What this is
A drawing of a sugya's argument, and the 1742 sefer that says what to draw.
In short
Ramchal's Derech Tevunos gives every sentence of a sugya a label: what the sentence does to the one before it, what it is once you state it plainly, and how it is warranted. This site labels passages that way and draws the result, so the shape of an argument is something you can look at instead of hold in your head.
I built it because the labelling is enormous, patient, mechanical work, and that is the first kind of work machines got good at.
The sefer
Derech Tevunos, The Ways of Reason, is a manual for reading how a sugya moves. Chapter 9 gives the moves a closed vocabulary: 7 elements and 19 leaves, each one a thing a sentence does to an earlier sentence, each with a test that separates it from the label sitting next to it. Chapter 3 reduces any utterance, however it was phrased, to a subject and a predicate in one of 11 manners. Chapter 4 says how two statements can stand to each other, and when an apparent dispute is not one. Chapters 7 and 8 say what a proof has to accomplish to land, what a rebuttal does to the claim it hits, and where a claim sits when nothing has been established about it.
Closed vocabularies, a procedure, and a discriminating test for each label against its nearest neighbour. That is a specification, written 284 years before anything could execute it.
Why it went unused
The work is enormous. In chapter 10 Ramchal takes four short Aramaic sentences from Shabbos 5b and reconstructs what is underneath them: two claims, four syllogisms and an elucidation, roughly a dozen steps, none of them written in the Gemara and every one of them required. It is the most impressive passage in the sefer and it is also the bill. Four sentences took him pages. Nobody has the hours to do that to a daf, let alone a masechta.
What changed
The labelling stopped being expensive. The question a model answers here is narrow: what does this sentence do to the sentence before it? The answer has to be one of 19 labels, it has to name the sentence it acts on, and it has to survive Ramchal's own test for telling that label from its nearest neighbour. An answer outside the vocabulary fails validation and the file never draws.
A model still gets things wrong, so every passage here was checked by hand. The drawing is what makes checking fast: a sentence labelled wrong takes the picture with it, and you catch it before you have finished reading the row.
Why a drawing
Two reasons, and the second one is the honest one. A picture is the quickest way to check a labelling. And a great many more people will look at a well-made drawing than will ever read an argument about dialectical method. If this sefer gets used, it will be because somebody saw the shape of an argument, found it beautiful, and got curious about what produced it.
Where to start
The passages are worked examples, 5 of Ramchal's own and 4 opening sugyot. The sefer itself is here in Hebrew and English, all 11 chapters, with chapter 9 as the vocabulary the drawing reads. A file you label yourself gets drawn by the same code the shipped passages get.
The longer version of the argument on this page is the most underrated sefer I own.