing small business debt limit. The CARES Act raised the limit to $7.5 million in March 2020 as a temporary measure, later extended to March 2022. That increase lapsed on 27 March 2022. On 21 June 2022 Congress restored it for two more years, for cases commenced on or after 27 March 2020 and still pending, so the restoration reached back over the gap. The higher limit expired again in June 2024, and after the triennial adjustment in April 2025 it stands at $3,424,000. For a case filed in May 2022, the release point printed at the time gives the lower limit, and only a later Act changes the answer. The post's conclusion: a useful history must carry the text, the date each change took force, and the scope of any law that reaches back.

The post then cites two lines of research on models and time. A Technical University of Munich study from this year, by Max Prior, Andreas Schultz and Matthias Grabmair, found two failures in models answering questions on German statutes: they apply rules that were superseded after their training cutoff, and they prefer the newer provision even when an older one governs the facts. One model answered 24% of post-cutoff questions correctly on its own and 87% when the relevant statutory text was retrieved for it. The study's authors conclude that temporal validity has to be treated as a hard constraint. Other recent work, by Wei Fan and others, trains search agents to respect the date a question is about, with large gains in consistency.

Asking a model to write the old text does not fix this, the post argues. When the same Munich group had a model apply German amending laws to earlier versions, the output looked 93 to 99% similar to the official text, but only half of single amendments and a fifth of chains of amendments matched exactly, and experts found legally significant errors in the near misses. So the history has to be checkable.

Lexifina's own test: for each pair of release points from 2017 onward, the authors took the public laws enacted in between, carried out their amending instructions on the older text in date order, and compared the result with the newer official version. Of 365 changes across nine titles, 361 reproduced the next release point exactly. The other four are shown with a label saying how far they could be confirmed. Each change carries the law and the date it took effect: 261 took effect on enactment by default, 72 by an effective-date note and 29 by a date in the amendment itself. One change depends on an agency determination that may never come, and it is left out of the text rather than guessed.

What a reader gets: pick a date and see the text in force on it, with the words that read differently today marked. Each phrase is coloured by the Act that wrote it. A change to a defined term lists every other provision that uses it, and a reader can follow a section to see what has changed since their last visit.

The post closes with stated limits. The tool does not yet show a later law's reach back on the earlier dates it covers, so for a case filed in the 2022 gap it shows the lower limit without warning that the June 2022 Act changed the answer. Effective-date notes are read by pattern, and unusual wording falls back to the date of enactment. Each change links to the law's own words, and for anything that matters the original should be read. The authors say they are still thinking about the right form factor.

Key facts

  • Lexifina's essay proposes a version history for the US Code: choose a date, see the text in force then, with each phrase coloured by the Act that wrote it.
  • Replaying public laws' amending instructions in date order on older release-point text reproduced the next official release exactly in 361 of 365 changes across nine titles, for release-point pairs from 2017 onward.
  • The 11 USC § 1182(1) small business debt limit illustrates the problem: $7.5 million from the CARES Act in March 2020, a lapse on 27 March 2022, a retroactive restoration on 21 June 2022, and $3,424,000 after the April 2025 adjustment.
  • A Technical University of Munich study found one model answered 24% of post-cutoff questions correctly alone and 87% with the relevant statutory text retrieved; the same group found generated old versions looked 93 to 99% similar to the official text but only half of single amendments and a fifth of chains matched exactly.
  • Stated limits: later laws' reach back over earlier dates is not yet shown, effective-date notes are read by pattern with a fallback to the enactment date, and the original should be read for anything that matters.

Why it matters

The US Code is published as release points, with the authorship of each sentence recorded in prose notes and effective dates often kept elsewhere. That makes the question "what did the law say on this day, and who wrote this phrase?" slow for a person and unreliable for a language model. The post ties this to the growing use of models for first-pass legal research. Its small business debt example shows how the printed answer for a May 2022 case can differ from the correct one, because a June 2022 Act reached back over the gap. A history that carries text, effective date and scope of retroactive laws targets exactly that failure.

Who it affects

Lawyers who read federal statutes and need the text in force on a past date are the main audience. The post also points at people building or relying on language-model legal research, since the cited Munich study found models apply rules superseded after their training cutoff and prefer newer provisions even when an older one governs the facts. Readers who want to follow a section and see what changed since their last visit are named as users of the proposed system.

How to use it

The post does not name a product, give a public URL for the tool, or state availability or pricing, and it says the authors are still thinking about the right form factor. As described, a user would pick a date to get the text in force, see words that read differently today marked, see phrases coloured by the Act that wrote them, look up every provision that uses a changed defined term, and follow a section for changes. Each change links to the law's own words, and the authors say that for anything that matters the original should be read.

How solid is it

The headline result is the authors' own replay test: 361 of 365 changes reproduced the next release point exactly, with the other four labelled by how far they could be confirmed. This is self-reported, and the post does not list which nine titles were tested or what the four non-exact changes were. The effective-date counts (261, 72 and 29) sum to 362, and the post does not reconcile that with the 365 total. The Munich figures are quoted from a cited paper (arXiv:2605.23497, 22 May 2026), and the post says only "one model" for the 24% and 87% result. Of the Fan et al. work, the post reports only "large gains in consistency". The piece is an essay by a company, not an official release or an independent evaluation.

Risks and caveats

The authors list their own limits. The tool does not yet show a later law's reach back on earlier dates, so for a case filed in the 2022 gap it shows the lower limit with no warning that the June 2022 Act changed the answer. Effective-date notes are read by pattern, and unusual wording falls back to the date of enactment. One change that depends on an agency determination that may never come is left out of the text. Outside positive-law titles the Code is only prima facie evidence, with the Statutes at Large governing where they differ. No timescale for fixing the limits is given.