This revision repairs most of the specific problems I identified. The correcting findings now stand separately from the withdrawn propositions; external endpoints are declared; reconstruction scopes are distinguished; and historical claims carry attestation information.
I independently checked the eleven manifest hashes, the forty bundled first-seating text hashes, and the fifty quotations assigned resolved introduction dates against those bundled historical copies. Those checks pass. I haven’t independently verified their Git commits.
The strongest addition is the research-question layer. It gives an unsettled claim a next investigation: sources, instrument, procedure, and consequential outcomes. That makes the kernel more useful for advancing the argument.
Four things still need attention:
1. The review’s provenance is misattributed.
§6.2 credits DeepSeek with the five-point review reproduced from this ChatGPT conversation. If DeepSeek relayed or evaluated it, record that additional step while retaining the originating attribution. This is exactly the kind of provenance distinction the kernel should preserve.
2. Q8 reverses its own retreat condition.
It says the framework retreats if “Column C fills.” But #1602 defines Column C as evidence excluding an independently transmitted source. Filling it supports establishing the dependency.
The retreat condition should concern evidence defeating the proposed derivation—for example, contrary direction or a demonstrated independent source. The aconite result concerns direction; it doesn’t make filling Column C adverse.
3. One withdrawn “claim” remains a noun phrase.
`mo-1572-lexical-original` quotes:
“the instruments that read lexical distribution”
That identifies instruments, without stating the proposition withdrawn about them. The surrounding source reports their failure under controls. Quote the actual original inference if recoverable; otherwise mark its wording unseated, as you correctly do for two other originals.
4. Research outcomes need a named scope of consequence.
Q2 labels exclusion of Xenophon and Aristophanes a retreat, while correctly saying it would leave the four positions intact. That could instead establish a successful boundary. Each outcome should identify which claim or proposed extension advances, retreats, or remains unchanged. Otherwise a local result can again be substituted for the standing of the whole.
The one self-dividing maker line remains intact. The package now gives a reader substantially better means to inspect how it developed and investigate what remains open. The next repair is chiefly about making those investigative consequences as precise as the claim states.