Compact grammar and design questions

14.1 A formal sketch

This is a structural sketch, not a full parser for inherited Yiddish syntax. Gender and case features are absent. Semantic subject number distinguishes singular and plural zey.

assertion        := [fronted_constituent] finite_clause
epistemic_head   := EF_agreement strength evidential+
strength         := signed_finite_bits | av_repetition | un_av_repetition
                  | shats(finite_number_or_range)
evidential       := =zen | =zog | =pon | =zikh
quantified_modal := (zol | vil | ken | darf) bound_degree
                  | modal_root (named_scale; numerical_measurement)
question         := freg question_content QUESTION_END
extended_question := freg lang question_content oys
frame_open       := context_operator [lang] [source_or_model_label]
context_operator := nem | tsit | tsit-genoy | tsit-fri
                  | tsit-iber | boy-mod | bild | volt
short_act        := (tu | heys | vintsh | fil | halt | banem
                  | zog-tsu | shpil | rik | tsurik) content SENTENCE_END
extended_act     := short_act_operator lang content oys
av_repetition    := av repeated n times, n >= 1
un_av_repetition := un- + av_repetition

State rules: opening a frame pushes a context. Oys pops the innermost still-open one. QUESTION_END closes its ordinary freg and any contained local children, restoring the parent; it does not close freg lang. Punctuation in a quoted/displayed child belongs to that child, not to an enclosing question. PARAGRAPH_END pops all ordinary local frames opened in that paragraph; a lang frame survives. A lang opener is allowed only at root or under another lang frame, not beneath an unclosed ordinary frame. A short act's ending also closes its children. SENTENCE_END clears the local zelb register. Square brackets expose grouping but neither extend scope nor independently pop the stack. A source voice shifts only through attributed speech or an explicit document voice declaration; loyt does not shift it.

The finite strength slot remains numerical and finite for empirical assertion. The modal-degree domain is selected by its root or named scale; it is not the EF domain. No person ending attaches to a quantified modal. Independently asserted subordinate clauses take their own epistemic heads. Nonassertive and stipulated content follows its act's rules; it is not a general escape from assertion marking.

14.2 A reader and writer checklist

1. Identify the act: assertion, question, quotation, command, wish, definition, model, or other specified act.
2. For an assertion, identify the exact proposition and its negation.
3. Use invariant articles, adjectives, and pronouns; identify roles through order and prepositions. Conjugate EF for the subject's person and number only.
4. State a finite bit value, or an explicitly numerical approximate value/range.
5. Choose the evidence route from the current speaking voice. Use an attributed voice frame, not mere source citation, to change that voice.
6. Put time and aspect in the event predicate, not into the bit value.
7. Give separately asserted subordinate clauses their own marking.
8. End each ordinary question at its own ?; retain enclosing frames. Mark paragraph boundaries for other local context frames. Add lang for extended frames and close them with oys. Never leave a redundant question closer that would now pop its parent.
9. When updating, keep the hypothesis fixed and distinguish the old value, increment, and new value.
10. For quantified modals, identify the degree, scale, subject, comparison, and conditions separately from EF confidence.
11. Resolve zey number and zelb antecedents; reset zelb at each sentence.
12. When coining a term, record its spelling, pronunciation, plural, intended sense, and design status.

14.3 The release standard

The adopted sentence engine combines invariant nominal morphology, personal zey with semantic number agreement, subject zikh and local zelb, finite speaker-owned log-odds, four evidence routes, the av repetition rule, quantified modal constructions, nonassertive operators, automatic question closure, paragraph expiration for other context frames, and lang extensions. The short decimal digits and modal scale anchors are the release conventions. The worked examples and longer translations follow Version 0.

There are two distinct levels of scrutiny. A sentence can be grammatically well formed while making a poorly supported claim. A translation can have perfect bit arithmetic while sounding awkward in its Yiddish substrate. Both need review, and success at one does not guarantee the other.

14.4 Questions for further study

Naturalness. Which auxiliary placements and compounds would a fluent Yiddish reader find plausible? Further literary review should focus on idiom, calques, and relative-clause readability. The current essay already applies the adopted genderless and caseless morphology.

Spoken usability. Can listeners reliably distinguish efst, eft, and efn at conversational speed, and can they count repeated av syllables? Does a spoken numerical range remain manageable?

Frame load. Paragraph expiration and lang blocks are now adopted. Test whether listeners hear paragraph boundaries, remember interrupted extended frames, and distinguish zelb sentence scope from broader frame scope. Test ordinary question endings against explicitly extended freg lang … oys. For other context frames, oys remains available to clarify an early ending.

Numerical honesty. Do speakers need interval-valued expressions more often than point estimates? The proposed shats forms should be tested without turning them into content-free hedges.

Lexical growth. AI, causality, calibration, and decision vocabulary needs repeated use before spellings and derivations are settled. Several terms in the vocabulary list intentionally remain proposals.

Perspective. Can readers reliably distinguish citing a source from entering its voice? Indirect reports do not implicitly shift the speaking voice. Test loyt source phrases against explicit tsit-iber voice frames, including a return to the narrator at paragraph end.

Translation reconstruction. Mandatory numerical commitments intentionally add precision. Compare independent numerical readings of the same source, checking voice ownership, H versus “possibly H,” and coherence across repeated propositions. An unknown-value translation escape hatch is not part of Version 0.

Modal scales. Test the ordinal anchors, task rubrics, and distinction between recommendation, desire, ability, and need. The construction is adopted; an empirically validated universal scale is not claimed.

14.5 What Moloch tests and changes

The Moloch specimen is a translation workshop as well as a reading text. It retains all eight sections and supplies the English alongside the Yuddish. Extended third-party quotations without verified reuse permission are replaced by labeled short summaries; they are not silently presented as translated quotations. The paragraph identifiers below refer to the translation’s source-block labels.

The Zey model, made explicit. Let N = 1000, unpolluted profit B = 1000 dollars/(farm·month), filter cost c = 300 dollars/(farm·month), and k be the number of nonfiltering farms. The translation reads the stipulated pollution damage as d = 1 dollar/(farm·month) per nonfiltering farm. Then a nonfilterer receives B − k·d, while a filter user receives B − k·d − c. At k = 1 these are 999 and 699; at k = 400 they are 600 and 300. When everyone filters, each receives 700; when nobody filters, each receives 0. These zeros are exact monetary quantities in a stipulated model, not forbidden probability endpoints.

A unilateral switch from filtering to not filtering raises the switcher's profit by c − d = 299, because their own new pollution costs them 1. It lowers the other 999 farms' profits by 1 each, making the total change 299 − 999 = −700 dollars/month. Moving the other way reduces the switcher's profit by 299 but improves total profit by 700. This is the precise externality behind the story. The source's reference to “$300 extra profit” describes the filter expense avoided; the net gain after the new pollution is 299. The translation preserves the 300 wording and explains the distinction here rather than silently correcting the author.

A useful Yuddish calculation display is:

Boy-mod Ozere-rekhen [B = 1000; c = 300; d = 1; Δgevin_Stiv = c − d = 299; Δgevin_al = c − N·d = −700.] oys.

Its equations are stipulated calculations. A claim that this model describes an actual lake still needs a finite numerical EF and evidence. No update in credence follows merely from computing a payoff difference.

Claims the grammar must not certify. In blocks 193 and 197, “51% of the power” is the source's rough threshold, not a defined universal theorem about coordination. Power need not be a single additive scalar; some arrangements need unanimity, others do not require majority coercive power. In block 294, being more intelligent does not by definition prove recursive self-improvement, explosive speed, permanent dominance, or alignment. Those are separate assumptions or forecasts. In block 295, “unconstrained by natural law” is left as the source's provocative phrasing; an empirical account should distinguish escaping a competitive pressure from violating physics. In block 151, the claim that a monarch has no incentives is an idealization the language can report, not something a compound such as onrayts-geber makes true.

Corpus coverage. Both longer essay bodies and the standalone Feeling Rational reading copies follow Version 0, including automatic question closure. Their source alignment is preserved; spanning frames carry lang, personal zey follows semantic number, and reconstructed quotations retain numerical EF. The translation collection explains selective modal-grade interpretations and the ozere / zey and naydik / nay vocabulary distinctions. The spoken sign particles, interaction operators, update notation, quotation suffixes, and technical vocabulary used in this corpus are release conventions. Items still explicitly labeled Proposed are outside those adopted rules. Further spelling and idiom review can improve the translations without changing their release designation.

14.6 Release conventions

EF belongs to the speaking voice, including reconstructed authors and quoted characters. Loyt names a source or standard without shifting voice. Singular personal zey shares its form with plural zey, distinguished by ef/efn. Subject reflexive zikh and sentence-local backward-reference zelb have separate binding rules. Ordinary context frames expire at paragraph end; lang explicitly extends them across paragraphs. Ordinary questions follow section 14.7. Quantified modal stems place degree under independently quantified EF, and word formation follows the layered policy.

The four modal roots and conventional grade anchors, the placement of lang, semantic number agreement for zey, and the spoken paragraph convention are part of Version 0. The scale-specific measurement constructor is available in technical notation. No singular ment, arbitrary indexed-pronoun paradigm, missing-credence translation marker, universal bit-valued utility scale, or implicit sentence expiration of nem is adopted.

When translating, preserve the proposition and speaking voice. Use personal zey where appropriate while preserving number. Add lang to every frame that must cross a logical paragraph boundary; use oys for an earlier ending within a paragraph. Introduce zelb only where its exact local antecedent is intended. Quantified modals add semantic information; they do not automatically replace every kenen or darfn. Review Latin, Hebrew, and gloss together. A change to paragraph layout can change scope and requires a scope check.

The reference and vocabulary list describe the same release. The vocabulary records definitions, forms, and translation examples in one alphabetical list. Translator’s notes explain interpretive choices and remaining questions. Any edits to a translated work must keep source alignment, Hebrew script, glosses where supplied, and vocabulary examples synchronized.

14.7 Question boundaries

An ordinary freg closes at its terminal ? or completed interrogative intonation. A surrounding frame remains active. Freg lang survives question and paragraph boundaries and requires oys. Other context frames retain paragraph scope unless explicitly extended.

Identify the frame each closer ends. An ordinary question needs no following oys; its gloss may mark AUTO-CLOSE. Retain closers for enclosing quotations, suppositions, models, and extended voices. Give each independent question its own freg, or deliberately choose freg lang for one extended inquiry. Quoted punctuation and enclosing frames still need their own boundaries.

These conventions apply to the reference, phrasebook, unified vocabulary, both aligned essays, and alternate Feeling Rational reading copies. The English source texts and reconstructed numerical commitments are preserved, and stable vocabulary anchors connect definitions to examples.