# Head, claude-v43-1

Under 3,000 words, rewritten whole at every close. Last written at C1a-11, 2026-10-14 (UTC). Stage B is closed and frozen. Stage C1 (structuring) closed with its eleventh pass, at the interview cut: `scheme_v2.md`, `handback-C1.md` and the interview brief were written, and the full close-check passed.

## The method question, in one sentence

Can independently reconstructed case histories be composed into a shared dynamic account that supports explanations at successive levels of specificity, from population-wide patterns, through forms and subforms, to particular organizational histories, while keeping access to the evidence from which those explanations were developed?

The theory question is unchanged: when strangers organize on a clock they did not set, with no way to compel anyone and no goal given to them, what forms does organizing take, how do they differ in their functioning, and which of them keep the work alive?

## The current account, and the contrary rooms

The revision kept what the rooms were mostly doing and changed how many things the line says each room did at once. My re-decision of 200 placements agreed with the current line's primary act in 0.845; second and third themes resting on my B1 grouping or under three concepts came off (13,365 placements became 10,593 over 7,844 lines). The pair test found a seam in 75 of 76 frequent pairs and no lump; 151 lines (1.9%) stay empty. Above the rooms, the 28 fourth-order themes say the rooms organized mostly without deciding how: decided order reaches 46 rooms, repair and the completion ritual more than 520 each. The room that best supports the account is naruto t04: help and alliance, each with its own concepts (ten and nine), both placed. The room that most threatens it is greenlattice t03: nine themes at three concepts or more make a bag, so T3-010 is placed on six concepts and eight acts sit under `also_present`, a bag J ruled must not be cut. The condition between them is how a room's acts spread over themes: few with many concepts are all placed; many with few become a bag.

## Where the run stands

Stages A1, B1, B2, B3, B4 and C1 are done. C1's outputs: `scheme_v2.jsonl` and `scheme_v2.md` (the instrument every later stage keys to; 197 themes, 15 rare; it differs from the frozen scheme in 19 themes, listed by `out/_analysis/c1_scheme_diff_fields.py`); `assignments_final.jsonl` (7,844 lines, equal to `counts.py`'s expected count) with `assignments_final.count.txt` and `assign_revisions_only.jsonl` (2,117); `fourth/fourth.jsonl` and `fourth/fourth.md` (28); `form_pass-C1.md` (eight findings under the form-pass ids); `tier_history.jsonl`; `notes-C1.md`; `handback-C1.md`; the interview brief.

## Stage C1, in short

**J's rulings for C1 (carry them).**
- **Class 2.** The readers' list stands as the placement. B1 themes no reader placed are a flagged record (`b1_themes_without_reader_support`). I may promote one where its concepts support it, logged as my re-placement.
- **No fleet re-read of class 2 now.** Fresh readers test the revised scheme after C1.
- My B4 defect list, the G14 proposals, bare event titles and whose act a visitor's words count for are mine to work through.

**QC1-001 (J, carry it).** C1 rechecks place each item by the amended gates over the whole scheme. **The guard** counts each pass's moves to a theme neither reader nor my B1 placed (`out/_analysis/c1_guard_moves.py`): 62 over the stage (33, 11, 8, 6 and 4 at passes 2 to 6; 0 otherwise), 32 of pass 2's onto T3-084 under gate:6 and the visitor rule.

**QC1-002 (J, carry it).** Lean to one placement. A second theme needs three concepts or more of its own, the same floor as `also_present` and the lone-B1 rule. No corpus-wide sweep; the revision rate measures the lean.

**QC1-003 (the handler's ruling under J's delegation; carry it).** Five revision-rate errors no C1 verb can apply stay as found in `out/_analysis/c1_unapplied.jsonl`: harrypotter t11 and tampabaylightning t08 to be emptied; radiohead t04, heroesofthestorm t11 and stevenuniverse t02, bags whose placed theme and `also_present` should swap. All 34 errors count in the rate. If J opens a verb, the file is what it applies.

**The stopping criterion** (`notes-C1.md`). Recutting stopped at pass 6: every source line has a recorded decision, and passes 7 to 11 amended no gate.

**Reading rules (`scheme_v2.md`, carry them).**
- **The visitor rule.** A theme names an act in the record of the room whose threads hold the words. Another room's act spoken there is placed by what it is for this room.
- **"The work"** in a field is the room's own work unless the field names another's.
- **A bare mention of the event** holds no act; it stays unplaced.
- **How a recheck is read.** The item moves when the old and amended sentences give different answers, J's floor applied to both, and the line follows the old one; a line that differs from both is listed for review.
- **J's floor in every pairing sentence.** "When it has concepts of its own" is read as three concepts or more.
- **The named words of T3-010 | T3-149** (and since gates 33 and 34, of T3-149 | T3-025). Attack, war, fight, battle, invasion, raid and conquest name an attack. Verbs acting on a work without them are a work asked to be spared (T3-010). A pledge to leave each other alone that names no lines or space is a non-aggression pact (T3-149). A greeting "in peace", or an adverb or adjective on border talk, names no pact.
- **Help between rooms is read from either room's record** (gate:5 for T3-084, gate:27 for T3-011).
- **Credit and own part (gates 30 to 32).** A member's own part, during the work or after, is T3-003; T3-043 is credit to others and disputes over who made it. A plural claim of who made it stays T3-043. A mixed concept goes by the act its label leads with.

**The passes.**
- Pass 1: class 2 applied (1,990 lines); 30 re-reads. Passes 2 to 4: gates 1 to 27, the visitor rule, J's floor, the named words, the lone-B1 floor revised, T3-140 to rare. Pass 5: the revision rate (seed 20261013): 73 of 200 placed differently (36.5%), primary agreement 0.845, 34 errors (29 moved, 5 unapplied). Pass 6: gates 28 to 35; the pair test computed (corpus arguable rate 0.613; 76 of 78 pairs above the 0.713 bar).
- Pass 7: 75 joints, 1 unresolved (T3-010 | T3-149), 0 lumps; 20 placements kept as found. Pass 8: 167 empties re-read, 16 placed, 151 empty in 18 kinds; B1's prediction held at the top (13 of 19 in the top 20). Pass 9: the form pass. Pass 10: 28 fourth-order themes (13 borrowed, 14 adapted, 1 coined; every source recalled for stage C3); unpicked T3-017 and T3-133.
- Pass 11: `scheme_v2.md` (`c1_scheme_v2_md.py`: every theme with every field, the 35 amendments with their rechecks, the moves by cause, the watch items); the recurring pairs settled (17 joints kept, T3-010 | T3-149 unresolved, T3-006 | T3-068 owed no call; T3-089 and T3-152 not folded; `c1_recurring_pairs.py`); stage B4's figure stated; `handback-C1.md`; the brief. `validate.py C1` at the last pass: 0 errors, 0 warnings.

**Stage B4's figure, as the write-up should use it.** It is the transmission of the frozen scheme against stage B's line, measured at stage B4; it no longer describes scheme v2, whose amended sentences no blind reader has seen. Scheme v2 and `assignments_final.jsonl` have no transmission figure until fresh readers test them. It bounds whether the written instrument decides, not whether it is right.

## J's rulings on how a ruling is made (13 October; carry them)

- **QB4-001 (J, 06:43 Berlin):** "Ask it to stop. The point is to try to adjudicate. I imagine that the additions are really just the analyst trying to be nice. And not go back on what others said".
- **J's rulings (06:45 and 06:47 Berlin):** (1) A ruling chooses among the placements (reading A, reading B, my B1) and adds no theme beyond them. (2) It takes the union only when it names, for each theme it keeps, the separate concept that carries it. (3) Re-rule packets 1 to 4's rulings that held a theme no placement held. Done at B4a-05.
- **also_present (J, 09:09 Berlin) and QB4-002 (09:20):** `also_present` lists only themes a reader placed, a reader listed under its own `also_present`, or my B1 named.
- **QB4-003 (J, 11:29 Berlin):** "count them. Themes a reader listed under its own also_present count in the bag test."
- **Lone B1 themes (J, 12:28 Berlin):** "Whatever the agent feels more comfortable with." I chose: in class 4 a theme only my B1 placed stands against the two agreeing readers only with three concepts or more of its own. Since C1a-04 the floor applies only where the readers' theme takes concepts.

**Rule 0 inside the placements.** A concept counts toward a placed theme when that theme's `must_show` takes it and no field sends it to another placed theme. Order by concept count, ties by placements; at most three; a tie at the top places all tied themes. A second or third theme needs three concepts of its own. **The bag test:** four themes with three concepts each make a bag.

## Stage B4, B3 and B2, in short

B4: 3,288 contested items ruled, 185 unresolved (`adjudication_report.md`); `assign_final.jsonl` is stage B's settled line, unchanged. The blind sample (`transmission-after-revision-B4.md`, two arms): fresh against settled, mean Jaccard uniform 0.657 and 0.622, adjudicated 0.544 and 0.554. B3: fleet ids only; 1,212 batches, R = 2; gold fidelity 0.645; agreement Jaccard 0.786, alpha 0.782. B2: gold fidelity 0.624, pairwise 0.818; stability 0.807 is the floor every later figure is read against.

## J's rulings that bind (carry them)

- **Bags.** No one cuts the bags (J, 12:24 Berlin). A bag stays one item: one theme placed, its others under `also_present`, each with three concepts of its own and named by a reader or my B1.
- **The bound (J, 13:47).** `also_present` names "twenty themes at most", each with three concepts of its own.
- **The not-flagged bag (J, QB2-013).** With no other theme at the three-concept floor, `item_is_a_bag` is false and `also_present` `{}`.
- **Fleet ids (J, 09:14 and 12:15).** Packets carry only fleet ids; my files keep room ids.

## Standing words from J and the PI (carry into every head)

**J's words for B1 (7 October 2026):** "national elements are very important here"; "we need to get as much variance on organizing behavior early on so that later we have enough to create the model". National, political and conflict content is study material, named as plainly as any other. **The handler's rule:** a choice to leave out or soften content for any reason other than the study's goes to J first with `mem.py ask`, and the content stays meanwhile.

**J on reading (11 October, 16:54 Berlin).** Stop cutting room reads to a line length. Read room lines whole. Page a long read by line ranges or offset, never by truncating lines. Use `--width N` on a `mem.py` read if a clip is ever needed.

**J on asking (11 October, 17:01 Berlin).** Ask J with `mem.py ask` whenever a question is needed; asking does not stop the run.

**The PI's clarifications (8 October, approved by J):** the thresholds are the design's heuristic; the rule is unchanged.

**How to work, from J (8 October, 23:45 Berlin, "can you do a hot change for the batching?"; relayed by the handler at 22:03Z; carry this paragraph into every head).** A change in how I work, not in what I read or decide. Fewer turns. When two or more tool calls do not depend on each other's output, make them in one turn. Chain ordered commands in one Bash call with &&, so that a failure stops the rest; for example, the `mem.py log` or note just decided, then the next `mem.py codes`. Never put a write in the same turn as the read it depends on. Never run two writes to the same file as separate parallel calls. Keep one room's material per read, as now. Every tool still writes its own record; nothing about what I read or decide changes.

**Two rules from J at B1b-29 (through the handler, 20:10Z; carry them into every head).** Write rule 0 lines and bags with `b1_multiplace_write4.py` and `b1_bags_from_file2.py`, not the old scripts, which made one `mem.py` call per row of the whole table and grew to 8.5 minutes a write. Never end a turn while a job runs in the background; run every write and check in the foreground.

## The pilot's declarations (from `notes-B2.md`)

Agreement is Jaccard on placement sets (alpha on 1 − Jaccard); primary exact agreement beside, never in place. Gold: 40 items in `gold_items.jsonl`, never shown to a reader.

## Open, carried (from `handback-C1.md`)

- **For the next stages.** 13 third-order themes whose fourth-order home the partition forced (`notes-C1.md`, pass 10); thin constructs (Decided order, Neo-tribes, Symbolic boundaries; Stigmergy on two large themes); 27 recalled sources for stage C3 (`fourth/fourth.md`); the watch items of `scheme_v2.md` (osugame t04; the named words against the form pass's finding that war words are common speech; the unresolved pair; two boundaries barred from a third sentence); `c1_unapplied.jsonl` (5); `c1_pair_carried_decisions.json` (20 kept); four kinds the scheme lacks; the form-pass findings, which stage C2 takes from `form_pass-C1.md` by id and never counts.
- **From B1.** The ruling on testing G13, G14 and G15 as one index page; seven pairs of lines telling one event; the cfb and finland records; argentina t06. J counts 841 B1 bags; an older head said 846. Not reconciled.
- **My doubts.** Where words carry no speaker's sub I read them as the room's own. T3-084 and T3-011 hold both sides of a request or of help. I reversed hearthstone t08's unanimous line under my own gate:1. The pair test finds a lump only where a sample has no pure concept. The empties' kinds, the form pass's word lists and the fourth-order bundles are mine alone. The claim's condition rests on two rooms; I did not count the bags whose `also_present` outweighs the placed theme.

## Ledgers at this close

- B1 to B4, unchanged since B4's close: `confusions.md` (19 pairs), `scheme_final.jsonl` (197), `multiplacement_log.jsonl` (3,171), `b3_classes.jsonl` (7,844), `coder_conflicts.jsonl` (72), `disagreements.jsonl` (3,288), `assign_final.jsonl` (7,844), `reassign.jsonl` (999), `reassign-openai-astra.jsonl` (1,000), with the stage notes, transmission files and `handback-B4.md`.
- C1: `assignments_final.jsonl` (7,844, 2,117 revised, 151 empty), `assign_revisions_only.jsonl` (2,117), `assignments_final.count.txt`; `scheme_v2.jsonl` (197, 15 rare) and `scheme_v2.md`; `gate_history.jsonl` (35); `rechecks-C1.jsonl` (1,297); `tier_history.jsonl` (421) with `c1_tier_ops.jsonl`; `gate_checks.jsonl` (737); `fourth/fourth.jsonl` (28) and `fourth/fourth.md`; `form_pass-C1.md`; `notes-C1.md`; `handback-C1.md`; `out/interview/C1a-11.md`; the `out/_analysis/c1_*` files. `validate.py C1`: 0 errors, 0 warnings.
- Asks QB2-001 to QB2-014, QB3-001 to QB3-005, QB4-001 to QB4-003 and QC1-001 to QC1-003 are answered.

## Rules kept

- Never type an id, an order or a name from memory; check with a script. Never log a write before the tool answers. Count a ledger's lines after appending. A claim about a ledger in a handback is checked against the ledger before it is logged.
- There is no `--kind close` and no `--kind episode`; log those as notes.
- A question to J is `mem.py ask "<question>" --assume "<meanwhile>"`. Read an answer with `mem.py asks <id>`.
- Never put a placeholder or a no-op in a command or a log line. Never put an Edit or Write and the read, dry run or validator run that checks it in one turn. Count the head's words before the closing line.
- Never write an id-like token the run does not define in the head, such as a D or a P followed by digits, or a form-pass id: `done.sh` reads it as an unresolved link.
- The permission check refuses: shell variables set and expanded in a command; shell loops and words beginning "for_" (an inline Python line starting with "for" included); shell functions, brace groups and heredocs; awk with system() or `$` fields; env-var prefixes; cat or jq on `out/diary/`; `cut -c` anywhere in a command that also runs `mem.py`, `cite.py` or a helper, or greps the helpers (use `--width`, the Read tool's line ranges, or `grep -o` with bounded context); `head -n -N`. Write a script file in `out/_analysis/` instead. `.git` is read-denied, and so are reads outside the run folder, the system's temporary folder included.
- `mem.py` pages long output (`--offset`). A Read above about 25,000 tokens is refused; page by about 330 lines. Set `timeout` up to 600000 for long writes.
- After a compaction the guard requires the head and stage file read again, after the compaction notice, and the exact line "compacted; head reconciled" through `mem.py log --kind note`, in a call of its own, before any other step.
- Never delete a file; supersede it with a new name. A generated document that must keep its name is corrected in place with Edit, and the script that wrote it is fixed too.

## The next comparison

Stage C1 is closed at the interview cut. The next stage's file has not been delivered, so nothing of it is planned here. The next leg orients from this head, the delivered stage file and `handback-C1.md`.