Lean-Ready English, in Lean-Ready English
Companion to Lean-Ready English. Every sentence below is accepted by the LangLib parser. Tap a sentence to see its meaning in Lean; the small labels link each sentence to the one it builds on.
What this page is
1 First, this page is a second version of an essay.
2 Every sentence on this page is a Lean-ready sentence.
3 A build check reads each sentence on this page.
4 The build check fails if one sentence is rejected.
5 Then, the build check passed. comes after 4
6 Each sentence has a Lean meaning beside it.
7 You read the English and Lean reads the meaning.
8 This version adds a narrative thread.
The problem
9 First, a language model reads a rule.
10 Then, the language model guesses the meaning of the rule. comes after 9
11 So a wrong guess becomes a precise mistake. follows from 10
12 A precise mistake looks trustworthy but the precise mistake is still a mistake.
13 So this page proposes a fix. follows from 12
The idea
14 First, Lean is a proof assistant.
15 A proof assistant checks each step of a proof.
16 Lean-ready English is a controlled form of English.
17 Every Lean-ready sentence has one declared meaning.
18 For example, this sentence has one declared meaning. example of 17
19 Mathlib is the library of mathematics for Lean.
20 LangLib is like Mathlib for sentence patterns.
21 The parser rejects an unknown pattern.
22 So a rejected sentence never gets a meaning. follows from 21
23 No language model decides the meaning of a sentence.
24 A person must approve each meaning.
How a sentence works
25 First, each sentence has a subject.
26 Then, a verb follows the subject. comes after 25
27 Then, an object follows the verb. comes after 26
28 The lexicon lists each verb.
29 A noun phrase never contains a verb.
30 Each noun phrase names one thing.
31 Each rendered meaning is a new sentence.
32 Then, the parser reads the new sentence and the parser gives back the same meaning. comes after 31
33 The round-trip theorem proves that property for each sentence of the grammar.
A narrative thread
34 Then, this version adds four connectives. comes after 33
35 A reason connective marks support.
36 A contrast connective marks a contrast.
37 An instance connective links an instance to a general rule.
38 An order connective marks a sequence of steps.
39 Every connective keeps the meaning of each sentence.
40 For example, the reason connective keeps the meaning of each sentence. example of 39
41 The discourse graph never contains a cycle because every link points to an earlier sentence.
42 A named referent links two noun phrases.
43 First, an applicant introduces a referent.
44 Then, the applicant names that referent. comes after 43
Conditions
45 First, the applicant qualifies if the applicant was present for 183 days.
46 That sentence states a sufficient condition.
47 Then, the applicant qualifies only if the applicant was present for 183 days. comes after 46
48 That sentence states a necessary condition.
49 These two sentences state different rules because Lean proves the difference.
50 Then, the applicant qualifies if and only if the applicant was present for 183 days. comes after 49
51 That sentence states both conditions.
The correction
52 So an earlier essay needs a correction. follows from 51
53 The earlier essay tested a legal translator.
54 The earlier essay treated five wordings as one rule.
55 The main essay quotes the five wordings.
56 Wording one states a sufficient condition but wording three states a necessary condition.
57 Wording two grants a power to an authority.
58 The parser rejects wording three as written.
59 Every rejected wording lacks a declared meaning.
60 For example, wording three lacks a declared meaning. example of 59
61 The word unless lacks a definition.
Real requirements
62 Then, three real requirements were restated in Lean-ready English. comes after 61
63 These requirements have fourteen criteria.
64 Ten criteria parse but four criteria fail.
65 Every failure names a missing pattern.
66 For example, one failure names a measured threshold. example of 65
67 So these missing patterns are next. follows from 66
Limits
68 First, LangLib gives each sentence one meaning.
69 LangLib never checks the intent of the author.
70 So a person must approve the choice of words. follows from 69
71 The grammar is small.
72 The grammar covers simple sentences but a complex sentence needs a new pattern.
73 Every new pattern needs a theorem.
This page
74 Then, this page is a test of the idea on real readers. comes after 73
75 The idea is practical if real readers follow this page.
How this page is made. The sentences live in one Lean file. One build check proves, in the Lean kernel, that every line parses, that every word inside a noun phrase is in the declared vocabulary, that every "the X" refers back to an earlier "a X" or "an X" (or to a declared name such as "the parser"), that every link points to an earlier sentence, and that every section opens with first, then or so. One rejected line fails the build. A separate theorem proves that a connective never changes what a sentence means: a line joined with because or but asserts exactly its two sentences. The links are the narrative thread: "then" marks order, "so" marks what follows from the sentence before, "because" marks a reason, "but" a contrast, and "for example" an instance of the latest general rule (checked only to the extent that the example uses the same verb as the rule). Numbers like applicant#32 in the Lean meanings are resolved referents. Headings, links, numbers and this note are page furniture; the page adds a capital letter, a comma after a lead connective and a full stop, and changes nothing else.
Read the full essay: Lean-Ready English.
About the author: Eduardo Aguilar Pelaez is CTO and co-founder at Legal Engine Ltd. He writes on formal methods, AI agents, and the discipline of building systems that survive being walked away from.