BFM-decomp/docs/cookbook-index.md

# Cookbook index — find the entry by the SYMPTOM you are looking at

> **Generated by `tools/cookbook_index.py` — do not hand-edit** (R33). Regenerate after adding a cookbook section.
>
> `docs/matching-cookbook.md` is ~716 KB / 512 sections. Grepping it blind is how three P30 wave-1 agents each "discovered" an idiom that was already written down. **Start here, then read the section.** A section appears under every symptom it addresses.

**How to use:** name what you SEE in the diff (a stolen delay slot, an extra `la`, a swapped register pair, a `conflicting types` error), find that symptom below, read those sections first. If nothing fits, THEN grind — and add a section when you win.


## Start here — the symptoms that come up most, with the section that fixes them

> Hand-curated from what agents actually hit and, in several cases, RE-DERIVED because keyword matching alone did not surface the entry (P30 wave-2 feedback). If your symptom is here, read the named section before anything else.

- **wrong branch sense / arms swapped / `beq` where the target has `bne` (match_one prints this class as **BRANCH-POLARITY**)** → **§3-T4** (invert the source condition) + **§32**.2; for a trailing `return 0` vs an early return see the shared-ret0 note (cookbook L1344)
- **`conflicting types` on YOUR OWN function's definition, where the fleet canon is `(void)`** → **§73** (the PARAMS axis) + **§42** — keep the `(void)` signature and read the incoming arg via `register s32 a0v __asm__("$4")`; §42 is otherwise indexed only under regalloc
- **a shared global declared at a conflicting type (u8 vs s32, signedness) blocking your draft** → **§37** asm-label alias `extern T X __asm__("D_x")` — beats `*(T*)&X`, whose address-of perturbs regalloc. Works for FUNCTION definitions too (P30 wave 2)
- **target reuses ONE address register across two different offsets of the same global** → **§20** (the pointer-var-to-the-global bullet, cookbook L1907-1918) — take `T *p = &D_x;` and index off `p`, never the bare symbol twice
- **an extra `la` / the address hoisted into a callee-saved register across calls** → **§20** + `gcc-2.7.2-map/cse_expr.md` §H — `*(T*)&sym` force_regs the address; the asm-label alias (§37) keeps the direct `%lo` mem form
- **`andi $x,0xFF` folded away in your output but present in the target** → **§1/I2** + **§12** name the family; if the prescribed `& 0xff` at the use folds, hold the masked byte in a **u16** local so only a QI->HI extend survives (P30 wave 2, byte-tested)
- **`slti` where the target has `sltiu` (or vice versa)** → **§35** — a separate SIGNED int copy of an unsigned load keeps `slti`; chained bounds get range-folded, so write each bound as its own `if`/`goto` (**§21**)
- **you are about to hand-derive a body that some overlay already matched** → **§71** — grep the callees/globals for an already-matched SIBLING first; in wave 2 this alone produced iteration-1 MATCHes on 4 of 19 targets
- **gcc stole an instruction into a branch delay slot that the target leaves as `nop`** → **§5a** + the zero-byte `__asm__("")` fence (**§34** toolkit) — `reorg.c stop_search_p` halts the eager filler on an asm insn
- **`void` vs `s32` return — is promoting it byte-neutral?** → **NO, not always: §41d** (byte-proven; a `void` body with no `return` gains an instruction). P30 adds a second mechanism: an `s32` return keeps `$v0` live-out and blocks dbr from filling a loop-back delay slot
- **`match_one` says **SIZE-MISMATCH/short** and the target has a frame-pointer prologue (`addu $fp,$sp,$zero` / `21F0A003`)** → **The target is -O0.** Pass `--o0` to `match_one`/`rtu_match` — nothing else will ever match (§116: opt level is a property of the FILE). Known -O0 regions: boot, `ov_SC01_077_o0*`, the whale `_o0b`, and 0x8013B568..0x8013C98C — a target outside all four still needs checking
- **same instruction multiset, one contiguous window, loads/registers ROTATED inside it** → pure sched1 statement-order (`gcc-2.7.2-map/sched.md` §S1/§S4). **Cheapest lever: brute-force it** — N independent statements, script all N! orders through `match_one` (24 runs ≈ 2 min) instead of reasoning about `rank_for_schedule`
- **a two-constant `if/else` or `?:` result lands in **$v1** where the target reuses the condition's **$v0**** → **§76** — fold the condition into a NAMED local and overwrite that SAME variable with the two constants (an s16 temp, a fresh temp, an inner scope, and both ternary polarities all stay $v1; only reusing one `s32` local coalesces onto $v0). §76 reads as behemoth-only; it is not
- **a load HOISTED above a store (match_one may call it a WIDTH class)** → usually **§76 regalloc**, not a width bug: a missing WAR dependency lets sched2 hoist it. Same fix — one variable for the compare temp and the result
- **`ori $v0,0xffd8` in yours vs `addiu $v0,-0x28` in the target** → a NEGATIVE constant stored into an **unsigned** narrow local materialises via zero-extended `ori`; make the local **signed** to get `addiu`. It does NOT cost you the `lhu` on readback — gcc-2.7.2 emits `lhu` for any plain HImode load feeding an `sh` (P30 wave 4, byte-tested)
- **LENGTH-DRIFT −1 and the missing instruction is a reg-to-reg COPY in a `jal` delay slot** → a narrow **prototyped** param (plain ANSI `s16 arg3`) — §43 is indexed as the K&R-definition lever, but the ANSI form is the fix as often
- **your narrow load lost its load-delay `nop` right after an inline block move** → an `lwl/lwr`+`swl/swr` pair is a delay-slot **SPONGE** for the following load — the inverse of the §5a/§34 fence case. Align-1 4×u8 struct assign emits the inline form (§38 covers only the -O0 memcpy-call form)
- **your `e = param_1` copy VANISHED (target addresses every field off a copy at a join block)** → cse.c `make_regs_eqv` keeps a param copy only when the new pseudo's live range escapes the cse block AND outlives the param's last mention — reach that by REUSING the same variable in a later block. Copy-in-the-arms gets hoisted into the delay slot; copy-assigned-in-an-arm goes global allocno → callee-saved + 2 prologue insns
- **⚠️ the `.run/ghidra_c/<fn>.c` seed looks like a DIFFERENT function entirely** → it may be — overlays share VAs, so a seed can be decompiled from another overlay mapped at the same address. Trust the target `.s`, not the seed (P30 wave 4)

## By symptom


### delay slots & branches  (6)

- **§3-T4** — Branch polarity: invert the source condition to flip gcc's chosen branch  <sub>L90</sub>
- **§5a** — Cross-jump tail-merge — gcc collapses two byte-identical blocks the original kept separate (FIX FOUND)  <sub>L211</sub>
- **§3-D.** — THE CROSS-JUMP RATCHET (the sharpest new trap — `func_80131340` L-C)  <sub>L3508</sub>
- **§3-C.** — Inner-block declaration does NOT delay slot allocation — BYTE-REFUTED  <sub>L10065</sub>
- **§162** — CROSS-JUMP DIRECTION: the surviving copy is always the LATER one, so a BACKWARD `j` into a sibling arm is a source `goto` (P30 S48)  <sub>L11263</sub>
- **§162** — The cross-jump "CALL veto" is a COUNT law, not a CALL law (BOUNDS §88a; P30 S48, `func_80189540`)  <sub>L11299</sub>

### instruction scheduling  (19)

- **§3-T2** — Source statement order drives instruction scheduling  <sub>L78</sub>
- **§3** — When a diff is pure scheduling → decomp-permuter (harness built, Phase 6)  <sub>L107</sub>
- **§10** — Closing the regalloc/scheduling hard tail by hand (LZSS, Phase 7 session F — the full close)  <sub>L835</sub>
- **§25** — The "schedule" class is mostly COALESCING (pin-crackable), not scheduling; + the gate two-stage + h_exact over-counts ×134 (Phase 21, cont.6)  <sub>L2132</sub>
- **§3-The** — crack: the residual was gcc-2.7.2 COPY-COALESCING, not the instruction scheduler  <sub>L2144</sub>
- **§3-The** — genuine scheduler — `rank_for_schedule` (sched.c), for when it IS scheduling  <sub>L2160</sub>
- **§3-The** — genuine schedule WALLS (do NOT re-grind — stub)  <sub>L2169</sub>
- **§32** — The region-a CAMERA-GIANT idiom set: struct-base hoisting + 4 sibling levers (Phase 24 T7, Fable5-cracked on `func_80129CF8` 191 ins, match_one MATCH; transferable to the 6 sibling giants)  <sub>L2422</sub>
- **§34** — The `func_80138ED0` giant crack: gcc-2.7.2's **3-qty sort bug** + the **zero-byte asm allocation toolkit** + the **giv-init fence** (Phase 24 T5; Opus→close=21, Fable5→MATCH ×134)  <sub>L2454</sub>
- **§35** — The region-a sibling-giant harvest: difficulty ≠ $s-reg count (it's global-array hoisting) + the banking recipe + new loop idioms (Phase 24 T7, 5 parallel Opus-Max agents)  <sub>L2471</sub>
- **§3-A4** — SINK THE CONSUMER CALL INTO THE ARMS (the inverse of A1; `func_8016AB6C`, byte-proven)  <sub>L3444</sub>
- **§49** — The LUID DIAL: a zero-byte SCHEDULING dial (the sched.c analogue of §47) — `func_8017A4AC` (536 ins ×134), Phase 26 session 8  <sub>L3527</sub>
- **§66d-5** — `residual_class`'s "structural ⇒ permuter CPU is waste" is WRONG for schedule permutations (measured, Phase 29 SESSION-18)  <sub>L5491</sub>
- **§3-The** — attribution primitive (use this before calling anything a scheduling residual)  <sub>L6057</sub>
- **§3-The** — scheduling rules (refining §135-2 and §135-4)  <sub>L8909</sub>
- **Consequence** — for the family (a real scheduling decision)  <sub>L10092</sub>
- **§148** — The loop.c hoisting THRESHOLD is arithmetic you can compute, and the `?:` clamp that folds to MIN_EXPR (P30 S42, `func_8017C6F4`, 947 ins)  <sub>L10142</sub>
- **§3-A.** — `move_movables` hoists iff `threshold × savings × lifetime ≥ insn_count` — and you can read it  <sub>L10148</sub>
- **§16Z** — SHARPENS  *(sharpens §37 "the /s-DEP LATTICE", §136-13, §136-14, §16Xy, §162q, `gcc-2.7.2-map/sched.md` §64)*  <sub>L14337</sub>

### register allocation & pins  (42)

- **§10** — Closing the regalloc/scheduling hard tail by hand (LZSS, Phase 7 session F — the full close)  <sub>L835</sub>
- **Residual** — A — commutative `|`/`&`/`+` result lands in the wrong source-operand register  <sub>L856</sub>
- **Register-allocation** — ORDER (call-crossing $s0/$s1 swap) → FORCE it with register pins (byte-proven)  <sub>L1384</sub>
- **§3-The** — LOOSE-TYPING wall is real for narrow params (Phase 16, reconfirmed)  <sub>L1428</sub>
- **§3-THE** — CAP: at this tail the match_one→gate gap is the LOOSE-TYPING CALL-GRAPH wall (not cheap plumbing)  <sub>L1637</sub>
- **§22** — DEF-side loose-typing recovery + grinder blacklist (Phase 21)  <sub>L2005</sub>
- **§24** — The `ov_SC01_077_a.c` split-file vein: split-aware propagation, but loose-typing-gated (Phase 21)  <sub>L2085</sub>
- **§25** — The "schedule" class is mostly COALESCING (pin-crackable), not scheduling; + the gate two-stage + h_exact over-counts ×134 (Phase 21, cont.6)  <sub>L2132</sub>
- **§42** — The F-band ≤28 regalloc crack wave: register-pin/DENSITY levers beat the permuter (Phase 25 T7 F-band, 2026-07-10; Ultracode 9-worker wave, 4/9 banked byte-identical, 266 swept ×134)  <sub>L2900</sub>
- **§43** — The K&R s16-param definition DISSOLVES the "narrow-param wall" for by-value register args (Phase 25 task A, Fable5 crack of the 369-ins giant `func_80166994` ×134, 2026-07-11)  <sub>L3195</sub>
- **§45** — The flagship `func_80133CD4` crack (399 ins ×134): the merged-variable permutation-breaker + the 1-death local-alloc gate (Phase 25 task A giant escalation, Fable5 gdb-on-cc1, 2026-07-11)  <sub>L3298</sub>
- **§47** — The live-length SLIDER: splitting a global.c allocno-priority TIE with one zero-byte asm (Phase 26 session 8, Fable5 Max, byte-proven on `func_8017BEBC` 952 ins ×113)  <sub>L3365</sub>
- **§48** — The 12-core jr crack wave: the ALLOCNO-PRICING dials and the EBB rule (Phase 26 session 8, Ultracode, 9/12 MATCH first pass)  <sub>L3405</sub>
- **§3-A.** — ALLOCNO-PRICING DIALS — move a value into the register you want, byte-neutrally  <sub>L3413</sub>
- **§52** — The WALKER-FAMILY skeleton: 6 regalloc-order levers + a deeply-characterized intrinsic wall (`func_80178004`, 165 ins ×134; Phase 26, Fable5, 2026-07-15)  <sub>L3913</sub>
- **§52a** — The regalloc sibling wave: new levers + two new wall classes (cheap-Opus applying §52, 2026-07-15)  <sub>L3959</sub>
- **§52b** — Sibling wave 2: more de-pin levers, a third wall class, and the match_one→whole-binary gap at scale (2026-07-15)  <sub>L3997</sub>
- **§66d-1** — What transfers between giants is the LOOP, not the PIN  <sub>L5280</sub>
- **§70** — The giv-init base register: walk the PARAMETER, not a copy of it (Phase 29 SESSION-18, `func_801777BC`)  <sub>L5619</sub>
- **§71** — Before mapping a giant, look for an already-matched SIBLING beside it (Phase 29 SESSION-18, `func_8017D960`, 3,338 ins)  <sub>L5660</sub>
- **§72** — A `register __asm__` pin is a PREFERENCE, not a reservation (Phase 29 SESSION-18, `func_8017F510`)  <sub>L5726</sub>
- **§74** — Auditing a pinned draft: the §72 hazard is CALLER-SAVED pins spanning a call, and only the disassembly can tell you (Phase 29 SESSION-19, `func_8017D960` b2, 5 pins)  <sub>L5822</sub>
- **§76** — The allocno CLASS (local vs global) is the dominant regalloc lever, and C reaches it ONLY through declaration scope and variable reuse (Phase 29 SESSION-19, behemoth #3 `func_8017F510` 1,511 ins, 97 → MATCH, pin-free)  <sub>L6028</sub>
- **§78** — A LENGTH drift can be a register grant in disguise; and `fold` never leaves a literal first in an `|` chain (Phase 29 SESSION-19, behemoth #2 `func_8017D960` 3,338 ins, 1806 → 0, pin-free)  <sub>L6174</sub>
- **§76** — confirmed at scale, and a pin nuance  <sub>L6251</sub>
- **§80** — A do-not-re-buy entry is scoped to its BASE, not to the function; and the pin's hidden cost is an unconditional `qty_phys_sugg` (Phase 29 SESSION-19, `func_8017BF14` 45 → 0)  <sub>L6270</sub>
- **§3-The** — pin's hidden cost, with the citation  <sub>L6291</sub>
- **§83** — The parameterised-repeat law, the spill-area trap, and why a per-case edit cannot move a per-case symptom (Phase 29 SESSION-20, `func_80183814` 5,122 ins, cold-ish start → 36 structural / 99.3%)  <sub>L6412</sub>
- **§83c** — TRAP: a "dead local" in a prior draft may be gcc's OWN spill area  <sub>L6445</sub>
- **§86** — Pinned-exemplar templatability is a PER-FAMILY property, not a per-member rate; and the §42e pin guard is now over-conservative (Phase 29 SESSION-20)  <sub>L6589</sub>
- **§3-Two** — further notes worth keeping  <sub>L8248</sub>
- **§137** — REGALLOC-PERM is a TWO-COMPILE ARITHMETIC PROBLEM, not a permuter job  <sub>L9341</sub>
- **§3-The** — same swallow, twice more, in the integration spine  <sub>L9706</sub>
- **§3-B.** — A `?:` on MEMORY operands costs ~16 bytes of invisible frame; on REGISTER operands, zero  <sub>L10054</sub>
- **§3-D.** — A lone `$t8`/`$t9` in the target is RELOAD SCRATCH — reproduce the spill, don't pin the register  <sub>L10070</sub>
- **§3-C.** — A zero-byte ALLOCNO-PRIORITY slider  <sub>L10181</sub>
- **§150** — A register ROTATION across symmetric blocks is VARIABLE-IDENTITY evidence, not an allocator tie (P30 S43, `func_8017C6F4`, 947 ins ×4)  <sub>L10309</sub>
- **§152** — BYTE SIZE is the family key that name- and h_seq-grouping both miss (P30 S43, the 0xECC family: 1 crack → 12 overlays → 11,364 ins)  <sub>L10415</sub>
- **§155** — hi/lo literal scanning MUST track base registers (S45)  <sub>L10555</sub>
- **§158** — The RANGE-EXTENDER: a fourth zero-emission asm lever completes the allocno toolkit (P30 S46 tier-3, `func_8017CE58`, 733 ins)  <sub>L10709</sub>
- **Bonus** — facts worth keeping  <sub>L10754</sub>
- **§156** — THE PREFERENCE-DONOR MERGE: cross-region variable reuse is what fills a0-a3, and a call-arg use in ONE region steers the fill in ALL of them (P30 S46 tier-3, `func_80186E24`, 611 ins: 236-off "S11 regalloc-order" → MATCH, zero new pins)  <sub>L10769</sub>

### CSE / redundancy / rematerialization  (3)

- **§46** — The `func_80178D40` crack (890 ins ×134, the heaviest core in the game): four LOOP-STRUCTURE levers cheap-Opus found by reading loop.c/jump.c/cse.c (Phase 26 session 8, 2026-07-13)  <sub>L3313</sub>
- **§83d** — CSE's quantity budget is WHOLE-FUNCTION, so a local rewrite cannot fix a local symptom  <sub>L6452</sub>
- **§153** — THE ADDRESS-REMATERIALISATION LAUNDER: a third zero-emission asm lever (P30 S43, `func_8018D98C`, 710 ins)  <sub>L10463</sub>

### loops & induction variables  (9)

- **§3-T1** — Loop pointer: top-of-body for `addu` induction, not constant-folded `addiu`  <sub>L71</sub>
- **§34** — The `func_80138ED0` giant crack: gcc-2.7.2's **3-qty sort bug** + the **zero-byte asm allocation toolkit** + the **giv-init fence** (Phase 24 T5; Opus→close=21, Fable5→MATCH ×134)  <sub>L2454</sub>
- **§35** — The region-a sibling-giant harvest: difficulty ≠ $s-reg count (it's global-array hoisting) + the banking recipe + new loop idioms (Phase 24 T7, 5 parallel Opus-Max agents)  <sub>L2471</sub>
- **§46** — The `func_80178D40` crack (890 ins ×134, the heaviest core in the game): four LOOP-STRUCTURE levers cheap-Opus found by reading loop.c/jump.c/cse.c (Phase 26 session 8, 2026-07-13)  <sub>L3313</sub>
- **§66d** — The permuter⇄reader loop: alternate a random search with a byte-verified idiom, and let `residual_class` decide whose turn it is (Phase 29 SESSION-17)  <sub>L5245</sub>
- **§66d-1** — What transfers between giants is the LOOP, not the PIN  <sub>L5280</sub>
- **§70** — The giv-init base register: walk the PARAMETER, not a copy of it (Phase 29 SESSION-18, `func_801777BC`)  <sub>L5619</sub>
- **§145** — Three loop/combine levers from the S40 wave-2 drafters (16/16 match_one)  <sub>L9924</sub>
- **§148** — The loop.c hoisting THRESHOLD is arithmetic you can compute, and the `?:` clamp that folds to MIN_EXPR (P30 S42, `func_8017C6F4`, 947 ins)  <sub>L10142</sub>

### structs, block moves & memcpy  (35)

- **§3-T2** — Source statement order drives instruction scheduling  <sub>L78</sub>
- **§5** — Known hard-residual classes (instruction-identical, one byte-exact blocker)  <sub>L199</sub>
- **§13** — Add a location overlay — the canonical runbook (Phase 13; the Phase-15 fleet recipe)  <sub>L1056</sub>
- **§15** — Struct-heavy shared-core pipeline (Phase 16) — empirical determinations (S0)  <sub>L1313</sub>
- **§16** — Guided hand-matching the struct-heavy core (Phase 17 — beats the §15 brute-force)  <sub>L1335</sub>
- **§3-The** — crack: the residual was gcc-2.7.2 COPY-COALESCING, not the instruction scheduler  <sub>L2144</sub>
- **§28** — Banking a "close=0 gate-rejected" giant: the canonical-extern recovery (Phase 22 T1, byte-proven on func_8015126C ×134)  <sub>L2325</sub>
- **§28b** — The struct-walled close=0 giant (§28 case #3) IS bankable: the engine_types.h type-lift (Phase 22 T2, byte-proven on func_80156B74 ×134)  <sub>L2357</sub>
- **§30** — Fable5Max cracks §20/§10 "unsteerable" from the gcc SOURCE: store-vs-load is a `/s` aliasing flag, the def-side wall has a macro escape, + the birthing-boost (Phase 23, byte-proven on giant `func_8014EE14` 248 ins ×134)  <sub>L2382</sub>
- **§32** — The region-a CAMERA-GIANT idiom set: struct-base hoisting + 4 sibling levers (Phase 24 T7, Fable5-cracked on `func_80129CF8` 191 ins, match_one MATCH; transferable to the 6 sibling giants)  <sub>L2422</sub>
- **§38** — The WHALE `func_80144B9C` (770 ins): the -O0 struct-assign memcpy idiom + the -O0 reach-134 ×134 rollout (Phase 24 T7 §G, cheap Opus — no Fable5, no calls.c)  <sub>L2527</sub>
- **§40** — Structural families: the MECHANICAL symbol-remap (crack one exemplar → remap the rest, ~0 tokens) (Phase 25 T3/T7, 2026-07-08)  <sub>L2553</sub>
- **§41** — The DEF-SIDE canonical-sig wall: mechanically banking a drafted giant past `conflicting types` (Phase 25 T5b batch-2, 2026-07-09; `tools/canon_sig_reconcile.py`, byte-proven on `func_8013B274`)  <sub>L2710</sub>
- **§42a** — addendum — wave 2 (residuals + 29-100 band): iso-MATCH ≠ real-TU bank, the memcpy→struct-assign fix, +5 levers (2026-07-10b)  <sub>L2953</sub>
- **§44** — The Phase-25 cheap-Opus giant batch: 5 structural levers + the §43 extension (2026-07-11, 6 crackers over the frontier giants)  <sub>L3242</sub>
- **§46** — The `func_80178D40` crack (890 ins ×134, the heaviest core in the game): four LOOP-STRUCTURE levers cheap-Opus found by reading loop.c/jump.c/cse.c (Phase 26 session 8, 2026-07-13)  <sub>L3313</sub>
- **§3-Why** — 0/8 was structural, and predictable from two words  <sub>L4045</sub>
- **§54** — `--fix-def-sig`: the member's CANONICAL DECLARATION is a build step too (tiny-IMM mega-pools, +4,801, Phase 29 T6, 2026-07-16)  <sub>L4163</sub>
- **§56b** — PROPAGATING an h_seq giant: the exemplar's externs MUST be fleet-canonical, not the draft's types  <sub>L4232</sub>
- **§65a** — The blast-radius taxonomy (makes §61's law structural instead of remembered)  <sub>L5030</sub>
- **§66d-5** — `residual_class`'s "structural ⇒ permuter CPU is waste" is WRONG for schedule permutations (measured, Phase 29 SESSION-18)  <sub>L5491</sub>
- **§83** — The parameterised-repeat law, the spill-area trap, and why a per-case edit cannot move a per-case symptom (Phase 29 SESSION-20, `func_80183814` 5,122 ins, cold-ish start → 36 structural / 99.3%)  <sub>L6412</sub>
- **§84** — The DERIVED-OFFSET remap bug: a hand-computed literal that encodes the DISTANCE between two per-overlay symbols, and why `match_one` is structurally blind to it (Phase 29 SESSION-20, `func_8013D53C`)  <sub>L6472</sub>
- **§3-The** — construct  <sub>L6477</sub>
- **§91** — A structure-TRANSFER is only valid where the structure corresponds: the `--like` role trap (Phase 29 SESSION-21, `func_8012AAAC` ×137)  <sub>L6842</sub>
- **§3-The** — three-hypothesis trail, because two of them were wrong and the wrongness is instructive  <sub>L6871</sub>
- **§129** — Post-carve, `rtu_match`/`match_one` COUNT THE JUMP TABLE AS INSTRUCTIONS; and a carve must never be committed without its owner (P30 S28, `func_8013BD74`)  <sub>L8437</sub>
- **§129a** — the target instruction count is INFLATED after a carve  <sub>L8441</sub>
- **§3-The** — DEFINITION-side alias is the only escape when the fleet canon disagrees on a promoting param  <sub>L9470</sub>
- **§3-Two** — errors of mine, both instructive  <sub>L10006</sub>
- **§160** — THE REACH-15 WAVE HARVEST: an align-1 block move, and the instrument that called it a wall (P30 S47)  <sub>L10890</sub>
- **§16Xy** — SHARPENS  *(sharpens §136d-3, §37 /s-DEP LATTICE, §135-2, §136-13, §162q)*  <sub>L12297</sub>
- **§3-The** — `/s` drop clause is in ALL THREE dependence predicates, so the FIXED-ADDRESS STORE is what floats (P30 S48, `func_8017CA18`, ov_MAIN_012)  <sub>L12299</sub>
- **§16Z** — SHARPENS  *(sharpens §37 "the /s-DEP LATTICE", §136-13, §136-14, §16Xy, §162q, `gcc-2.7.2-map/sched.md` §64)*  <sub>L14337</sub>
- **§3-The** — ADDRESS-CLASS TABLE: which load/store pairs even REACH the `/s` clause (P30 S48 wave 4, `func_80185B44`, ov_SC03_014)  <sub>L14339</sub>

### types, signedness & load/store width  (34)

- **§3-I1** — Unsigned range check: `(x - lo) < (hi-lo)` → `addiu`+`sltiu`  <sub>L41</sub>
- **§3-I2** — Byte mask forces `andi` even after `lbu`  <sub>L47</sub>
- **§3-T3** — Types  <sub>L85</sub>
- **§7** — PsyQ SDK types & symbols (the library-call prerequisite)  <sub>L303</sub>
- **§14c** — Callee-signature-aware harvest — breaking the extern-type-conflict wall (Phase 15, T6)  <sub>L1213</sub>
- **Phase-20** — RESOLUTION — `tools/cast_call_sites.py` BUILT + the cap re-diagnosed (R14, byte-proven)  <sub>L1701</sub>
- **§23** — Giant `func_80153E00` cracked (scalar-data CAST); BUT most giant drafts are STALE+INCOMPLETE, and a diag MUST remove artifacts (Phase 21 — byte-proven + a self-correction)  <sub>L2032</sub>
- **§28b** — The struct-walled close=0 giant (§28 case #3) IS bankable: the engine_types.h type-lift (Phase 22 T2, byte-proven on func_80156B74 ×134)  <sub>L2357</sub>
- **§33** — Automating the giant decl-reconcile: `tools/reconcile_decls.py` (the DATA analog of §20's `cast_call_sites`) + a fleet-majority type oracle (Phase 24 T7b, byte-proven on `func_80129CF8`)  <sub>L2434</sub>
- **§40a** — The DECL-RECONCILE pass (type-lift so remapped drafts compile in the sibling TU) (Phase 25 T7.2, 2026-07-08)  <sub>L2601</sub>
- **§41** — The DEF-SIDE canonical-sig wall: mechanically banking a drafted giant past `conflicting types` (Phase 25 T5b batch-2, 2026-07-09; `tools/canon_sig_reconcile.py`, byte-proven on `func_8013B274`)  <sub>L2710</sub>
- **§41d** — `void`→`s32` is NOT always byte-neutral: gate the RAW draft FIRST (Phase 26 session 6, byte-proven)  <sub>L2878</sub>
- **§42b** — addendum — wave 3 (Max, 2026-07-10c): THE STALE-OBJECT GATE TRAP + the read-global `&`-cast drift + fix  <sub>L2994</sub>
- **§43** — The K&R s16-param definition DISSOLVES the "narrow-param wall" for by-value register args (Phase 25 task A, Fable5 crack of the 369-ins giant `func_80166994` ×134, 2026-07-11)  <sub>L3195</sub>
- **§3-C.** — TYPE- AND SHAPE-DRIVEN CODEGEN (the C type literally selects the addressing mode)  <sub>L3484</sub>
- **§56b** — PROPAGATING an h_seq giant: the exemplar's externs MUST be fleet-canonical, not the draft's types  <sub>L4232</sub>
- **§64** — The §20 type-lift's three laws: fold the tagged typedef, check VISIBILITY, and strip only what is TEXTUALLY IDENTICAL (`lift_types.py`, Phase 29 SESSION-14, 2026-07-23)  <sub>L4908</sub>
- **§64a** — VARIANT types: UNIQUIFY the camps, do not reconcile them (`uniquify_type.py`, Phase 29 SESSION-14)  <sub>L4957</sub>
- **§65g** — Where the cheap levers STOP: the local-type and in-TU-self-decl classes did not yield  <sub>L5139</sub>
- **§68** — A comment-only line halted the extern scan, and the skip label blamed the type cap (Phase 29 SESSION-18)  <sub>L5524</sub>
- **§73** — A def-side self-decl conflict has TWO axes: RETURN (fleet widen, T2) and PARAMS (casts at each use, T0). Diagnose which before paying for the expensive one (Phase 29 SESSION-19, `func_8014F3E8` + `func_8014D4C0`)  <sub>L5783</sub>
- **§75c** — Class-B's remedy is the FULL §17a-1 PAIR (decl **and** call-site cast); a decl-only fix moves the error and looks like a new wall (Phase 29 SESSION-19, `func_8012F14C`)  <sub>L5997</sub>
- **§3-The** — defect this exposed: a shared type that is present but invisible  <sub>L6354</sub>
- **§92** — Conforming a DECLARATION: pointer changes are caller-neutral, scalar-WIDTH changes are not (Phase 29 SESSION-21, `tools/conform_decls.py`)  <sub>L6891</sub>
- **§94** — A family sweep's `0/N` is a TYPE-CARRY failure until proven otherwise: lift the exemplar's local types, ALL of them, transitively (Phase 29 SESSION-21, `func_8016B6BC` 0/137 → 137/137)  <sub>L6947</sub>
- **§100** — Prefer the DRAFT-LOCAL fix: a type only one function uses belongs in its BODY, not in a shared header (Phase 29 SESSION-22, `func_80175DA8` 0/137 → 137/137)  <sub>L7216</sub>
- **§120** — Uniquify draft-defined TYPE names; and check which of N staging sites you actually patched (Phase 29 T93)  <sub>L8026</sub>
- **§3-Do** — NOT "strip the duplicate typedef" — it breaks the extern that uses it  <sub>L8037</sub>
- **§3-The** — type-form rules  <sub>L8879</sub>
- **§143** — `cast_call_sites` read a RETURN STATEMENT as a prototype and deleted it. A 0/39 sweep became 18/39. (P30 S40)  <sub>L9831</sub>
- **Then** — propagation returned 0/137 TWICE — both times a missing TYPE  <sub>L9997</sub>
- **§154** — Reading a disc payload: the module-id word, static base derivation, and "type 1 = uncompressed overlay" (P30 S44)  <sub>L10519</sub>
- **§3-C.** — PAC type 1 = the same payload class as type 4, just NOT compressed  <sub>L10545</sub>
- **§155b** — check the TYPE your oracle returns before comparing against it (S45 p5)  <sub>L10584</sub>

### declarations, prototypes & K&R  (57)

- **§3-T4** — Branch polarity: invert the source condition to flip gcc's chosen branch  <sub>L90</sub>
- **§8c** — Splitting a TU means rebuilding its DECLARATION ENVIRONMENT, not moving text (Phase 26 session 6)  <sub>L437</sub>
- **§8d** — Templating a body INTO a TU must not CHANGE its declaration environment — demote the carried data externs (Phase 26 session 8, byte-proven on `func_8015AE2C` ×133)  <sub>L483</sub>
- **§13** — Add a location overlay — the canonical runbook (Phase 13; the Phase-15 fleet recipe)  <sub>L1056</sub>
- **§14c** — Callee-signature-aware harvest — breaking the extern-type-conflict wall (Phase 15, T6)  <sub>L1213</sub>
- **§3-The** — pipeline gotcha — match_one ≠ the gate; `sig_unify` is MANDATORY  <sub>L1421</sub>
- **§3-The** — gate two-stage — sig_unify is a FALLBACK, not unconditional (`gate_stage.py`, §19 folded in)  <sub>L2189</sub>
- **§28** — Banking a "close=0 gate-rejected" giant: the canonical-extern recovery (Phase 22 T1, byte-proven on func_8015126C ×134)  <sub>L2325</sub>
- **§28d** — The "macro-extern-injection" lever: freeing reach-134 inline matches dedup_propagate skips as "not self-contained" (Phase 23, `tools/inject_capped_externs.py`)  <sub>L2366</sub>
- **§33** — Automating the giant decl-reconcile: `tools/reconcile_decls.py` (the DATA analog of §20's `cast_call_sites`) + a fleet-majority type oracle (Phase 24 T7b, byte-proven on `func_80129CF8`)  <sub>L2434</sub>
- **§40a** — The DECL-RECONCILE pass (type-lift so remapped drafts compile in the sibling TU) (Phase 25 T7.2, 2026-07-08)  <sub>L2601</sub>
- **§41** — The DEF-SIDE canonical-sig wall: mechanically banking a drafted giant past `conflicting types` (Phase 25 T5b batch-2, 2026-07-09; `tools/canon_sig_reconcile.py`, byte-proven on `func_8013B274`)  <sub>L2710</sub>
- **§41b** — addendum — M4 "reconcile_decls" tier is ALSO a probe over-count: 0/8 mechanical (Phase 25 T7-M4, 2026-07-10)  <sub>L2858</sub>
- **§41d** — `void`→`s32` is NOT always byte-neutral: gate the RAW draft FIRST (Phase 26 session 6, byte-proven)  <sub>L2878</sub>
- **§43** — The K&R s16-param definition DISSOLVES the "narrow-param wall" for by-value register args (Phase 25 task A, Fable5 crack of the 369-ins giant `func_80166994` ×134, 2026-07-11)  <sub>L3195</sub>
- **§51g** — When the thing you are scanning has a GRAMMAR, parse the grammar (`tools/cdecl.py`)  <sub>L3769</sub>
- **§54** — `--fix-def-sig`: the member's CANONICAL DECLARATION is a build step too (tiny-IMM mega-pools, +4,801, Phase 29 T6, 2026-07-16)  <sub>L4163</sub>
- **§56** — Banking a hand-drafted GIANT into its exemplar TU: self-contained draft vs live-TU decls, 4 reconciliation tactics (`func_8013FAF8`, 312 ins ×138, Phase 29 T4, 2026-07-17)  <sub>L4187</sub>
- **§56b** — PROPAGATING an h_seq giant: the exemplar's externs MUST be fleet-canonical, not the draft's types  <sub>L4232</sub>
- **§57** — The SELF-decl normalize: the sibling's OWN caller declares the templated fn divergently (the third §17a-1 direction; `tools/normalize_self_decls.py`, func_801670E4 ×137, Phase 29, 2026-07-18)  <sub>L4256</sub>
- **§61** — Task 14: the gate ladder's missing stage is the ARITY pre-pass, and it is TU-side not draft-side (Phase 29, 2026-07-21)  <sub>L4545</sub>
- **§63** — The fresh-138 DEF-SIDE blocker: fix the HEADER decl, not the draft (`fix_header_decl.py`, Phase 29 SESSION-13, 2026-07-23)  <sub>L4877</sub>
- **§63** — UPDATE (Phase 29 SESSION-14) — `fix_header_decl`'s "SAFE" verdict is FLEET-BLIND; it MUST be R22-validated  <sub>L5000</sub>
- **§65c** — `rtu_match` MATCH → bank held 13/13 on self-decl, but broke on the FIRST callee-decl case  <sub>L5081</sub>
- **§65g** — Where the cheap levers STOP: the local-type and in-TU-self-decl classes did not yield  <sub>L5139</sub>
- **§66a** — The widest write in a pipeline is the one most likely to be UNDECLARED  <sub>L5181</sub>
- **§68** — A comment-only line halted the extern scan, and the skip label blamed the type cap (Phase 29 SESSION-18)  <sub>L5524</sub>
- **§73** — A def-side self-decl conflict has TWO axes: RETURN (fleet widen, T2) and PARAMS (casts at each use, T0). Diagnose which before paying for the expensive one (Phase 29 SESSION-19, `func_8014F3E8` + `func_8014D4C0`)  <sub>L5783</sub>
- **§75** — A propagation cap is usually a MINORITY-SPELLING SOURCE OVERLAY, not a wall: census the carried extern before believing the exclusion message (Phase 29 SESSION-19, `func_8014F3E8` ×4 → ×138)  <sub>L5866</sub>
- **§75b** — A body's preamble can carry `#define`s, not just `extern`s; extraction lifts only the externs (Phase 29 SESSION-19, `func_80165CA0` ×3 → fleet)  <sub>L5948</sub>
- **§75c** — Class-B's remedy is the FULL §17a-1 PAIR (decl **and** call-site cast); a decl-only fix moves the error and looks like a new wall (Phase 29 SESSION-19, `func_8012F14C`)  <sub>L5997</sub>
- **§76** — The allocno CLASS (local vs global) is the dominant regalloc lever, and C reaches it ONLY through declaration scope and variable reuse (Phase 29 SESSION-19, behemoth #3 `func_8017F510` 1,511 ins, 97 → MATCH, pin-free)  <sub>L6028</sub>
- **§79** — For a 0-callee giant, fingerprint by DATA symbols (§71 cannot fire); and the STACK-SLOT ORDER is a declaration-order oracle (Phase 29 SESSION-19, `func_8017BF14` 4,763 ins, cold start → 45/4763)  <sub>L6220</sub>
- **§3-NEW** — LEVER — the frame layout reads back the original declaration order  <sub>L6240</sub>
- **§3-1.** — The inlined-helper signature  <sub>L6375</sub>
- **§92** — Conforming a DECLARATION: pointer changes are caller-neutral, scalar-WIDTH changes are not (Phase 29 SESSION-21, `tools/conform_decls.py`)  <sub>L6891</sub>
- **§95** — `reconcile_tu` dropped the SIBLING declarators of a multi-symbol `extern` line (Phase 29 SESSION-21, `func_80176218`)  <sub>L6983</sub>
- **§96** — The same rewrite, one shape down: `reconcile_tu` matched statements to lines by TEXT, so every COMMENTED declaration was silently skipped (Phase 29 SESSION-22, `func_80176218` banked)  <sub>L7018</sub>
- **§98** — `conform_decls` had three defects, and only the third needed R22 to find (Phase 29 SESSION-22, `func_8014CF04`)  <sub>L7113</sub>
- **§99** — The narrow-param wall is a DEF-side problem with a ZERO-blast-radius fix: convert the definition to K&R (Phase 29 SESSION-22, `func_80175AB8` + `func_80175DA8`)  <sub>L7167</sub>
- **§102** — A PLUMBING verdict can MASK a DIFF; and K&R is not always a codegen change (Phase 29 SESSION-22, `func_8016EC0C`)  <sub>L7272</sub>
- **§103** — A FILE-scope `extern` in a shared overlay TU is a GLOBAL constraint on every LATER function; move the DECL, not the draft (Phase 29 T48/T51, `func_80135260` — the fleet-wide half)  <sub>L7307</sub>
- **§104** — Two silent-skip defects in one scan: match on MASKED text, emit from the ORIGINAL (Phase 29 T53, `gather_externs`)  <sub>L7395</sub>
- **§112** — A macro-scoped declaration only collides where the macro is INSTANTIATED (Phase 29 T67/T69, `audit_header_sigs.py`)  <sub>L7730</sub>
- **§113** — An ARITY blocker only exists if the macro CALLS the function; an address-taken use has no call site (Phase 29 T72)  <sub>L7776</sub>
- **§114** — The THIRD decl axis: a CALLEE the draft declares differently from the target TU (Phase 29 T76/T77)  <sub>L7808</sub>
- **§3-Do** — NOT "strip the duplicate typedef" — it breaks the extern that uses it  <sub>L8037</sub>
- **§121** — Synthesise externs for macro-DEFINED callees from the macro's own definition head (Phase 29 T95)  <sub>L8061</sub>
- **§3-The** — declaration surface (integration, not codegen)  <sub>L8938</sub>
- **§136f** — Two declaration sub-cases the reconcile lane surfaced (lane now 15/15 lifetime)  <sub>L9183</sub>
- **§138** — The propagation lanes: a gate refusal is a DECLARATION, and which lever you owe depends on blast radius  <sub>L9412</sub>
- **Reconciling** — a gate-refused draft: which way you edit depends on WHERE the TU's decl is  <sub>L9562</sub>
- **§141** — The §134 class is CLOSED: every line-shape decision now routes through `cdecl._mask` (P30 S39)  <sub>L9737</sub>
- **§143** — `cast_call_sites` read a RETURN STATEMENT as a prototype and deleted it. A 0/39 sweep became 18/39. (P30 S40)  <sub>L9831</sub>
- **§3-C.** — Inner-block declaration does NOT delay slot allocation — BYTE-REFUTED  <sub>L10065</sub>
- **§159** — THE DECLARATION AXIS: conform to byte-truth, and make every guard state its COVERAGE (P30 S47; ~10,930 sites across 8 axes, fleet byte-identical)  <sub>L10825</sub>
- **§168** — THE COUSIN TIER (P30 S49, 2026-08-12): h_seq's exact-hash brittleness, measured — and the similarity map above it  <sub>L16208</sub>

### jump tables & switches  (27)

- **§8** — rodata island (compiler jump tables) — the `.data→.rodata→.data` sandwich (Phase 7)  <sub>L320</sub>
- **§8a** — rodata island in a flat OVERLAY — the tail sandwich, per matched jr-function (Phase 26 — PoC PROVEN)  <sub>L342</sub>
- **§8a-pad** — a trailing `.word 0x00000000` under a jtbl dlabel is `.align` PAD, not an entry (Phase 26 session 6, byte-proven)  <sub>L380</sub>
- **§8b** — MULTI-jtbl per overlay — the `ld_interleave --order` sandwich + the same-subseg cases (Phase 26 session 4)  <sub>L404</sub>
- **§8e** — The jtbl ALIGNMENT LAW + the pad-spec filter — multi-table .rodata spans (Phase 29, byte-proven; `.run/probe_jtbl/verdict.md`)  <sub>L530</sub>
- **§3-The** — clean object-level metric (use this, not the permuter score, for jtbl/rodata functions)  <sub>L844</sub>
- **§41b** — T7 execution: the object-only probe OVER-counts BANKABLE by two link/rodata classes (Phase 25 T7-M1, 2026-07-10)  <sub>L2809</sub>
- **§48** — The 12-core jr crack wave: the ALLOCNO-PRICING dials and the EBB rule (Phase 26 session 8, Ultracode, 9/12 MATCH first pass)  <sub>L3405</sub>
- **§53** — SWEEP A FAMILY WITH THE TOOL ITS EXEMPLAR NEEDED: the jr/switch carve, and how omitting it manufactured the "families don't template" doctrine (Phase 28 T1, 2026-07-15)  <sub>L4028</sub>
- **§61b** — The jtbl gate stage: built, and the ORDERING law it exposed (Phase 29 Task-14 stage 4, 2026-07-21)  <sub>L4658</sub>
- **§61c** — The jtbl bank is INCREMENTALLY valid and CLEAN-INVALID (Phase 29, 2026-07-21) — the blocking finding  <sub>L4713</sub>
- **§62** — The jtbl RECONCILE must also follow the carve: the post-carve draft reconcile (`harvest_verify._jtbl_reconcile`, Phase 29 SESSION-11, 2026-07-22)  <sub>L4833</sub>
- **§81** — Banking a jr (jump-table) function: the 3-step carve chain, and why `match_one` cannot see the problem (Phase 29 SESSION-19, `func_8017C954`)  <sub>L6323</sub>
- **§88d** — BANKING ORDER: run the §81 carve chain BEFORE banking, never after  <sub>L6701</sub>
- **§97** — The gate's own tree hygiene: a refused carve, an unchecked recovery, and a snapshot that captured a dirty tree (Phase 29 SESSION-22)  <sub>L7066</sub>
- **§105** — A gate's revert must survive an EXCEPTION, not just a failure (Phase 29 T53, `jtbl_family_bank`)  <sub>L7431</sub>
- **§125** — Split the CARVE from the BODY before calling a jr residue a wall — and measure it by SHA from a CLEAN tree (P30 SESSION-28; **this section's first draft was WRONG and the method caught it**)  <sub>L8206</sub>
- **§126** — The carve-within-a-carve: an ADDRESS RANGE is not an OPTIMIZATION REGION (P30 T2, byte-proven end-to-end)  <sub>L8264</sub>
- **§129** — Post-carve, `rtu_match`/`match_one` COUNT THE JUMP TABLE AS INSTRUCTIONS; and a carve must never be committed without its owner (P30 S28, `func_8013BD74`)  <sub>L8437</sub>
- **§129a** — the target instruction count is INFLATED after a carve  <sub>L8441</sub>
- **§129b** — never commit a carve whose owner is still a stub (it strands the carve)  <sub>L8462</sub>
- **§130** — An INCREMENTAL build can report BYTE-IDENTICAL for a change the CLEAN build cannot even LINK (P30 S28, the jr pair)  <sub>L8486</sub>
- **§131** — The jtbl OVER-SPAN: `sltiu N` is ground truth in BOTH directions, and the zero-word rule only guards one (P30 S28, `func_80191C50`)  <sub>L8533</sub>
- **§132** — The `JR-PAIR-IN-ONE-O0-OBJECT` "wall" was TWO instrument defects: a merged-double span the carve could not see, and a truncated object no rule deleted (P30 S29, `func_8013B83C` + `func_8013BD74`)  <sub>L8570</sub>
- **§132b** — When the span's already-matched owner is ITSELF multi-switch: `--span-rel` (P30 S1, `func_8014032C`)  <sub>L8660</sub>
- **§139** — A GATE THAT GREPS FOR VERDICTS MUST ASSERT 1:1 ACCOUNTING; and a `--src` filter must not survive a carve (P30 S38, wave 6: 10 of 16 drafts vanished)  <sub>L9591</sub>
- **§161** — THE RETRY-WAVE HARVEST: a jump table indexed from zero, and two allocator traps (P30 S47)  <sub>L10966</sub>

### optimisation level (-O0/-O2)  (12)

- **§6** — Per-module optimization mixing — the -O0 boot module (Phase 7)  <sub>L246</sub>
- **Detecting** — the opt level (do this first)  <sub>L254</sub>
- **Build** — mechanism — per-file opt override (splat resegmentation)  <sub>L288</sub>
- **§18** — Per-file `-O0` split inside an overlay/blob (Phase 19 T1)  <sub>L1519</sub>
- **§38** — The WHALE `func_80144B9C` (770 ins): the -O0 struct-assign memcpy idiom + the -O0 reach-134 ×134 rollout (Phase 24 T7 §G, cheap Opus — no Fable5, no calls.c)  <sub>L2527</sub>
- **§39** — The ×1→×134 giant-endgame: propagate a matched **-O2** giant via the NATIVE DEFINE-macro path (Phase 24 T7 §G close, 2026-07-08)  <sub>L2537</sub>
- **§116** — Optimization level is a property of the FILE, not the function: read a family 0/N against the member's stub HOME (Phase 29 T79)  <sub>L7877</sub>
- **§3-The** — finding: opt level is per FILE, so the file's contents must be opt-HOMOGENEOUS  <sub>L8279</sub>
- **§127** — The `-O0` regime: the CONSTANT-OFFSET FOLD, and why `-O0` needs its own idiom set (P30 T3 wave, 15 targets)  <sub>L8349</sub>
- **§3-The** — rest of the `-O0` regime (write PLAIN C, and mean it)  <sub>L8367</sub>
- **§127a** — §71 (sibling-first) is the strongest `-O0` lever, and it beats the index  <sub>L8377</sub>
- **§132** — The `JR-PAIR-IN-ONE-O0-OBJECT` "wall" was TWO instrument defects: a merged-double span the carve could not see, and a truncated object no rule deleted (P30 S29, `func_8013B83C` + `func_8013BD74`)  <sub>L8570</sub>

### family propagation & sweeps  (84)

- **§8d** — Templating a body INTO a TU must not CHANGE its declaration environment — demote the carried data externs (Phase 26 session 8, byte-proven on `func_8015AE2C` ×133)  <sub>L483</sub>
- **§11** — Cross-binary dedup & code-sharing (Phase 11 — "one match unlocks many")  <sub>L908</sub>
- **§3-The** — mechanism: game-code dedup is SOURCE-LEVEL, not an object swap (R-D1, the key lesson)  <sub>L926</sub>
- **§3-The** — flat-blob overlay config (what the template encodes)  <sub>L1071</sub>
- **§3-THE** — NON-4-ALIGNED-OVERLAY GOTCHA (≈75% of the fleet; fixed in the template + Makefile, automatic)  <sub>L1079</sub>
- **§14** — Propagate a matched function across the fleet — `tools/dedup_propagate.py` (Phase 15)  <sub>L1126</sub>
- **§19** — Scaling the toolkit waves — the recovery PIPELINE + the propagation CAP (Phase 19 T3)  <sub>L1597</sub>
- **§24** — The `ov_SC01_077_a.c` split-file vein: split-aware propagation, but loose-typing-gated (Phase 21)  <sub>L2085</sub>
- **§25** — The "schedule" class is mostly COALESCING (pin-crackable), not scheduling; + the gate two-stage + h_exact over-counts ×134 (Phase 21, cont.6)  <sub>L2132</sub>
- **§3-h** — _exact OVER-COUNTS ×134 — verify shareability before crediting a class's "reach-134" (R14)  <sub>L2177</sub>
- **Where** — this leaves the reach-134 tail (cont.6 option-3, now CONFIRMED byte-backed)  <sub>L2233</sub>
- **§28** — Banking a "close=0 gate-rejected" giant: the canonical-extern recovery (Phase 22 T1, byte-proven on func_8015126C ×134)  <sub>L2325</sub>
- **§28c** — The close=0 recovery is NOT fully exhausted (§26 corrected, R14); + the dedup_propagate registry-skip recovery (Phase 22 T2)  <sub>L2352</sub>
- **§28b** — The struct-walled close=0 giant (§28 case #3) IS bankable: the engine_types.h type-lift (Phase 22 T2, byte-proven on func_80156B74 ×134)  <sub>L2357</sub>
- **§28d** — The "macro-extern-injection" lever: freeing reach-134 inline matches dedup_propagate skips as "not self-contained" (Phase 23, `tools/inject_capped_externs.py`)  <sub>L2366</sub>
- **§30** — Fable5Max cracks §20/§10 "unsteerable" from the gcc SOURCE: store-vs-load is a `/s` aliasing flag, the def-side wall has a macro escape, + the birthing-boost (Phase 23, byte-proven on giant `func_8014EE14` 248 ins ×134)  <sub>L2382</sub>
- **§34** — The `func_80138ED0` giant crack: gcc-2.7.2's **3-qty sort bug** + the **zero-byte asm allocation toolkit** + the **giv-init fence** (Phase 24 T5; Opus→close=21, Fable5→MATCH ×134)  <sub>L2454</sub>
- **§38** — The WHALE `func_80144B9C` (770 ins): the -O0 struct-assign memcpy idiom + the -O0 reach-134 ×134 rollout (Phase 24 T7 §G, cheap Opus — no Fable5, no calls.c)  <sub>L2527</sub>
- **§39** — The ×1→×134 giant-endgame: propagate a matched **-O2** giant via the NATIVE DEFINE-macro path (Phase 24 T7 §G close, 2026-07-08)  <sub>L2537</sub>
- **§40** — Structural families: the MECHANICAL symbol-remap (crack one exemplar → remap the rest, ~0 tokens) (Phase 25 T3/T7, 2026-07-08)  <sub>L2553</sub>
- **§40a** — The DECL-RECONCILE pass (type-lift so remapped drafts compile in the sibling TU) (Phase 25 T7.2, 2026-07-08)  <sub>L2601</sub>
- **§40b** — The reloc-tracker blind spot: the indexed-global idiom that hid the "reach-1 tail" (Phase 26 Task 1, 2026-07-11, byte-verified V0/V1)  <sub>L2634</sub>
- **§40c** — The h_seq per-sibling reconcile: templating a reconcile-class crack ×134 (Phase 26 Task 8, 2026-07-12, byte-proven)  <sub>L2676</sub>
- **§41c** — T7-M2: the ×134 def-side-wall sweep via per-sibling RE-reconcile (Phase 25, 2026-07-10; 4,389 banks, ~0 agent tokens)  <sub>L2839</sub>
- **§42** — The F-band ≤28 regalloc crack wave: register-pin/DENSITY levers beat the permuter (Phase 25 T7 F-band, 2026-07-10; Ultracode 9-worker wave, 4/9 banked byte-identical, 266 swept ×134)  <sub>L2900</sub>
- **§42e** — propagating a CRACK ×134: the def-finder bug + the byte-drift residual (the "remap-fail" misdiagnosis)  <sub>L3121</sub>
- **§43** — The K&R s16-param definition DISSOLVES the "narrow-param wall" for by-value register args (Phase 25 task A, Fable5 crack of the 369-ins giant `func_80166994` ×134, 2026-07-11)  <sub>L3195</sub>
- **§45** — The flagship `func_80133CD4` crack (399 ins ×134): the merged-variable permutation-breaker + the 1-death local-alloc gate (Phase 25 task A giant escalation, Fable5 gdb-on-cc1, 2026-07-11)  <sub>L3298</sub>
- **§46** — The `func_80178D40` crack (890 ins ×134, the heaviest core in the game): four LOOP-STRUCTURE levers cheap-Opus found by reading loop.c/jump.c/cse.c (Phase 26 session 8, 2026-07-13)  <sub>L3313</sub>
- **§47** — The live-length SLIDER: splitting a global.c allocno-priority TIE with one zero-byte asm (Phase 26 session 8, Fable5 Max, byte-proven on `func_8017BEBC` 952 ins ×113)  <sub>L3365</sub>
- **§49** — The LUID DIAL: a zero-byte SCHEDULING dial (the sched.c analogue of §47) — `func_8017A4AC` (536 ins ×134), Phase 26 session 8  <sub>L3527</sub>
- **§52** — The WALKER-FAMILY skeleton: 6 regalloc-order levers + a deeply-characterized intrinsic wall (`func_80178004`, 165 ins ×134; Phase 26, Fable5, 2026-07-15)  <sub>L3913</sub>
- **§3-The** — 6 levers (the "walker-family skeleton" — apply to the 12 siblings; levers 1-5 retire ~half the residual)  <sub>L3924</sub>
- **§53** — SWEEP A FAMILY WITH THE TOOL ITS EXEMPLAR NEEDED: the jr/switch carve, and how omitting it manufactured the "families don't template" doctrine (Phase 28 T1, 2026-07-15)  <sub>L4028</sub>
- **§55c** — Sizing the propagate: a TARGETED propagate is ~4 min/core, and "it's slow" was a BROKEN-TREE ARTIFACT  <sub>L4141</sub>
- **§54** — `--fix-def-sig`: the member's CANONICAL DECLARATION is a build step too (tiny-IMM mega-pools, +4,801, Phase 29 T6, 2026-07-16)  <sub>L4163</sub>
- **§56** — Banking a hand-drafted GIANT into its exemplar TU: self-contained draft vs live-TU decls, 4 reconciliation tactics (`func_8013FAF8`, 312 ins ×138, Phase 29 T4, 2026-07-17)  <sub>L4187</sub>
- **§56b** — PROPAGATING an h_seq giant: the exemplar's externs MUST be fleet-canonical, not the draft's types  <sub>L4232</sub>
- **§57** — The SELF-decl normalize: the sibling's OWN caller declares the templated fn divergently (the third §17a-1 direction; `tools/normalize_self_decls.py`, func_801670E4 ×137, Phase 29, 2026-07-18)  <sub>L4256</sub>
- **§57a** — Two NSD corrections + the SURGICAL-ONLY law + the honest broad-sweep yield (Phase 29, 2026-07-18)  <sub>L4303</sub>
- **§59** — Three h_seq sweep-residual classes match_one/the-exemplar-bank don't reveal (Phase 29 crack-wave close, 2026-07-18)  <sub>L4377</sub>
- **§3-Two** — corollaries worth remembering  <sub>L4450</sub>
- **§65a** — The blast-radius taxonomy (makes §61's law structural instead of remembered)  <sub>L5030</sub>
- **§3-Giv** — record order (the §70 family)  <sub>L5760</sub>
- **§75** — A propagation cap is usually a MINORITY-SPELLING SOURCE OVERLAY, not a wall: census the carried extern before believing the exclusion message (Phase 29 SESSION-19, `func_8014F3E8` ×4 → ×138)  <sub>L5866</sub>
- **§75a** — The exclusion classes, enumerated with named causes (Phase 29 SESSION-19, the 134-binary `dedup_extend` sweep)  <sub>L5925</sub>
- **§75b** — A body's preamble can carry `#define`s, not just `extern`s; extraction lifts only the externs (Phase 29 SESSION-19, `func_80165CA0` ×3 → fleet)  <sub>L5948</sub>
- **§76** — The allocno CLASS (local vs global) is the dominant regalloc lever, and C reaches it ONLY through declaration scope and variable reuse (Phase 29 SESSION-19, behemoth #3 `func_8017F510` 1,511 ins, 97 → MATCH, pin-free)  <sub>L6028</sub>
- **§77** — Every extraction tool carries a NARROW hard-coded preamble set; anything outside it silently caps the body's reach. Diff the preamble before you gate. (Phase 29 SESSION-19 — three variants in one session, two different tools)  <sub>L6084</sub>
- **§84** — The DERIVED-OFFSET remap bug: a hand-computed literal that encodes the DISTANCE between two per-overlay symbols, and why `match_one` is structurally blind to it (Phase 29 SESSION-20, `func_8013D53C`)  <sub>L6472</sub>
- **§86** — Pinned-exemplar templatability is a PER-FAMILY property, not a per-member rate; and the §42e pin guard is now over-conservative (Phase 29 SESSION-20)  <sub>L6589</sub>
- **§3-THE** — LAW: all-or-nothing PER FAMILY  <sub>L6600</sub>
- **§3-Why** — the two live families differ from the three dead ones — the open question  <sub>L6627</sub>
- **§3-The** — cheap discriminator, before spending a sweep  <sub>L6666</sub>
- **§89b** — the parallel gate farm existed; the family path could not reach it (`tools/sweep_parallel.py`)  <sub>L6745</sub>
- **§91** — A structure-TRANSFER is only valid where the structure corresponds: the `--like` role trap (Phase 29 SESSION-21, `func_8012AAAC` ×137)  <sub>L6842</sub>
- **§94** — A family sweep's `0/N` is a TYPE-CARRY failure until proven otherwise: lift the exemplar's local types, ALL of them, transitively (Phase 29 SESSION-21, `func_8016B6BC` 0/137 → 137/137)  <sub>L6947</sub>
- **§105** — A gate's revert must survive an EXCEPTION, not just a failure (Phase 29 T53, `jtbl_family_bank`)  <sub>L7431</sub>
- **§107** — A lever wired into ONE gate path is a lever most families cannot reach (Phase 29 T56, `func_80144090` 0/136 → 136/136)  <sub>L7513</sub>
- **§108** — Diagnosing a family `0/N`: the four causes, and the third opt-in lever (Phase 29 T59)  <sub>L7551</sub>
- **§111** — The distinct-code metric is not noisy: a family pays it only if its members are byte-VARIANT (Phase 29 T66)  <sub>L7681</sub>
- **§116** — Optimization level is a property of the FILE, not the function: read a family 0/N against the member's stub HOME (Phase 29 T79)  <sub>L7877</sub>
- **§123** — PROPAGATE A FAMILY WITH THE TOOL ITS TIER NEEDS: `dedup_propagate` is h_exact-only; its refusals are statements about the TOOL (P30 wave 1, 2026-07-30)  <sub>L8116</sub>
- **§124** — A "not matched" verdict can mean the definition is there under a DIFFERENT C NAME: the asm-label alias blind spot (P30 SESSION-28, `func_8016191C` ×137)  <sub>L8153</sub>
- **§124a** — a family sweep's `0 matched-exemplar families` may be a FILTER, not a wall  <sub>L8198</sub>
- **§134** — MULTI-LINE BLINDNESS: one root cause, four faces, in `family_remap`'s preamble scanner (P30 S6, 190 zero-crack families)  <sub>L8711</sub>
- **§136c** — SIBLING-FIRST is a DERIVATION shortcut, not just a conflict fix (the fastest route in a family wave)  <sub>L9066</sub>
- **§138** — The propagation lanes: a gate refusal is a DECLARATION, and which lever you owe depends on blast radius  <sub>L9412</sub>
- **§142** — An open stub whose `h_exact` class is MATCHED elsewhere is FREE. Propagate the body; do not gate a draft. (P30 S39, +7,710 ins in two commands)  <sub>L9775</sub>
- **§143** — `cast_call_sites` read a RETURN STATEMENT as a prototype and deleted it. A 0/39 sweep became 18/39. (P30 S40)  <sub>L9831</sub>
- **Then** — propagation returned 0/137 TWICE — both times a missing TYPE  <sub>L9997</sub>
- **§3-A.** — The frame has THREE strata, and stratum 3 is unreachable from C  <sub>L10038</sub>
- **§3-E.** — A `qty_compare` TIE is not spelling-reachable — recognise it and stop  <sub>L10080</sub>
- **Consequence** — for the family (a real scheduling decision)  <sub>L10092</sub>
- **§150** — A register ROTATION across symmetric blocks is VARIABLE-IDENTITY evidence, not an allocator tie (P30 S43, `func_8017C6F4`, 947 ins ×4)  <sub>L10309</sub>
- **§151** — THE GHOST WEDGE: when a load-before-store transposition is unreachable by ANY statement order (P30 S43, `func_8017EF68`, 969 ins)  <sub>L10364</sub>
- **When** — to reach for it  <sub>L10404</sub>
- **§152** — BYTE SIZE is the family key that name- and h_seq-grouping both miss (P30 S43, the 0xECC family: 1 crack → 12 overlays → 11,364 ins)  <sub>L10415</sub>
- **§156** — an ORPHANED reconcile poisons the fleet: `dedup_propagate`'s kept edit (S45 p6/p7)  <sub>L10637</sub>
- **§160** — THE REACH-15 WAVE HARVEST: an align-1 block move, and the instrument that called it a wall (P30 S47)  <sub>L10890</sub>
- **§162** — S48 WAVE-1 HARVEST (P30, 2026-08-11): the reach-ordered sibling campaign's first 12 targets  <sub>L11011</sub>
- **§163z** — THE UNVETTED REMAINDER (do not cite as law; each needs a dedupe pass)  <sub>L11825</sub>
- **§3-The** — ADDRESS-CLASS TABLE: which load/store pairs even REACH the `/s` clause (P30 S48 wave 4, `func_80185B44`, ov_SC03_014)  <sub>L14339</sub>
- **§168** — THE COUSIN TIER (P30 S49, 2026-08-12): h_seq's exact-hash brittleness, measured — and the similarity map above it  <sub>L16208</sub>

### integration / TU plumbing  (39)

- **§8c** — Splitting a TU means rebuilding its DECLARATION ENVIRONMENT, not moving text (Phase 26 session 6)  <sub>L437</sub>
- **§8d** — Templating a body INTO a TU must not CHANGE its declaration environment — demote the carried data externs (Phase 26 session 8, byte-proven on `func_8015AE2C` ×133)  <sub>L483</sub>
- **§9.3** — Make it the build: resegment + swap + resolve (`tools/psyq_integrate.py`, libcd DONE)  <sub>L685</sub>
- **§9.4** — Integrating a SECOND library (libgs block 6 after libcd) — multi-library gotchas  <sub>L712</sub>
- **§9.5** — Integrating a WHOLE multi-block library in one call (full libgs — Phase 7 session G)  <sub>L740</sub>
- **§17a** — The TOOLKIT at WAVE scale (Phase-18 Step-3b/Step-1 — measured) + the pipeline-integration gotchas  <sub>L1445</sub>
- **§3-THE** — CAP: at this tail the match_one→gate gap is the LOOSE-TYPING CALL-GRAPH wall (not cheap plumbing)  <sub>L1637</sub>
- **Diagnostic** — lesson: a failed in-TU build leaves a STALE `.o`  <sub>L1654</sub>
- **§28d** — The "macro-extern-injection" lever: freeing reach-134 inline matches dedup_propagate skips as "not self-contained" (Phase 23, `tools/inject_capped_externs.py`)  <sub>L2366</sub>
- **§30a** — §30 generalizes to the FULL near-miss backlog (via STANDARD Opus agents, not just Fable5) + 2 more steer levers + the mechanical-integration throughput unlock (Phase 23 (a))  <sub>L2394</sub>
- **§40a** — The DECL-RECONCILE pass (type-lift so remapped drafts compile in the sibling TU) (Phase 25 T7.2, 2026-07-08)  <sub>L2601</sub>
- **§42a** — addendum — wave 2 (residuals + 29-100 band): iso-MATCH ≠ real-TU bank, the memcpy→struct-assign fix, +5 levers (2026-07-10b)  <sub>L2953</sub>
- **§42c** — addendum — wave 3 (Max orchestrator + CORRECTED Ultracode fan-out, 2026-07-10c): the real-TU-faithful parallel harness (rtu_match) + 7/9 crack, ZERO iso-drift  <sub>L3033</sub>
- **§56** — Banking a hand-drafted GIANT into its exemplar TU: self-contained draft vs live-TU decls, 4 reconciliation tactics (`func_8013FAF8`, 312 ins ×138, Phase 29 T4, 2026-07-17)  <sub>L4187</sub>
- **§61** — Task 14: the gate ladder's missing stage is the ARITY pre-pass, and it is TU-side not draft-side (Phase 29, 2026-07-21)  <sub>L4545</sub>
- **§61a** — The Task-5 wave: 11/12 MATCH, 0 banked — three DISTINCT integration walls, each now named (Phase 29, 2026-07-21)  <sub>L4606</sub>
- **§63** — The fresh-138 DEF-SIDE blocker: fix the HEADER decl, not the draft (`fix_header_decl.py`, Phase 29 SESSION-13, 2026-07-23)  <sub>L4877</sub>
- **§63** — UPDATE (Phase 29 SESSION-14) — `fix_header_decl`'s "SAFE" verdict is FLEET-BLIND; it MUST be R22-validated  <sub>L5000</sub>
- **§65** — The stranded-draft recovery: BLAST-RADIUS TIERS, and the per-overlay de-macroize that refutes §20's DEF-conflict wall (`tools/demacroize.py` + `tools/blocker_probe.py`, Phase 29 SESSION-16, 2026-07-24)  <sub>L5015</sub>
- **§65b** — The escape: de-macroize the instantiation, don't touch the shared header  <sub>L5045</sub>
- **§65f** — The de-macroize lever's BOUNDARY, measured: byte-neutral 14 times of 15, and the 15th is why the gate exists  <sub>L5118</sub>
- **§65g** — Where the cheap levers STOP: the local-type and in-TU-self-decl classes did not yield  <sub>L5139</sub>
- **§85** — The RETURN-axis fleet widen is ALL-OR-NOTHING: widening the shared header alone guarantees a conflict in the source overlay (Phase 29 SESSION-20, `func_8012CC88` / `func_8014D12C`)  <sub>L6539</sub>
- **§95** — `reconcile_tu` dropped the SIBLING declarators of a multi-symbol `extern` line (Phase 29 SESSION-21, `func_80176218`)  <sub>L6983</sub>
- **§96** — The same rewrite, one shape down: `reconcile_tu` matched statements to lines by TEXT, so every COMMENTED declaration was silently skipped (Phase 29 SESSION-22, `func_80176218` banked)  <sub>L7018</sub>
- **§3-Two** — `reconcile_tu` bugs found underneath, one introduced while fixing the other  <sub>L7190</sub>
- **§100** — Prefer the DRAFT-LOCAL fix: a type only one function uses belongs in its BODY, not in a shared header (Phase 29 SESSION-22, `func_80175DA8` 0/137 → 137/137)  <sub>L7216</sub>
- **§102** — A PLUMBING verdict can MASK a DIFF; and K&R is not always a codegen change (Phase 29 SESSION-22, `func_8016EC0C`)  <sub>L7272</sub>
- **§103** — A FILE-scope `extern` in a shared overlay TU is a GLOBAL constraint on every LATER function; move the DECL, not the draft (Phase 29 T48/T51, `func_80135260` — the fleet-wide half)  <sub>L7307</sub>
- **§109** — Conforming a definition to a shared header: fix the NAMES, then check the RETURN precondition (Phase 29 T60)  <sub>L7600</sub>
- **§112** — A macro-scoped declaration only collides where the macro is INSTANTIATED (Phase 29 T67/T69, `audit_header_sigs.py`)  <sub>L7730</sub>
- **§114** — The THIRD decl axis: a CALLEE the draft declares differently from the target TU (Phase 29 T76/T77)  <sub>L7808</sub>
- **§3-The** — integration idioms (these decide whether a byte-correct draft BANKS)  <sub>L8804</sub>
- **§3-The** — declaration surface (integration, not codegen)  <sub>L8938</sub>
- **Reconciling** — a gate-refused draft: which way you edit depends on WHERE the TU's decl is  <sub>L9562</sub>
- **§3-The** — same swallow, twice more, in the integration spine  <sub>L9706</sub>
- **§3-A.** — Payload word0 is a global MODULE ID; code starts after the header  <sub>L10524</sub>
- **§162** — THE LICM PAIR: what makes an address a movable AT ALL, and why the preheader order is the body order (P30 S47, `ov_MAIN_012`)  <sub>L11140</sub>
- **§166** — THE DESTINATION-TU ORACLE (P30 S48): the seven-attempt bug that was never codegen  <sub>L14899</sub>

### build graph, splat & the harness  (109)

- **§4** — Flag/toolchain gotchas  <sub>L190</sub>
- **Build** — mechanism — per-file opt override (splat resegmentation)  <sub>L288</sub>
- **§8c** — Splitting a TU means rebuilding its DECLARATION ENVIRONMENT, not moving text (Phase 26 session 6)  <sub>L437</sub>
- **§9** — Link real PsyQ library objects byte-exact (Phase 7 — GO proven)  <sub>L613</sub>
- **§9.1** — Generalised per-object linker — `tools/psyq_link.py` (+ `psyq_link_lib.py`), 18/18 libcd byte-exact  <sub>L642</sub>
- **§9.2** — Wire a library region into the build with NOLOAD — no data carving (`tools/psyq_link_region.py`)  <sub>L666</sub>
- **§9.3** — Make it the build: resegment + swap + resolve (`tools/psyq_integrate.py`, libcd DONE)  <sub>L685</sub>
- **§9.6** — Scaling library linking to the whole EXE (Phase 8 — 8 libs linked, 20%→50% byte-identical)  <sub>L766</sub>
- **§9.7** — Binary-agnostic toolchain refactor (Phase 9) — the reusable pattern for Gen2  <sub>L802</sub>
- **§3-The** — clean object-level metric (use this, not the permuter score, for jtbl/rodata functions)  <sub>L844</sub>
- **§3-The** — mechanism: game-code dedup is SOURCE-LEVEL, not an object swap (R-D1, the key lesson)  <sub>L926</sub>
- **Per-binary** — toolchain provenance (R24)  <sub>L959</sub>
- **§12** — Ultracode harvest — parallel-draft + byte-gate at scale (Phase 12, resident: 1.4%→71.7% in one session)  <sub>L964</sub>
- **One-command** — onboarding — `tools/new_overlay.sh <SCxx> <FILE_nnn>`  <sub>L1062</sub>
- **§3-THE** — NON-4-ALIGNED-OVERLAY GOTCHA (≈75% of the fleet; fixed in the template + Makefile, automatic)  <sub>L1079</sub>
- **Fleet** — build — `make build-all` / `make check-all`  <sub>L1120</sub>
- **§14** — Propagate a matched function across the fleet — `tools/dedup_propagate.py` (Phase 15)  <sub>L1126</sub>
- **§14b** — Harvesting the UNMATCHED shared core — the match_one wall (Phase 15)  <sub>L1192</sub>
- **§14c** — Callee-signature-aware harvest — breaking the extern-type-conflict wall (Phase 15, T6)  <sub>L1213</sub>
- **§17** — The compiler-quirk wall — the matching TOOLKIT (Phase 18; gcc-2.7.2 source + byte-gated)  <sub>L1376</sub>
- **§3-The** — pipeline gotcha — match_one ≠ the gate; `sig_unify` is MANDATORY  <sub>L1421</sub>
- **§17a** — The TOOLKIT at WAVE scale (Phase-18 Step-3b/Step-1 — measured) + the pipeline-integration gotchas  <sub>L1445</sub>
- **§19** — Scaling the toolkit waves — the recovery PIPELINE + the propagation CAP (Phase 19 T3)  <sub>L1597</sub>
- **§20** — The wave-at-scale GATE CAP + residual-class verdicts (Phase 20)  <sub>L1632</sub>
- **§3-THE** — CAP: at this tail the match_one→gate gap is the LOOSE-TYPING CALL-GRAPH wall (not cheap plumbing)  <sub>L1637</sub>
- **Diagnostic** — lesson: a failed in-TU build leaves a STALE `.o`  <sub>L1654</sub>
- **Phase-20** — RESOLUTION — `tools/cast_call_sites.py` BUILT + the cap re-diagnosed (R14, byte-proven)  <sub>L1701</sub>
- **§24** — The `ov_SC01_077_a.c` split-file vein: split-aware propagation, but loose-typing-gated (Phase 21)  <sub>L2085</sub>
- **§25** — The "schedule" class is mostly COALESCING (pin-crackable), not scheduling; + the gate two-stage + h_exact over-counts ×134 (Phase 21, cont.6)  <sub>L2132</sub>
- **§3-The** — gate two-stage — sig_unify is a FALLBACK, not unconditional (`gate_stage.py`, §19 folded in)  <sub>L2189</sub>
- **§28** — Banking a "close=0 gate-rejected" giant: the canonical-extern recovery (Phase 22 T1, byte-proven on func_8015126C ×134)  <sub>L2325</sub>
- **§28c** — The close=0 recovery is NOT fully exhausted (§26 corrected, R14); + the dedup_propagate registry-skip recovery (Phase 22 T2)  <sub>L2352</sub>
- **§28d** — The "macro-extern-injection" lever: freeing reach-134 inline matches dedup_propagate skips as "not self-contained" (Phase 23, `tools/inject_capped_externs.py`)  <sub>L2366</sub>
- **§29** — Reasoning-model (GLM5.2) DEF-side reconciliation idioms + the wall's hard limit (Phase 23 T10.7, `tools/glm_reconcile.py`)  <sub>L2374</sub>
- **§33** — Automating the giant decl-reconcile: `tools/reconcile_decls.py` (the DATA analog of §20's `cast_call_sites`) + a fleet-majority type oracle (Phase 24 T7b, byte-proven on `func_80129CF8`)  <sub>L2434</sub>
- **§34** — The `func_80138ED0` giant crack: gcc-2.7.2's **3-qty sort bug** + the **zero-byte asm allocation toolkit** + the **giv-init fence** (Phase 24 T5; Opus→close=21, Fable5→MATCH ×134)  <sub>L2454</sub>
- **§35** — The region-a sibling-giant harvest: difficulty ≠ $s-reg count (it's global-array hoisting) + the banking recipe + new loop idioms (Phase 24 T7, 5 parallel Opus-Max agents)  <sub>L2471</sub>
- **§37** — The T7 §G giant endgame: 6/8 cracked, meta-laws + transferable levers (Phase 24, 2026-07-07; FULL byte-verified detail + gcc-2.7.2 line cites in `docs/gcc-2.7.2-map/t7g-giant-harvest.md`)  <sub>L2506</sub>
- **§39** — The ×1→×134 giant-endgame: propagate a matched **-O2** giant via the NATIVE DEFINE-macro path (Phase 24 T7 §G close, 2026-07-08)  <sub>L2537</sub>
- **§41** — The DEF-SIDE canonical-sig wall: mechanically banking a drafted giant past `conflicting types` (Phase 25 T5b batch-2, 2026-07-09; `tools/canon_sig_reconcile.py`, byte-proven on `func_8013B274`)  <sub>L2710</sub>
- **§41a** — v3.1: the def-side wall was ~71% TOOL-shaped — the five measured defects + the laws that dissolve them (Phase 25 T6, Fable5, 2026-07-09)  <sub>L2754</sub>
- **§41b** — T7 execution: the object-only probe OVER-counts BANKABLE by two link/rodata classes (Phase 25 T7-M1, 2026-07-10)  <sub>L2809</sub>
- **§41d** — `void`→`s32` is NOT always byte-neutral: gate the RAW draft FIRST (Phase 26 session 6, byte-proven)  <sub>L2878</sub>
- **§42b** — addendum — wave 3 (Max, 2026-07-10c): THE STALE-OBJECT GATE TRAP + the read-global `&`-cast drift + fix  <sub>L2994</sub>
- **§45** — The flagship `func_80133CD4` crack (399 ins ×134): the merged-variable permutation-breaker + the 1-death local-alloc gate (Phase 25 task A giant escalation, Fable5 gdb-on-cc1, 2026-07-11)  <sub>L3298</sub>
- **§51** — TOOLING INTEGRITY: the silent skip, and how to hunt it  <sub>L3630</sub>
- **§51b** — Why the byte-gate cannot save you  <sub>L3652</sub>
- **§51f** — Checklist for any new corpus-scanning tool  <sub>L3755</sub>
- **§51g** — When the thing you are scanning has a GRAMMAR, parse the grammar (`tools/cdecl.py`)  <sub>L3769</sub>
- **§53** — SWEEP A FAMILY WITH THE TOOL ITS EXEMPLAR NEEDED: the jr/switch carve, and how omitting it manufactured the "families don't template" doctrine (Phase 28 T1, 2026-07-15)  <sub>L4028</sub>
- **§55** — Core-crack wave levers + the GATE-ORCHESTRATION law (Phase 29 T3, 13-agent ultracode wave, 2026-07-17)  <sub>L4093</sub>
- **§55b** — THE GATE-ORCHESTRATION LAW (3 traps, ~3.5h lost; all recovered, 0 data lost)  <sub>L4122</sub>
- **§55c** — Sizing the propagate: a TARGETED propagate is ~4 min/core, and "it's slow" was a BROKEN-TREE ARTIFACT  <sub>L4141</sub>
- **§54** — `--fix-def-sig`: the member's CANONICAL DECLARATION is a build step too (tiny-IMM mega-pools, +4,801, Phase 29 T6, 2026-07-16)  <sub>L4163</sub>
- **§57** — The SELF-decl normalize: the sibling's OWN caller declares the templated fn divergently (the third §17a-1 direction; `tools/normalize_self_decls.py`, func_801670E4 ×137, Phase 29, 2026-07-18)  <sub>L4256</sub>
- **§61** — Task 14: the gate ladder's missing stage is the ARITY pre-pass, and it is TU-side not draft-side (Phase 29, 2026-07-21)  <sub>L4545</sub>
- **§61b** — The jtbl gate stage: built, and the ORDERING law it exposed (Phase 29 Task-14 stage 4, 2026-07-21)  <sub>L4658</sub>
- **§61d** — The undo was eating the tree: two tools, one defect, invisible to the byte-gate (Phase 29, 2026-07-22)  <sub>L4779</sub>
- **§62** — The jtbl RECONCILE must also follow the carve: the post-carve draft reconcile (`harvest_verify._jtbl_reconcile`, Phase 29 SESSION-11, 2026-07-22)  <sub>L4833</sub>
- **§65** — The stranded-draft recovery: BLAST-RADIUS TIERS, and the per-overlay de-macroize that refutes §20's DEF-conflict wall (`tools/demacroize.py` + `tools/blocker_probe.py`, Phase 29 SESSION-16, 2026-07-24)  <sub>L5015</sub>
- **§65d** — Existing-ladder baseline, measured (do this before building a recovery stage)  <sub>L5091</sub>
- **§65e** — Two oracles, and the disagreement is the finding (R34 in practice)  <sub>L5101</sub>
- **§65f** — The de-macroize lever's BOUNDARY, measured: byte-neutral 14 times of 15, and the 15th is why the gate exists  <sub>L5118</sub>
- **§66b** — A metric parsed out of another tool's prose goes NULL silently when the label changes  <sub>L5200</sub>
- **§67a** — Run the symbol-set guard BEFORE you pay for a gate (`tools/symcheck.py`, Phase 29 SESSION-18)  <sub>L5405</sub>
- **§77** — Every extraction tool carries a NARROW hard-coded preamble set; anything outside it silently caps the body's reach. Diff the preamble before you gate. (Phase 29 SESSION-19 — three variants in one session, two different tools)  <sub>L6084</sub>
- **§3-The** — CANDIDATE gate and the REAL gate need DIFFERENT preambles — keep the difference out of the bank  <sub>L6124</sub>
- **§79** — For a 0-callee giant, fingerprint by DATA symbols (§71 cannot fire); and the STACK-SLOT ORDER is a declaration-order oracle (Phase 29 SESSION-19, `func_8017BF14` 4,763 ins, cold start → 45/4763)  <sub>L6220</sub>
- **§82** — Two source-shape oracles from behemoth #6: a duplicated `addiu $aN,$sp,K` across a `jal` means the block was INLINED, and scalar-vs-aggregate decides WHEN a stack slot is allocated (Phase 29 SESSION-19, `func_8017C730` 1,061 ins)  <sub>L6370</sub>
- **§3-2.** — Scalar vs aggregate decides *when* the slot is allocated  <sub>L6389</sub>
- **§3-Why** — it survived every candidate gate  <sub>L6495</sub>
- **§3-THE** — FIX IS MECHANICAL — the tool already holds the answer  <sub>L6510</sub>
- **§87** — `match_one` COMPILES but never LINKS, so an unresolvable data symbol reads as MATCH; and stored drafts go STALE against the tree (Phase 29 SESSION-20)  <sub>L6633</sub>
- **§88f** — the missing rung: a RELOCATION gate between `match_one` and the binary  <sub>L6718</sub>
- **§89a** — MEASURE the write set; do not assert its tier (`tools/blast_radius.py`)  <sub>L6732</sub>
- **§89b** — the parallel gate farm existed; the family path could not reach it (`tools/sweep_parallel.py`)  <sub>L6745</sub>
- **§90** — Five tool-integrity laws from one session, each of which changed an answer (Phase 29 SESSION-21, 2026-07-27)  <sub>L6766</sub>
- **§90a** — A comparison tool MUST share its reference oracle's index space, exactly  <sub>L6772</sub>
- **§92** — Conforming a DECLARATION: pointer changes are caller-neutral, scalar-WIDTH changes are not (Phase 29 SESSION-21, `tools/conform_decls.py`)  <sub>L6891</sub>
- **§97** — The gate's own tree hygiene: a refused carve, an unchecked recovery, and a snapshot that captured a dirty tree (Phase 29 SESSION-22)  <sub>L7066</sub>
- **§105** — A gate's revert must survive an EXCEPTION, not just a failure (Phase 29 T53, `jtbl_family_bank`)  <sub>L7431</sub>
- **§107** — A lever wired into ONE gate path is a lever most families cannot reach (Phase 29 T56, `func_80144090` 0/136 → 136/136)  <sub>L7513</sub>
- **§3-Why** — it survived so long: a MASKED oracle will MATCH a wrong symbol  <sub>L7953</sub>
- **§122** — GATE RAW BEFORE TRANSFORMING; the undo belongs to the WRITER, as a per-edit journal (P30 T0a, 2026-07-30)  <sub>L8084</sub>
- **§123** — PROPAGATE A FAMILY WITH THE TOOL ITS TIER NEEDS: `dedup_propagate` is h_exact-only; its refusals are statements about the TOOL (P30 wave 1, 2026-07-30)  <sub>L8116</sub>
- **§126a** — a bare `except: continue` around a coverage-asserting oracle re-creates the silent skip (P30 S28)  <sub>L8317</sub>
- **§130** — An INCREMENTAL build can report BYTE-IDENTICAL for a change the CLEAN build cannot even LINK (P30 S28, the jr pair)  <sub>L8486</sub>
- **§132** — The `JR-PAIR-IN-ONE-O0-OBJECT` "wall" was TWO instrument defects: a merged-double span the carve could not see, and a truncated object no rule deleted (P30 S29, `func_8013B83C` + `func_8013BD74`)  <sub>L8570</sub>
- **§133** — The DEFAULT-FILTER class: three times in one session, a tool silently answered a narrower question than the one asked (P30 S1–S3)  <sub>L8685</sub>
- **§134** — MULTI-LINE BLINDNESS: one root cause, four faces, in `family_remap`'s preamble scanner (P30 S6, 190 zero-crack families)  <sub>L8711</sub>
- **§3-The** — splitting/merging rules (each closed a residual, byte-gated)  <sub>L8853</sub>
- **§137a** — A gate verdict has a TIMESTAMP; re-check it against the draft's mtime  <sub>L9383</sub>
- **§138** — The propagation lanes: a gate refusal is a DECLARATION, and which lever you owe depends on blast radius  <sub>L9412</sub>
- **§134** — again, in a second tool — and the waiter rule corrected  <sub>L9523</sub>
- **Reconciling** — a gate-refused draft: which way you edit depends on WHERE the TU's decl is  <sub>L9562</sub>
- **§139** — A GATE THAT GREPS FOR VERDICTS MUST ASSERT 1:1 ACCOUNTING; and a `--src` filter must not survive a carve (P30 S38, wave 6: 10 of 16 drafts vanished)  <sub>L9591</sub>
- **§140** — A METRIC IS NOT A MEASUREMENT UNTIL IT IS REPRODUCIBLE FROM THE COMMITTED TREE (P30 S1e: a phantom regression that gated the session's best lever)  <sub>L9657</sub>
- **§142** — An open stub whose `h_exact` class is MATCHED elsewhere is FREE. Propagate the body; do not gate a draft. (P30 S39, +7,710 ins in two commands)  <sub>L9775</sub>
- **§3-The** — measurement (do this before any wave; it is ~20 lines and needs no builds)  <sub>L9787</sub>
- **§155b** — check the TYPE your oracle returns before comparing against it (S45 p5)  <sub>L10584</sub>
- **§156** — an ORPHANED reconcile poisons the fleet: `dedup_propagate`'s kept edit (S45 p6/p7)  <sub>L10637</sub>
- **§158** — The RANGE-EXTENDER: a fourth zero-emission asm lever completes the allocno toolkit (P30 S46 tier-3, `func_8017CE58`, 733 ins)  <sub>L10709</sub>
- **§160** — THE REACH-15 WAVE HARVEST: an align-1 block move, and the instrument that called it a wall (P30 S47)  <sub>L10890</sub>
- **§161** — THE RETRY-WAVE HARVEST: a jump table indexed from zero, and two allocator traps (P30 S47)  <sub>L10966</sub>
- **§162** — S48 WAVE-1 HARVEST (P30, 2026-08-11): the reach-ordered sibling campaign's first 12 targets  <sub>L11011</sub>
- **§163** — S48 WAVES 2-3 HARVEST (P30, 2026-08-11/12): the five that were byte-probed and are actionable  <sub>L11757</sub>
- **§165** — S48 WAVE-4 HARVEST (P30, 2026-08-12): banked the same day the wave landed  <sub>L13686</sub>
- **§166** — THE DESTINATION-TU ORACLE (P30 S48): the seven-attempt bug that was never codegen  <sub>L14899</sub>
- **§167** — S48 WAVE-5/6 HARVEST (P30, 2026-08-12): the saturation point  <sub>L14955</sub>

### process, measurement & doctrine  (73)

- **§8e** — The jtbl ALIGNMENT LAW + the pad-spec filter — multi-table .rodata spans (Phase 29, byte-proven; `.run/probe_jtbl/verdict.md`)  <sub>L530</sub>
- **§3-The** — mechanism: game-code dedup is SOURCE-LEVEL, not an object swap (R-D1, the key lesson)  <sub>L926</sub>
- **Per-binary** — toolchain provenance (R24)  <sub>L959</sub>
- **§17a** — The TOOLKIT at WAVE scale (Phase-18 Step-3b/Step-1 — measured) + the pipeline-integration gotchas  <sub>L1445</sub>
- **Diagnostic** — lesson: a failed in-TU build leaves a STALE `.o`  <sub>L1654</sub>
- **Phase-20** — RESOLUTION — `tools/cast_call_sites.py` BUILT + the cap re-diagnosed (R14, byte-proven)  <sub>L1701</sub>
- **§23** — Giant `func_80153E00` cracked (scalar-data CAST); BUT most giant drafts are STALE+INCOMPLETE, and a diag MUST remove artifacts (Phase 21 — byte-proven + a self-correction)  <sub>L2032</sub>
- **§3-h** — _exact OVER-COUNTS ×134 — verify shareability before crediting a class's "reach-134" (R14)  <sub>L2177</sub>
- **§28c** — The close=0 recovery is NOT fully exhausted (§26 corrected, R14); + the dedup_propagate registry-skip recovery (Phase 22 T2)  <sub>L2352</sub>
- **§31-triage** — R17 applies to CODEGEN residuals, never to a compile ERROR (Phase 26 session 7, Drew asked)  <sub>L2693</sub>
- **§41a** — v3.1: the def-side wall was ~71% TOOL-shaped — the five measured defects + the laws that dissolve them (Phase 25 T6, Fable5, 2026-07-09)  <sub>L2754</sub>
- **§41b** — T7 execution: the object-only probe OVER-counts BANKABLE by two link/rodata classes (Phase 25 T7-M1, 2026-07-10)  <sub>L2809</sub>
- **§41b** — addendum — M4 "reconcile_decls" tier is ALSO a probe over-count: 0/8 mechanical (Phase 25 T7-M4, 2026-07-10)  <sub>L2858</sub>
- **§53** — SWEEP A FAMILY WITH THE TOOL ITS EXEMPLAR NEEDED: the jr/switch carve, and how omitting it manufactured the "families don't template" doctrine (Phase 28 T1, 2026-07-15)  <sub>L4028</sub>
- **§3-The** — meta-lesson (R35, and why this one is expensive)  <sub>L4078</sub>
- **§55** — Core-crack wave levers + the GATE-ORCHESTRATION law (Phase 29 T3, 13-agent ultracode wave, 2026-07-17)  <sub>L4093</sub>
- **§55b** — THE GATE-ORCHESTRATION LAW (3 traps, ~3.5h lost; all recovered, 0 data lost)  <sub>L4122</sub>
- **§57a** — Two NSD corrections + the SURGICAL-ONLY law + the honest broad-sweep yield (Phase 29, 2026-07-18)  <sub>L4303</sub>
- **§58** — match_one MATCH ≠ BANK: the four blind spots + the crack-wave reconcile-before-bank law (Phase 29 crack-wave, 2026-07-18)  <sub>L4343</sub>
- **§60** — Classify the residual, don't rank it: the deterministic residual→class classifier and what it measured about the backlog (Phase 29 Task-13A, 2026-07-21)  <sub>L4407</sub>
- **What** — it measured — the whole open backlog, byte-grounded  <sub>L4431</sub>
- **§3-The** — parallel-probe race this surfaced  <sub>L4463</sub>
- **§61b** — The jtbl gate stage: built, and the ORDERING law it exposed (Phase 29 Task-14 stage 4, 2026-07-21)  <sub>L4658</sub>
- **§62** — The jtbl RECONCILE must also follow the carve: the post-carve draft reconcile (`harvest_verify._jtbl_reconcile`, Phase 29 SESSION-11, 2026-07-22)  <sub>L4833</sub>
- **§63** — UPDATE (Phase 29 SESSION-14) — `fix_header_decl`'s "SAFE" verdict is FLEET-BLIND; it MUST be R22-validated  <sub>L5000</sub>
- **§65** — The stranded-draft recovery: BLAST-RADIUS TIERS, and the per-overlay de-macroize that refutes §20's DEF-conflict wall (`tools/demacroize.py` + `tools/blocker_probe.py`, Phase 29 SESSION-16, 2026-07-24)  <sub>L5015</sub>
- **§65a** — The blast-radius taxonomy (makes §61's law structural instead of remembered)  <sub>L5030</sub>
- **§65d** — Existing-ladder baseline, measured (do this before building a recovery stage)  <sub>L5091</sub>
- **§65e** — Two oracles, and the disagreement is the finding (R34 in practice)  <sub>L5101</sub>
- **§65f** — The de-macroize lever's BOUNDARY, measured: byte-neutral 14 times of 15, and the 15th is why the gate exists  <sub>L5118</sub>
- **§66c** — Before a wave, verify the FUEL exists; an "already attempted" set built from the wrong directory lies in BOTH directions  <sub>L5217</sub>
- **§66d-5** — `residual_class`'s "structural ⇒ permuter CPU is waste" is WRONG for schedule permutations (measured, Phase 29 SESSION-18)  <sub>L5491</sub>
- **§3-The** — honest fix was source-level and cheap  <sub>L5750</sub>
- **§3-The** — residual, and the honest read  <sub>L6260</sub>
- **§83** — The parameterised-repeat law, the spill-area trap, and why a per-case edit cannot move a per-case symptom (Phase 29 SESSION-20, `func_80183814` 5,122 ins, cold-ish start → 36 structural / 99.3%)  <sub>L6412</sub>
- **Scope** — , measured (do not over-generalise — §80)  <sub>L6518</sub>
- **§3-Two** — ladder lessons banked with it  <sub>L6528</sub>
- **§3-Do** — the WHOLE axis in one edit, then R22 once  <sub>L6567</sub>
- **§3-THE** — LAW: all-or-nothing PER FAMILY  <sub>L6600</sub>
- **§88b** — the `slti` literal-position law (extends §78 to comparisons)  <sub>L6686</sub>
- **§88e** — a wrong diagnosis, refuted properly (the model for how to treat an inherited lever)  <sub>L6708</sub>
- **§89a** — MEASURE the write set; do not assert its tier (`tools/blast_radius.py`)  <sub>L6732</sub>
- **§90d** — Do not measure a live wave's drafts (§87 in real time)  <sub>L6810</sub>
- **§98** — `conform_decls` had three defects, and only the third needed R22 to find (Phase 29 SESSION-22, `func_8014CF04`)  <sub>L7113</sub>
- **§106** — Persist the MEASUREMENT, derive the POLICY: a stored route let a stale file out-vote the live table (Phase 29 T54, `residual_class._ROUTE`)  <sub>L7470</sub>
- **§125** — Split the CARVE from the BODY before calling a jr residue a wall — and measure it by SHA from a CLEAN tree (P30 SESSION-28; **this section's first draft was WRONG and the method caught it**)  <sub>L8206</sub>
- **§3-The** — meta-lesson  <sub>L8257</sub>
- **Wave** — economics (measured, for the next batch's sizing)  <sub>L8950</sub>
- **§136b** — A prior wave's "genuine byte-DIFF" verdict is NOT reliable evidence (4 of 4 refuted)  <sub>L9030</sub>
- **§136g** — When the index points at the WRONG lever: two byte-refuted routings (func_801863B4)  <sub>L9208</sub>
- **§136h** — CORRECTION: the zero-crack pool does NOT "refill with cheap work" (my error, byte-measured)  <sub>L9237</sub>
- **§136j** — The failure MIX flips with function size (measured across four bands, one session)  <sub>L9299</sub>
- **Rank** — the lane by measured concentration, not by class count  <sub>L9478</sub>
- **§140** — A METRIC IS NOT A MEASUREMENT UNTIL IT IS REPRODUCIBLE FROM THE COMMITTED TREE (P30 S1e: a phantom regression that gated the session's best lever)  <sub>L9657</sub>
- **§3-The** — measurement (do this before any wave; it is ~20 lines and needs no builds)  <sub>L9787</sub>
- **§3-And** — the report-vs-bytes lesson attached to it  <sub>L9819</sub>
- **§146** — RE-MEASURE A WALL BEFORE YOU RESPECT IT. Both "permanent" giants fell to drafts already on disk. (P30 S6, +50,094 ins)  <sub>L9974</sub>
- **§147** — The three-stratum FRAME LAW, and four "stop searching" verdicts (P30 S42, `func_8017C294`, serial run)  <sub>L10033</sub>
- **§3-C.** — Inner-block declaration does NOT delay slot allocation — BYTE-REFUTED  <sub>L10065</sub>
- **§3-D.** — "Cheap fuel" that was never probed: 0 of 31 templatable  <sub>L10290</sub>
- **§3-Two** — corrections to the record  <sub>L10342</sub>
- **§3-The** — two fallouts, and how to close them (both measured, in order)  <sub>L10392</sub>
- **What** — does NOT work (14 byte-measured probes)  <sub>L10483</sub>
- **§157** — the cheap-tier size cliff, measured (S45 p6)  <sub>L10684</sub>
- **§162** — The cross-jump "CALL veto" is a COUNT law, not a CALL law (BOUNDS §88a; P30 S48, `func_80189540`)  <sub>L11299</sub>
- **§3-The** — law  <sub>L11546</sub>
- **DIAGNOSTIC** — TELL — two faces, one law  <sub>L11588</sub>
- **§163** — S48 WAVES 2-3 HARVEST (P30, 2026-08-11/12): the five that were byte-probed and are actionable  <sub>L11757</sub>
- **§163z** — THE UNVETTED REMAINDER (do not cite as law; each needs a dedupe pass)  <sub>L11825</sub>
- **§164z** — REFUTED CLAIMS: do NOT re-derive these  <sub>L13620</sub>
- **§165z** — REFUTED THIS WAVE: do NOT re-derive  <sub>L14855</sub>
- **§167z** — REFUTED IN WAVES 5/6: do NOT re-derive  <sub>L16151</sub>
- **§168** — THE COUSIN TIER (P30 S49, 2026-08-12): h_seq's exact-hash brittleness, measured — and the similarity map above it  <sub>L16208</sub>

### (unbucketed — title matched no symptom vocabulary)  (169)

- **§3-How** — to use this  <sub>L30</sub>
- **§1** — Idiom catalog (asm pattern → C that produces it)  <sub>L39</sub>
- **§3-I3** — Division by a constant → magic multiply  <sub>L54</sub>
- **§3-I4** — Runtime (variable) division → `divu` + zero-check `break` (NEEDS `--expand-div`)  <sub>L60</sub>
- **§2** — Writing matching C (what makes gcc emit X)  <sub>L69</sub>
- **§3a** — Escalation TIER above the permuter — web-research the compiler internals (HIGH VALUE, proven)  <sub>L128</sub>
- **§3b** — §31-directed permuter mutation — bias the search over the class's levers (Phase 24 T5)  <sub>L147</sub>
- **Residual** — B — a `return <const>` materialised late / merged instead of distributed per-site  <sub>L867</sub>
- **§3-The** — A→B→C per-overlay workflow  <sub>L1091</sub>
- **Dedup-credit** — (§11) for overlays — two shapes  <sub>L1108</sub>
- **§14a** — The fleet bulk run (Phase 15 — `--auto-from`, 947 REAL → 74,527; 3.82% → 22.14% in one pass)  <sub>L1169</sub>
- **§14d** — Deterministic recovery beats agent waves on the hard tail (Phase 15, the final-session finding)  <sub>L1269</sub>
- **§14e** — Two hard-tail dead-ends (Phase 15 — documented so they aren't re-attempted)  <sub>L1298</sub>
- **§3-The** — STEERABLE idioms (byte-confirmed this phase)  <sub>L1404</sub>
- **Strategic** — conclusion — the match-% lever post-research  <sub>L1435</sub>
- **§3-NEW** — / CONFIRMED residual classes (this session)  <sub>L1660</sub>
- **Operational** — gotchas (cost real time)  <sub>L1681</sub>
- **What** — WORKED — the Phase-20 reusable wins  <sub>L1689</sub>
- **§21** — wave-distilled idioms (Phase 21)  <sub>L1752</sub>
- **§26** — The cheap close=0 recovery lever is EXHAUSTED; `idiom_loop --assess` was DOUBLY inflated (Phase 21, cont.7)  <sub>L2199</sub>
- **§3-The** — `_a` close=0 recovery banks 0/20 — same def-side wall as MAIN (0/40)  <sub>L2205</sub>
- **§27** — Giant matching recipe (Phase 21 cont.7b — validated on func_80176D94, 152 ins)  <sub>L2259</sub>
- **§28a** — decomp.wiki GCC patterns worth trying on BFM giants (decomp.wiki/compilers/GCC, raw at decompals/decompedia; PS1-applicable subset)  <sub>L2345</sub>
- **§31** — THE gcc-2.7.2 CODEGEN MAP: pass → residual → C-lever catalog (Phase 23; 4 Fable5 agents read the compiler source)  <sub>L2402</sub>
- **§36** — Fable5 giant-crack levers (Phase 24 T7 Fable5 batch; accumulates as each lands — full RTL dumps in `.run/t7/fable/`)  <sub>L2487</sub>
- **§42d** — addendum — wave 4 (rtu_match fan-out over the mapped frontier, 2026-07-10c): 24/26 MATCH, +5 durable levers  <sub>L3082</sub>
- **§3-B.** — THE EBB RULE — the general form of §46-L2  <sub>L3468</sub>
- **§3-E.** — Meta  <sub>L3518</sub>
- **§50** — Refinements that BOUND §47/§48 (from the `func_80135EB0` wall, 21→6; Phase 26 session 8)  <sub>L3576</sub>
- **§51a** — The bug class  <sub>L3636</sub>
- **§51c** — THE METHOD (do not audit by reading the regex)  <sub>L3662</sub>
- **§51d** — THE LAWS  <sub>L3677</sub>
- **§51e** — The false-wall pipeline (why this is not just hygiene)  <sub>L3739</sub>
- **§3-Why** — the wall is (probably) intrinsic  <sub>L3945</sub>
- **§3-The** — case  <sub>L4033</sub>
- **§3-The** — rule  <sub>L4062</sub>
- **§55a** — New byte-proven levers (each from a banked or near draft)  <sub>L4098</sub>
- **§60a** — What the first DIRECTED grinder run exposed (Phase 29 Task-13B, 2026-07-21)  <sub>L4472</sub>
- **§60b** — The plateau autopsy's verdict: a `partial` drift is a WRONG DRAFT, not a missing transform (Phase 29 Task-13B close, 2026-07-21)  <sub>L4510</sub>
- **§66** — Exercise a banking driver's SUCCESS path before pointing it at a wave: the free re-bank test (Phase 29 SESSION-17, 2026-07-24)  <sub>L5159</sub>
- **§66d-2** — Two operational sharp edges  <sub>L5289</sub>
- **§66d-3** — Read the ILS per-cycle SERIES, not its final best: a repeated score and a still-falling one look identical in a summary line and mean opposite things  <sub>L5300</sub>
- **§67** — The arg-copy PLACEMENT lever: launder a parameter into a fresh pseudo AT the statement where the target's copy lands (Phase 29 SESSION-18, `func_8014D820` 25 → 16)  <sub>L5315</sub>
- **§66d-4** — "ILS converged" means converged FOR THAT WEIGHT PROFILE, not a floor (amends §66d-3; Phase 29 SESSION-18)  <sub>L5438</sub>
- **§69** — How to attack a behemoth: map it, don't draft it (Phase 29 SESSION-18, `func_80183814`, 5,122 ins)  <sub>L5569</sub>
- **What** — is left, and what is byte-recorded as SPENT  <sub>L5766</sub>
- **§3-The** — mechanism, with citations  <sub>L6037</sub>
- **§3-Two** — diagnosis traps this function proved  <sub>L6064</sub>
- **Practice** — Practice  <sub>L6074</sub>
- **§3-The** — drift was an allocation decision, not missing code  <sub>L6180</sub>
- **§3-The** — economics  <sub>L6211</sub>
- **§71** — has a blind spot, and this is it  <sub>L6226</sub>
- **§3-The** — flagged "#1 move" LOST — and why the failure is informative  <sub>L6302</sub>
- **§78** — 's attribution primitive, run and reproduced  <sub>L6312</sub>
- **Cold-start** — economics, now complete  <sub>L6318</sub>
- **Also** — reproduced on this function  <sub>L6401</sub>
- **§3-And** — the banking footnote (§75a class A, one line)  <sub>L6405</sub>
- **§83a** — READ THE HEADLINE NUMBER CORRECTLY: a LENGTH drift makes `match_one`'s count meaningless  <sub>L6419</sub>
- **§83b** — THE LEVER: find the parameterised REPEAT before decoding case-by-case  <sub>L6427</sub>
- **§83e** — §80 vindicated again, on the same day it was written  <sub>L6466</sub>
- **§3-The** — conflict  <sub>L6544</sub>
- **§3-THE** — FAILURE MODE — widening only `engine_core.h` is worse than not starting  <sub>L6552</sub>
- **§3-The** — precondition, and how to check it in one grep  <sub>L6559</sub>
- **Reading** — , for the next person  <sub>L6576</sub>
- **§3-The** — guard refuses a class that largely works  <sub>L6591</sub>
- **§3-The** — blindness ladder, now complete — FOUR classes `match_one` cannot see  <sub>L6649</sub>
- **Consequence** — for the backlog ledger  <sub>L6659</sub>
- **§88** — `cross_jump` will not merge a common suffix containing a CALL; and §78 is scoped to ORDERED comparisons only (Phase 29 SESSION-20, the behemoth close-out)  <sub>L6674</sub>
- **§88a** — repeated CALL-shaped blocks are left UNMERGED; call-free tails are merged for you  <sub>L6679</sub>
- **§89** — Two throughput rules the project already had written down and was not following (Phase 29 SESSION-20)  <sub>L6726</sub>
- **§3-The** — standing sequence  <sub>L6758</sub>
- **§90b** — "Byte-neutral" is not "wanted": undo on the SUCCESS path too  <sub>L6789</sub>
- **§90c** — A library-callable function must FAIL CLOSED on an unconfigured module  <sub>L6799</sub>
- **§90e** — An agent's CONCLUSION and its EVIDENCE fail independently — re-derive the premise, not the fix  <sub>L6818</sub>
- **§93** — `set -o pipefail` attributes a pipeline failure to the LAST stage, not the failing one (Phase 29 SESSION-21, `func_8014D820`)  <sub>L6922</sub>
- **§101** — The STALE DEFAULT class: a guard whose cause was removed is a silent skip wearing a safety label (Phase 29 SESSION-22, three instances in one session)  <sub>L7245</sub>
- **§110** — A unit must define exactly ONE function, and "ends in `;`" does not tell you which line defines it (Phase 29 T65)  <sub>L7640</sub>
- **§115** — A `func_XXXXXXXX` predicate rots by design: the same name-form assumption in THREE places (Phase 29 T78)  <sub>L7850</sub>
- **§3-The** — fix moves the DEFINITION, not the stub — and here is why the obvious shortcut fails  <sub>L7895</sub>
- **§117** — Spell the sibling's symbol from the SIBLING's address, not the exemplar's kind (Phase 29 T82)  <sub>L7926</sub>
- **§118** — Ordinal (positional) immediate resolution: compare C tokens to the DIFFERING asm uses (Phase 29 T87)  <sub>L7963</sub>
- **§119** — Two levers on the SAME axis, opposite directions: test the off-diagonal (Phase 29 T89)  <sub>L7997</sub>
- **§3-The** — wiring trap that cost two attempts  <sub>L8043</sub>
- **§3-The** — method (keep this)  <sub>L8213</sub>
- **§3-The** — instrument rules that make its answer trustworthy (this is where I failed)  <sub>L8223</sub>
- **§3-The** — corrected results (each SHA-verified, from a clean tree, restore re-verified)  <sub>L8238</sub>
- **§3-The** — instrument trap that hid it (and it is §124's shape again)  <sub>L8290</sub>
- **§3-The** — mechanics  <sub>L8299</sub>
- **§3-The** — idiom they kept re-deriving: the constant-offset fold  <sub>L8356</sub>
- **§127b** — the knowledge was in a SOURCE COMMENT, not the cookbook  <sub>L8386</sub>
- **§128** — A raw NUL in C source makes grep SILENTLY SKIP the file (P30 S28, 137 files)  <sub>L8393</sub>
- **§128a** — a negative control must corrupt a SCRATCH COPY, never the tracked file  <sub>L8425</sub>
- **§3-The** — real blocker underneath, for the record  <sub>L8477</sub>
- **§3-The** — diagnostic ladder that finally located it (reusable)  <sub>L8523</sub>
- **Defect** — 1 — a pre-§8e MERGED DOUBLE is not a single-table predecessor  <sub>L8579</sub>
- **Defect** — 2 — `as` writes a corpse and nothing deletes it  <sub>L8597</sub>
- **§3-The** — fingerprint, and the 30-second ladder that found it  <sub>L8607</sub>
- **§3-The** — transferable rule  <sub>L8628</sub>
- **§132a** — `--like` is for a sibling with NO record; against one that HAS a record it over-derives (P30 S29, `ov_SC07_010`)  <sub>L8635</sub>
- **§135** — Six byte-verified gcc-2.7.2 idioms from the P30 S6f-h waves (and the two-lane wave shape that found them)  <sub>L8765</sub>
- **§3-The** — codegen idioms  <sub>L8770</sub>
- **§3-The** — wave shape that produced these  <sub>L8819</sub>
- **§136** — The LOCAL-VARIABLE lever: how many C locals, at what scope (P30 wave 4a, 25 byte-verified banks)  <sub>L8837</sub>
- **§136a** — Blocker capture: classify on the OUTPUT, never on the exit status  <sub>L8973</sub>
- **§136d** — Four gcc-2.7.2 levers the redraft lane found (each closed a residual no other lever moved)  <sub>L9091</sub>
- **§136e** — §136c's PRECONDITION, and two more symptom keys (wave 4b batch 3)  <sub>L9135</sub>
- **§136i** — The drafter model LADDER: Haiku → **Sonnet** → Opus → Fable5 (Drew, 2026-08-03)  <sub>L9267</sub>
- **§3-The** — triage, cheapest first  <sub>L9418</sub>
- **THREE** — carry variants hide in one "CARRY-FIXABLE" bucket — and they need different fixes  <sub>L9487</sub>
- **STEP** — 0 of sibling-first: grep `src/` for a distinctive LITERAL from the `.s`  <sub>L9543</sub>
- **§3-The** — generalisation — three corollaries worth more than the bug  <sub>L9623</sub>
- **§3-And** — the inverse-lookup trap, same session  <sub>L9640</sub>
- **§3-The** — three-line proof (do this before diagnosing any metric movement)  <sub>L9677</sub>
- **§3-The** — two instrument defects it exposed  <sub>L9686</sub>
- **§3-Two** — wrong mechanisms I chased first, and why they were wrong  <sub>L9717</sub>
- **§3-The** — trap that hid it — SAME FUNCTION, TWO ROUTES, ONLY ONE IS FREE  <sub>L9802</sub>
- **Route** — selection (why `--addr` sometimes says "nothing changed")  <sub>L9811</sub>
- **§144** — THE LITERAL'S SPELLING PICKS THE IMMEDIATE ENCODING (P30 S40 wave 1, `func_801822E0`)  <sub>L9885</sub>
- **§3-Why** — a correct draft can read as an intrinsic wall  <sub>L9985</sub>
- **§3-The** — rule  <sub>L10020</sub>
- **§3-B.** — `(v < 0x40) ? v : 0x3F` is folded to `MIN_EXPR` and expands to the WRONG SHAPE  <sub>L10167</sub>
- **§3-D.** — Reproduce the original's BUGS verbatim  <sub>L10191</sub>
- **§149** — Four instrument defects in one session, and the two questions they were hiding (P30 S43)  <sub>L10240</sub>
- **§3-A.** — A prep step that returns its input on failure is indistinguishable from a search that found nothing  <sub>L10246</sub>
- **§3-B.** — Same address + same name ≠ same body — and the ledger keys on address  <sub>L10264</sub>
- **§3-C.** — `make: *** [...] Error N` is a summary, never a diagnosis  <sub>L10280</sub>
- **§3-The** — fix  <sub>L10316</sub>
- **§3-The** — method that found it (this is the transferable part)  <sub>L10326</sub>
- **Diagnostic** — order (adopt this)  <sub>L10353</sub>
- **§3-The** — mechanism (read from cc1's own `-dR` trace, not inferred)  <sub>L10369</sub>
- **§3-The** — lever — a zero-emission insn that absorbs the blocked tick  <sub>L10384</sub>
- **§3-The** — finding  <sub>L10420</sub>
- **§3-The** — key  <sub>L10429</sub>
- **§3-Two** — cautions that must travel with this technique  <sub>L10440</sub>
- **§3-The** — companion defect (open)  <sub>L10451</sub>
- **Symptom** — Symptom  <sub>L10471</sub>
- **Mechanism** — (gcc source + RTL dumps, not inferred)  <sub>L10476</sub>
- **§3-The** — cure — a fresh launder per site, each in its own block  <sub>L10489</sub>
- **Companion** — levers from the same function  <sub>L10499</sub>
- **§3-B.** — Two static base-derivation methods that must AGREE (use both)  <sub>L10533</sub>
- **§155a** — the same failure class, one level up: SHAPE-blind table scanning (S45 p5)  <sub>L10564</sub>
- **§155c** — the ZERO-REFERENCE trap: gcc splits a global-array address across the `lui` and the LOAD (S46)  <sub>L10607</sub>
- **Symptom** — Symptom  <sub>L10718</sub>
- **§3-Why** — the fork is chained (gcc source, validated insn-by-insn against -dS/-dR dumps)  <sub>L10724</sub>
- **§3-The** — method (dump-arithmetic first, then place — no probing)  <sub>L10737</sub>
- **§162a** — SHARPENS  *(sharpens §161a, §131, §8a-pad, §129a)*  <sub>L11038</sub>
- **§162b** — SHARPENS  *(sharpens §48-A3, §156, §150, §76)*  <sub>L11067</sub>
- **§162d** — SHARPENS  *(sharpens §31, §21, §30, §55a)*  <sub>L11096</sub>
- **§162e** — NEW  <sub>L11138</sub>
- **§162f** — SHARPENS  *(sharpens §42d, §41d, §73, §10)*  <sub>L11206</sub>
- **§162g** — NEW  <sub>L11261</sub>
- **§162h** — SHARPENS  *(sharpens §88, §88a, §50-B, §8)*  <sub>L11297</sub>
- **§162i** — SHARPENS  *(sharpens §135, §21, §42, §32)*  <sub>L11365</sub>
- **§162j** — SHARPENS  *(sharpens §25, §136d-1, §48-B, §46-L2)*  <sub>L11390</sub>
- **§162k** — SHARPENS  *(sharpens §1-I2, §12, §160d, §21)*  <sub>L11419</sub>
- **§162l** — SHARPENS  *(sharpens §48-B, §48-C1, §20, §21)*  <sub>L11482</sub>
- **§162m** — SHARPENS  *(sharpens §36, §158, §148, §153)*  <sub>L11537</sub>
- **§158a** — THE FIFTH LEVER IS NOT AN ASM: `do { } while (0)` is a REGION ref-multiplier you MINT (P30 S48, `func_8017CBC8`, ov_MAIN_012 / jr_801789AC, 188 ins → MATCH)  <sub>L11539</sub>
- **Size** — it before you write it (§158 step 1-2, applied)  <sub>L11557</sub>
- **§3-The** — wrap BOUNDARY is the dial — and it is indiscriminate  <sub>L11578</sub>
- **§3-Not** — a pure dial  <sub>L11584</sub>
- **§162n** — NEW  <sub>L11610</sub>
- **§162o** — SHARPENS  *(sharpens §158, §136-1, §136-6, §79)*  <sub>L11643</sub>
- **§162p** — SHARPENS  *(sharpens §48-B, §46-L2, §156, §136d-1)*  <sub>L11702</sub>
- **§162q** — SHARPENS  *(sharpens §30, §30a, §135-2, §136-13)*  <sub>L11719</sub>
- **§164** — S48 §163z SKEPTIC PASS (P30, 2026-08-12): 190 claims vetted, 82 banked  <sub>L11845</sub>
- **§169** — THE MICRO-ADAPT LANE (P30 S49): edit a proven body, don't crack a new one  <sub>L16260</sub>
- **§170** — THE A-PROP WORD-DIFF CARD (P30 S49): the lane that had no owner  <sub>L16307</sub>
- **§171** — THE STALE SEED SYMBOL (P30 S50, 2026-08-13): why §170's 91%→57% was never codegen  <sub>L16351</sub>


## All sections, in order

- **§3-How** — to use this  <sub>L30</sub>
- **§1** — Idiom catalog (asm pattern → C that produces it)  <sub>L39</sub>
- **§3-I1** — Unsigned range check: `(x - lo) < (hi-lo)` → `addiu`+`sltiu`  <sub>L41</sub>
- **§3-I2** — Byte mask forces `andi` even after `lbu`  <sub>L47</sub>
- **§3-I3** — Division by a constant → magic multiply  <sub>L54</sub>
- **§3-I4** — Runtime (variable) division → `divu` + zero-check `break` (NEEDS `--expand-div`)  <sub>L60</sub>
- **§2** — Writing matching C (what makes gcc emit X)  <sub>L69</sub>
- **§3-T1** — Loop pointer: top-of-body for `addu` induction, not constant-folded `addiu`  <sub>L71</sub>
- **§3-T2** — Source statement order drives instruction scheduling  <sub>L78</sub>
- **§3-T3** — Types  <sub>L85</sub>
- **§3-T4** — Branch polarity: invert the source condition to flip gcc's chosen branch  <sub>L90</sub>
- **§3** — When a diff is pure scheduling → decomp-permuter (harness built, Phase 6)  <sub>L107</sub>
- **§3a** — Escalation TIER above the permuter — web-research the compiler internals (HIGH VALUE, proven)  <sub>L128</sub>
- **§3b** — §31-directed permuter mutation — bias the search over the class's levers (Phase 24 T5)  <sub>L147</sub>
- **§4** — Flag/toolchain gotchas  <sub>L190</sub>
- **§5** — Known hard-residual classes (instruction-identical, one byte-exact blocker)  <sub>L199</sub>
- **§5a** — Cross-jump tail-merge — gcc collapses two byte-identical blocks the original kept separate (FIX FOUND)  <sub>L211</sub>
- **§6** — Per-module optimization mixing — the -O0 boot module (Phase 7)  <sub>L246</sub>
- **Detecting** — the opt level (do this first)  <sub>L254</sub>
- **Build** — mechanism — per-file opt override (splat resegmentation)  <sub>L288</sub>
- **§7** — PsyQ SDK types & symbols (the library-call prerequisite)  <sub>L303</sub>
- **§8** — rodata island (compiler jump tables) — the `.data→.rodata→.data` sandwich (Phase 7)  <sub>L320</sub>
- **§8a** — rodata island in a flat OVERLAY — the tail sandwich, per matched jr-function (Phase 26 — PoC PROVEN)  <sub>L342</sub>
- **§8a-pad** — a trailing `.word 0x00000000` under a jtbl dlabel is `.align` PAD, not an entry (Phase 26 session 6, byte-proven)  <sub>L380</sub>
- **§8b** — MULTI-jtbl per overlay — the `ld_interleave --order` sandwich + the same-subseg cases (Phase 26 session 4)  <sub>L404</sub>
- **§8c** — Splitting a TU means rebuilding its DECLARATION ENVIRONMENT, not moving text (Phase 26 session 6)  <sub>L437</sub>
- **§8d** — Templating a body INTO a TU must not CHANGE its declaration environment — demote the carried data externs (Phase 26 session 8, byte-proven on `func_8015AE2C` ×133)  <sub>L483</sub>
- **§8e** — The jtbl ALIGNMENT LAW + the pad-spec filter — multi-table .rodata spans (Phase 29, byte-proven; `.run/probe_jtbl/verdict.md`)  <sub>L530</sub>
- **§9** — Link real PsyQ library objects byte-exact (Phase 7 — GO proven)  <sub>L613</sub>
- **§9.1** — Generalised per-object linker — `tools/psyq_link.py` (+ `psyq_link_lib.py`), 18/18 libcd byte-exact  <sub>L642</sub>
- **§9.2** — Wire a library region into the build with NOLOAD — no data carving (`tools/psyq_link_region.py`)  <sub>L666</sub>
- **§9.3** — Make it the build: resegment + swap + resolve (`tools/psyq_integrate.py`, libcd DONE)  <sub>L685</sub>
- **§9.4** — Integrating a SECOND library (libgs block 6 after libcd) — multi-library gotchas  <sub>L712</sub>
- **§9.5** — Integrating a WHOLE multi-block library in one call (full libgs — Phase 7 session G)  <sub>L740</sub>
- **§9.6** — Scaling library linking to the whole EXE (Phase 8 — 8 libs linked, 20%→50% byte-identical)  <sub>L766</sub>
- **§9.7** — Binary-agnostic toolchain refactor (Phase 9) — the reusable pattern for Gen2  <sub>L802</sub>
- **§10** — Closing the regalloc/scheduling hard tail by hand (LZSS, Phase 7 session F — the full close)  <sub>L835</sub>
- **§3-The** — clean object-level metric (use this, not the permuter score, for jtbl/rodata functions)  <sub>L844</sub>
- **Residual** — A — commutative `|`/`&`/`+` result lands in the wrong source-operand register  <sub>L856</sub>
- **Residual** — B — a `return <const>` materialised late / merged instead of distributed per-site  <sub>L867</sub>
- **§11** — Cross-binary dedup & code-sharing (Phase 11 — "one match unlocks many")  <sub>L908</sub>
- **§3-The** — mechanism: game-code dedup is SOURCE-LEVEL, not an object swap (R-D1, the key lesson)  <sub>L926</sub>
- **Per-binary** — toolchain provenance (R24)  <sub>L959</sub>
- **§12** — Ultracode harvest — parallel-draft + byte-gate at scale (Phase 12, resident: 1.4%→71.7% in one session)  <sub>L964</sub>
- **§13** — Add a location overlay — the canonical runbook (Phase 13; the Phase-15 fleet recipe)  <sub>L1056</sub>
- **One-command** — onboarding — `tools/new_overlay.sh <SCxx> <FILE_nnn>`  <sub>L1062</sub>
- **§3-The** — flat-blob overlay config (what the template encodes)  <sub>L1071</sub>
- **§3-THE** — NON-4-ALIGNED-OVERLAY GOTCHA (≈75% of the fleet; fixed in the template + Makefile, automatic)  <sub>L1079</sub>
- **§3-The** — A→B→C per-overlay workflow  <sub>L1091</sub>
- **Dedup-credit** — (§11) for overlays — two shapes  <sub>L1108</sub>
- **Fleet** — build — `make build-all` / `make check-all`  <sub>L1120</sub>
- **§14** — Propagate a matched function across the fleet — `tools/dedup_propagate.py` (Phase 15)  <sub>L1126</sub>
- **§14a** — The fleet bulk run (Phase 15 — `--auto-from`, 947 REAL → 74,527; 3.82% → 22.14% in one pass)  <sub>L1169</sub>
- **§14b** — Harvesting the UNMATCHED shared core — the match_one wall (Phase 15)  <sub>L1192</sub>
- **§14c** — Callee-signature-aware harvest — breaking the extern-type-conflict wall (Phase 15, T6)  <sub>L1213</sub>
- **§14d** — Deterministic recovery beats agent waves on the hard tail (Phase 15, the final-session finding)  <sub>L1269</sub>
- **§14e** — Two hard-tail dead-ends (Phase 15 — documented so they aren't re-attempted)  <sub>L1298</sub>
- **§15** — Struct-heavy shared-core pipeline (Phase 16) — empirical determinations (S0)  <sub>L1313</sub>
- **§16** — Guided hand-matching the struct-heavy core (Phase 17 — beats the §15 brute-force)  <sub>L1335</sub>
- **§17** — The compiler-quirk wall — the matching TOOLKIT (Phase 18; gcc-2.7.2 source + byte-gated)  <sub>L1376</sub>
- **Register-allocation** — ORDER (call-crossing $s0/$s1 swap) → FORCE it with register pins (byte-proven)  <sub>L1384</sub>
- **§3-The** — STEERABLE idioms (byte-confirmed this phase)  <sub>L1404</sub>
- **§3-The** — pipeline gotcha — match_one ≠ the gate; `sig_unify` is MANDATORY  <sub>L1421</sub>
- **§3-The** — LOOSE-TYPING wall is real for narrow params (Phase 16, reconfirmed)  <sub>L1428</sub>
- **Strategic** — conclusion — the match-% lever post-research  <sub>L1435</sub>
- **§17a** — The TOOLKIT at WAVE scale (Phase-18 Step-3b/Step-1 — measured) + the pipeline-integration gotchas  <sub>L1445</sub>
- **§18** — Per-file `-O0` split inside an overlay/blob (Phase 19 T1)  <sub>L1519</sub>
- **§19** — Scaling the toolkit waves — the recovery PIPELINE + the propagation CAP (Phase 19 T3)  <sub>L1597</sub>
- **§20** — The wave-at-scale GATE CAP + residual-class verdicts (Phase 20)  <sub>L1632</sub>
- **§3-THE** — CAP: at this tail the match_one→gate gap is the LOOSE-TYPING CALL-GRAPH wall (not cheap plumbing)  <sub>L1637</sub>
- **Diagnostic** — lesson: a failed in-TU build leaves a STALE `.o`  <sub>L1654</sub>
- **§3-NEW** — / CONFIRMED residual classes (this session)  <sub>L1660</sub>
- **Operational** — gotchas (cost real time)  <sub>L1681</sub>
- **What** — WORKED — the Phase-20 reusable wins  <sub>L1689</sub>
- **Phase-20** — RESOLUTION — `tools/cast_call_sites.py` BUILT + the cap re-diagnosed (R14, byte-proven)  <sub>L1701</sub>
- **§21** — wave-distilled idioms (Phase 21)  <sub>L1752</sub>
- **§22** — DEF-side loose-typing recovery + grinder blacklist (Phase 21)  <sub>L2005</sub>
- **§23** — Giant `func_80153E00` cracked (scalar-data CAST); BUT most giant drafts are STALE+INCOMPLETE, and a diag MUST remove artifacts (Phase 21 — byte-proven + a self-correction)  <sub>L2032</sub>
- **§24** — The `ov_SC01_077_a.c` split-file vein: split-aware propagation, but loose-typing-gated (Phase 21)  <sub>L2085</sub>
- **§25** — The "schedule" class is mostly COALESCING (pin-crackable), not scheduling; + the gate two-stage + h_exact over-counts ×134 (Phase 21, cont.6)  <sub>L2132</sub>
- **§3-The** — crack: the residual was gcc-2.7.2 COPY-COALESCING, not the instruction scheduler  <sub>L2144</sub>
- **§3-The** — genuine scheduler — `rank_for_schedule` (sched.c), for when it IS scheduling  <sub>L2160</sub>
- **§3-The** — genuine schedule WALLS (do NOT re-grind — stub)  <sub>L2169</sub>
- **§3-h** — _exact OVER-COUNTS ×134 — verify shareability before crediting a class's "reach-134" (R14)  <sub>L2177</sub>
- **§3-The** — gate two-stage — sig_unify is a FALLBACK, not unconditional (`gate_stage.py`, §19 folded in)  <sub>L2189</sub>
- **§26** — The cheap close=0 recovery lever is EXHAUSTED; `idiom_loop --assess` was DOUBLY inflated (Phase 21, cont.7)  <sub>L2199</sub>
- **§3-The** — `_a` close=0 recovery banks 0/20 — same def-side wall as MAIN (0/40)  <sub>L2205</sub>
- **Where** — this leaves the reach-134 tail (cont.6 option-3, now CONFIRMED byte-backed)  <sub>L2233</sub>
- **§27** — Giant matching recipe (Phase 21 cont.7b — validated on func_80176D94, 152 ins)  <sub>L2259</sub>
- **§28** — Banking a "close=0 gate-rejected" giant: the canonical-extern recovery (Phase 22 T1, byte-proven on func_8015126C ×134)  <sub>L2325</sub>
- **§28a** — decomp.wiki GCC patterns worth trying on BFM giants (decomp.wiki/compilers/GCC, raw at decompals/decompedia; PS1-applicable subset)  <sub>L2345</sub>
- **§28c** — The close=0 recovery is NOT fully exhausted (§26 corrected, R14); + the dedup_propagate registry-skip recovery (Phase 22 T2)  <sub>L2352</sub>
- **§28b** — The struct-walled close=0 giant (§28 case #3) IS bankable: the engine_types.h type-lift (Phase 22 T2, byte-proven on func_80156B74 ×134)  <sub>L2357</sub>
- **§28d** — The "macro-extern-injection" lever: freeing reach-134 inline matches dedup_propagate skips as "not self-contained" (Phase 23, `tools/inject_capped_externs.py`)  <sub>L2366</sub>
- **§29** — Reasoning-model (GLM5.2) DEF-side reconciliation idioms + the wall's hard limit (Phase 23 T10.7, `tools/glm_reconcile.py`)  <sub>L2374</sub>
- **§30** — Fable5Max cracks §20/§10 "unsteerable" from the gcc SOURCE: store-vs-load is a `/s` aliasing flag, the def-side wall has a macro escape, + the birthing-boost (Phase 23, byte-proven on giant `func_8014EE14` 248 ins ×134)  <sub>L2382</sub>
- **§30a** — §30 generalizes to the FULL near-miss backlog (via STANDARD Opus agents, not just Fable5) + 2 more steer levers + the mechanical-integration throughput unlock (Phase 23 (a))  <sub>L2394</sub>
- **§31** — THE gcc-2.7.2 CODEGEN MAP: pass → residual → C-lever catalog (Phase 23; 4 Fable5 agents read the compiler source)  <sub>L2402</sub>
- **§32** — The region-a CAMERA-GIANT idiom set: struct-base hoisting + 4 sibling levers (Phase 24 T7, Fable5-cracked on `func_80129CF8` 191 ins, match_one MATCH; transferable to the 6 sibling giants)  <sub>L2422</sub>
- **§33** — Automating the giant decl-reconcile: `tools/reconcile_decls.py` (the DATA analog of §20's `cast_call_sites`) + a fleet-majority type oracle (Phase 24 T7b, byte-proven on `func_80129CF8`)  <sub>L2434</sub>
- **§34** — The `func_80138ED0` giant crack: gcc-2.7.2's **3-qty sort bug** + the **zero-byte asm allocation toolkit** + the **giv-init fence** (Phase 24 T5; Opus→close=21, Fable5→MATCH ×134)  <sub>L2454</sub>
- **§35** — The region-a sibling-giant harvest: difficulty ≠ $s-reg count (it's global-array hoisting) + the banking recipe + new loop idioms (Phase 24 T7, 5 parallel Opus-Max agents)  <sub>L2471</sub>
- **§36** — Fable5 giant-crack levers (Phase 24 T7 Fable5 batch; accumulates as each lands — full RTL dumps in `.run/t7/fable/`)  <sub>L2487</sub>
- **§37** — The T7 §G giant endgame: 6/8 cracked, meta-laws + transferable levers (Phase 24, 2026-07-07; FULL byte-verified detail + gcc-2.7.2 line cites in `docs/gcc-2.7.2-map/t7g-giant-harvest.md`)  <sub>L2506</sub>
- **§38** — The WHALE `func_80144B9C` (770 ins): the -O0 struct-assign memcpy idiom + the -O0 reach-134 ×134 rollout (Phase 24 T7 §G, cheap Opus — no Fable5, no calls.c)  <sub>L2527</sub>
- **§39** — The ×1→×134 giant-endgame: propagate a matched **-O2** giant via the NATIVE DEFINE-macro path (Phase 24 T7 §G close, 2026-07-08)  <sub>L2537</sub>
- **§40** — Structural families: the MECHANICAL symbol-remap (crack one exemplar → remap the rest, ~0 tokens) (Phase 25 T3/T7, 2026-07-08)  <sub>L2553</sub>
- **§40a** — The DECL-RECONCILE pass (type-lift so remapped drafts compile in the sibling TU) (Phase 25 T7.2, 2026-07-08)  <sub>L2601</sub>
- **§40b** — The reloc-tracker blind spot: the indexed-global idiom that hid the "reach-1 tail" (Phase 26 Task 1, 2026-07-11, byte-verified V0/V1)  <sub>L2634</sub>
- **§40c** — The h_seq per-sibling reconcile: templating a reconcile-class crack ×134 (Phase 26 Task 8, 2026-07-12, byte-proven)  <sub>L2676</sub>
- **§31-triage** — R17 applies to CODEGEN residuals, never to a compile ERROR (Phase 26 session 7, Drew asked)  <sub>L2693</sub>
- **§41** — The DEF-SIDE canonical-sig wall: mechanically banking a drafted giant past `conflicting types` (Phase 25 T5b batch-2, 2026-07-09; `tools/canon_sig_reconcile.py`, byte-proven on `func_8013B274`)  <sub>L2710</sub>
- **§41a** — v3.1: the def-side wall was ~71% TOOL-shaped — the five measured defects + the laws that dissolve them (Phase 25 T6, Fable5, 2026-07-09)  <sub>L2754</sub>
- **§41b** — T7 execution: the object-only probe OVER-counts BANKABLE by two link/rodata classes (Phase 25 T7-M1, 2026-07-10)  <sub>L2809</sub>
- **§41c** — T7-M2: the ×134 def-side-wall sweep via per-sibling RE-reconcile (Phase 25, 2026-07-10; 4,389 banks, ~0 agent tokens)  <sub>L2839</sub>
- **§41b** — addendum — M4 "reconcile_decls" tier is ALSO a probe over-count: 0/8 mechanical (Phase 25 T7-M4, 2026-07-10)  <sub>L2858</sub>
- **§41d** — `void`→`s32` is NOT always byte-neutral: gate the RAW draft FIRST (Phase 26 session 6, byte-proven)  <sub>L2878</sub>
- **§42** — The F-band ≤28 regalloc crack wave: register-pin/DENSITY levers beat the permuter (Phase 25 T7 F-band, 2026-07-10; Ultracode 9-worker wave, 4/9 banked byte-identical, 266 swept ×134)  <sub>L2900</sub>
- **§42a** — addendum — wave 2 (residuals + 29-100 band): iso-MATCH ≠ real-TU bank, the memcpy→struct-assign fix, +5 levers (2026-07-10b)  <sub>L2953</sub>
- **§42b** — addendum — wave 3 (Max, 2026-07-10c): THE STALE-OBJECT GATE TRAP + the read-global `&`-cast drift + fix  <sub>L2994</sub>
- **§42c** — addendum — wave 3 (Max orchestrator + CORRECTED Ultracode fan-out, 2026-07-10c): the real-TU-faithful parallel harness (rtu_match) + 7/9 crack, ZERO iso-drift  <sub>L3033</sub>
- **§42d** — addendum — wave 4 (rtu_match fan-out over the mapped frontier, 2026-07-10c): 24/26 MATCH, +5 durable levers  <sub>L3082</sub>
- **§42e** — propagating a CRACK ×134: the def-finder bug + the byte-drift residual (the "remap-fail" misdiagnosis)  <sub>L3121</sub>
- **§43** — The K&R s16-param definition DISSOLVES the "narrow-param wall" for by-value register args (Phase 25 task A, Fable5 crack of the 369-ins giant `func_80166994` ×134, 2026-07-11)  <sub>L3195</sub>
- **§44** — The Phase-25 cheap-Opus giant batch: 5 structural levers + the §43 extension (2026-07-11, 6 crackers over the frontier giants)  <sub>L3242</sub>
- **§45** — The flagship `func_80133CD4` crack (399 ins ×134): the merged-variable permutation-breaker + the 1-death local-alloc gate (Phase 25 task A giant escalation, Fable5 gdb-on-cc1, 2026-07-11)  <sub>L3298</sub>
- **§46** — The `func_80178D40` crack (890 ins ×134, the heaviest core in the game): four LOOP-STRUCTURE levers cheap-Opus found by reading loop.c/jump.c/cse.c (Phase 26 session 8, 2026-07-13)  <sub>L3313</sub>
- **§47** — The live-length SLIDER: splitting a global.c allocno-priority TIE with one zero-byte asm (Phase 26 session 8, Fable5 Max, byte-proven on `func_8017BEBC` 952 ins ×113)  <sub>L3365</sub>
- **§48** — The 12-core jr crack wave: the ALLOCNO-PRICING dials and the EBB rule (Phase 26 session 8, Ultracode, 9/12 MATCH first pass)  <sub>L3405</sub>
- **§3-A.** — ALLOCNO-PRICING DIALS — move a value into the register you want, byte-neutrally  <sub>L3413</sub>
- **§3-A4** — SINK THE CONSUMER CALL INTO THE ARMS (the inverse of A1; `func_8016AB6C`, byte-proven)  <sub>L3444</sub>
- **§3-B.** — THE EBB RULE — the general form of §46-L2  <sub>L3468</sub>
- **§3-C.** — TYPE- AND SHAPE-DRIVEN CODEGEN (the C type literally selects the addressing mode)  <sub>L3484</sub>
- **§3-D.** — THE CROSS-JUMP RATCHET (the sharpest new trap — `func_80131340` L-C)  <sub>L3508</sub>
- **§3-E.** — Meta  <sub>L3518</sub>
- **§49** — The LUID DIAL: a zero-byte SCHEDULING dial (the sched.c analogue of §47) — `func_8017A4AC` (536 ins ×134), Phase 26 session 8  <sub>L3527</sub>
- **§50** — Refinements that BOUND §47/§48 (from the `func_80135EB0` wall, 21→6; Phase 26 session 8)  <sub>L3576</sub>
- **§51** — TOOLING INTEGRITY: the silent skip, and how to hunt it  <sub>L3630</sub>
- **§51a** — The bug class  <sub>L3636</sub>
- **§51b** — Why the byte-gate cannot save you  <sub>L3652</sub>
- **§51c** — THE METHOD (do not audit by reading the regex)  <sub>L3662</sub>
- **§51d** — THE LAWS  <sub>L3677</sub>
- **§51e** — The false-wall pipeline (why this is not just hygiene)  <sub>L3739</sub>
- **§51f** — Checklist for any new corpus-scanning tool  <sub>L3755</sub>
- **§51g** — When the thing you are scanning has a GRAMMAR, parse the grammar (`tools/cdecl.py`)  <sub>L3769</sub>
- **§52** — The WALKER-FAMILY skeleton: 6 regalloc-order levers + a deeply-characterized intrinsic wall (`func_80178004`, 165 ins ×134; Phase 26, Fable5, 2026-07-15)  <sub>L3913</sub>
- **§3-The** — 6 levers (the "walker-family skeleton" — apply to the 12 siblings; levers 1-5 retire ~half the residual)  <sub>L3924</sub>
- **§3-Why** — the wall is (probably) intrinsic  <sub>L3945</sub>
- **§52a** — The regalloc sibling wave: new levers + two new wall classes (cheap-Opus applying §52, 2026-07-15)  <sub>L3959</sub>
- **§52b** — Sibling wave 2: more de-pin levers, a third wall class, and the match_one→whole-binary gap at scale (2026-07-15)  <sub>L3997</sub>
- **§53** — SWEEP A FAMILY WITH THE TOOL ITS EXEMPLAR NEEDED: the jr/switch carve, and how omitting it manufactured the "families don't template" doctrine (Phase 28 T1, 2026-07-15)  <sub>L4028</sub>
- **§3-The** — case  <sub>L4033</sub>
- **§3-Why** — 0/8 was structural, and predictable from two words  <sub>L4045</sub>
- **§3-The** — rule  <sub>L4062</sub>
- **§3-The** — meta-lesson (R35, and why this one is expensive)  <sub>L4078</sub>
- **§55** — Core-crack wave levers + the GATE-ORCHESTRATION law (Phase 29 T3, 13-agent ultracode wave, 2026-07-17)  <sub>L4093</sub>
- **§55a** — New byte-proven levers (each from a banked or near draft)  <sub>L4098</sub>
- **§55b** — THE GATE-ORCHESTRATION LAW (3 traps, ~3.5h lost; all recovered, 0 data lost)  <sub>L4122</sub>
- **§55c** — Sizing the propagate: a TARGETED propagate is ~4 min/core, and "it's slow" was a BROKEN-TREE ARTIFACT  <sub>L4141</sub>
- **§54** — `--fix-def-sig`: the member's CANONICAL DECLARATION is a build step too (tiny-IMM mega-pools, +4,801, Phase 29 T6, 2026-07-16)  <sub>L4163</sub>
- **§56** — Banking a hand-drafted GIANT into its exemplar TU: self-contained draft vs live-TU decls, 4 reconciliation tactics (`func_8013FAF8`, 312 ins ×138, Phase 29 T4, 2026-07-17)  <sub>L4187</sub>
- **§56b** — PROPAGATING an h_seq giant: the exemplar's externs MUST be fleet-canonical, not the draft's types  <sub>L4232</sub>
- **§57** — The SELF-decl normalize: the sibling's OWN caller declares the templated fn divergently (the third §17a-1 direction; `tools/normalize_self_decls.py`, func_801670E4 ×137, Phase 29, 2026-07-18)  <sub>L4256</sub>
- **§57a** — Two NSD corrections + the SURGICAL-ONLY law + the honest broad-sweep yield (Phase 29, 2026-07-18)  <sub>L4303</sub>
- **§58** — match_one MATCH ≠ BANK: the four blind spots + the crack-wave reconcile-before-bank law (Phase 29 crack-wave, 2026-07-18)  <sub>L4343</sub>
- **§59** — Three h_seq sweep-residual classes match_one/the-exemplar-bank don't reveal (Phase 29 crack-wave close, 2026-07-18)  <sub>L4377</sub>
- **§60** — Classify the residual, don't rank it: the deterministic residual→class classifier and what it measured about the backlog (Phase 29 Task-13A, 2026-07-21)  <sub>L4407</sub>
- **What** — it measured — the whole open backlog, byte-grounded  <sub>L4431</sub>
- **§3-Two** — corollaries worth remembering  <sub>L4450</sub>
- **§3-The** — parallel-probe race this surfaced  <sub>L4463</sub>
- **§60a** — What the first DIRECTED grinder run exposed (Phase 29 Task-13B, 2026-07-21)  <sub>L4472</sub>
- **§60b** — The plateau autopsy's verdict: a `partial` drift is a WRONG DRAFT, not a missing transform (Phase 29 Task-13B close, 2026-07-21)  <sub>L4510</sub>
- **§61** — Task 14: the gate ladder's missing stage is the ARITY pre-pass, and it is TU-side not draft-side (Phase 29, 2026-07-21)  <sub>L4545</sub>
- **§61a** — The Task-5 wave: 11/12 MATCH, 0 banked — three DISTINCT integration walls, each now named (Phase 29, 2026-07-21)  <sub>L4606</sub>
- **§61b** — The jtbl gate stage: built, and the ORDERING law it exposed (Phase 29 Task-14 stage 4, 2026-07-21)  <sub>L4658</sub>
- **§61c** — The jtbl bank is INCREMENTALLY valid and CLEAN-INVALID (Phase 29, 2026-07-21) — the blocking finding  <sub>L4713</sub>
- **§61d** — The undo was eating the tree: two tools, one defect, invisible to the byte-gate (Phase 29, 2026-07-22)  <sub>L4779</sub>
- **§62** — The jtbl RECONCILE must also follow the carve: the post-carve draft reconcile (`harvest_verify._jtbl_reconcile`, Phase 29 SESSION-11, 2026-07-22)  <sub>L4833</sub>
- **§63** — The fresh-138 DEF-SIDE blocker: fix the HEADER decl, not the draft (`fix_header_decl.py`, Phase 29 SESSION-13, 2026-07-23)  <sub>L4877</sub>
- **§64** — The §20 type-lift's three laws: fold the tagged typedef, check VISIBILITY, and strip only what is TEXTUALLY IDENTICAL (`lift_types.py`, Phase 29 SESSION-14, 2026-07-23)  <sub>L4908</sub>
- **§64a** — VARIANT types: UNIQUIFY the camps, do not reconcile them (`uniquify_type.py`, Phase 29 SESSION-14)  <sub>L4957</sub>
- **§63** — UPDATE (Phase 29 SESSION-14) — `fix_header_decl`'s "SAFE" verdict is FLEET-BLIND; it MUST be R22-validated  <sub>L5000</sub>
- **§65** — The stranded-draft recovery: BLAST-RADIUS TIERS, and the per-overlay de-macroize that refutes §20's DEF-conflict wall (`tools/demacroize.py` + `tools/blocker_probe.py`, Phase 29 SESSION-16, 2026-07-24)  <sub>L5015</sub>
- **§65a** — The blast-radius taxonomy (makes §61's law structural instead of remembered)  <sub>L5030</sub>
- **§65b** — The escape: de-macroize the instantiation, don't touch the shared header  <sub>L5045</sub>
- **§65c** — `rtu_match` MATCH → bank held 13/13 on self-decl, but broke on the FIRST callee-decl case  <sub>L5081</sub>
- **§65d** — Existing-ladder baseline, measured (do this before building a recovery stage)  <sub>L5091</sub>
- **§65e** — Two oracles, and the disagreement is the finding (R34 in practice)  <sub>L5101</sub>
- **§65f** — The de-macroize lever's BOUNDARY, measured: byte-neutral 14 times of 15, and the 15th is why the gate exists  <sub>L5118</sub>
- **§65g** — Where the cheap levers STOP: the local-type and in-TU-self-decl classes did not yield  <sub>L5139</sub>
- **§66** — Exercise a banking driver's SUCCESS path before pointing it at a wave: the free re-bank test (Phase 29 SESSION-17, 2026-07-24)  <sub>L5159</sub>
- **§66a** — The widest write in a pipeline is the one most likely to be UNDECLARED  <sub>L5181</sub>
- **§66b** — A metric parsed out of another tool's prose goes NULL silently when the label changes  <sub>L5200</sub>
- **§66c** — Before a wave, verify the FUEL exists; an "already attempted" set built from the wrong directory lies in BOTH directions  <sub>L5217</sub>
- **§66d** — The permuter⇄reader loop: alternate a random search with a byte-verified idiom, and let `residual_class` decide whose turn it is (Phase 29 SESSION-17)  <sub>L5245</sub>
- **§66d-1** — What transfers between giants is the LOOP, not the PIN  <sub>L5280</sub>
- **§66d-2** — Two operational sharp edges  <sub>L5289</sub>
- **§66d-3** — Read the ILS per-cycle SERIES, not its final best: a repeated score and a still-falling one look identical in a summary line and mean opposite things  <sub>L5300</sub>
- **§67** — The arg-copy PLACEMENT lever: launder a parameter into a fresh pseudo AT the statement where the target's copy lands (Phase 29 SESSION-18, `func_8014D820` 25 → 16)  <sub>L5315</sub>
- **§67a** — Run the symbol-set guard BEFORE you pay for a gate (`tools/symcheck.py`, Phase 29 SESSION-18)  <sub>L5405</sub>
- **§66d-4** — "ILS converged" means converged FOR THAT WEIGHT PROFILE, not a floor (amends §66d-3; Phase 29 SESSION-18)  <sub>L5438</sub>
- **§66d-5** — `residual_class`'s "structural ⇒ permuter CPU is waste" is WRONG for schedule permutations (measured, Phase 29 SESSION-18)  <sub>L5491</sub>
- **§68** — A comment-only line halted the extern scan, and the skip label blamed the type cap (Phase 29 SESSION-18)  <sub>L5524</sub>
- **§69** — How to attack a behemoth: map it, don't draft it (Phase 29 SESSION-18, `func_80183814`, 5,122 ins)  <sub>L5569</sub>
- **§70** — The giv-init base register: walk the PARAMETER, not a copy of it (Phase 29 SESSION-18, `func_801777BC`)  <sub>L5619</sub>
- **§71** — Before mapping a giant, look for an already-matched SIBLING beside it (Phase 29 SESSION-18, `func_8017D960`, 3,338 ins)  <sub>L5660</sub>
- **§72** — A `register __asm__` pin is a PREFERENCE, not a reservation (Phase 29 SESSION-18, `func_8017F510`)  <sub>L5726</sub>
- **§3-The** — honest fix was source-level and cheap  <sub>L5750</sub>
- **§3-Giv** — record order (the §70 family)  <sub>L5760</sub>
- **What** — is left, and what is byte-recorded as SPENT  <sub>L5766</sub>
- **§73** — A def-side self-decl conflict has TWO axes: RETURN (fleet widen, T2) and PARAMS (casts at each use, T0). Diagnose which before paying for the expensive one (Phase 29 SESSION-19, `func_8014F3E8` + `func_8014D4C0`)  <sub>L5783</sub>
- **§74** — Auditing a pinned draft: the §72 hazard is CALLER-SAVED pins spanning a call, and only the disassembly can tell you (Phase 29 SESSION-19, `func_8017D960` b2, 5 pins)  <sub>L5822</sub>
- **§75** — A propagation cap is usually a MINORITY-SPELLING SOURCE OVERLAY, not a wall: census the carried extern before believing the exclusion message (Phase 29 SESSION-19, `func_8014F3E8` ×4 → ×138)  <sub>L5866</sub>
- **§75a** — The exclusion classes, enumerated with named causes (Phase 29 SESSION-19, the 134-binary `dedup_extend` sweep)  <sub>L5925</sub>
- **§75b** — A body's preamble can carry `#define`s, not just `extern`s; extraction lifts only the externs (Phase 29 SESSION-19, `func_80165CA0` ×3 → fleet)  <sub>L5948</sub>
- **§75c** — Class-B's remedy is the FULL §17a-1 PAIR (decl **and** call-site cast); a decl-only fix moves the error and looks like a new wall (Phase 29 SESSION-19, `func_8012F14C`)  <sub>L5997</sub>
- **§76** — The allocno CLASS (local vs global) is the dominant regalloc lever, and C reaches it ONLY through declaration scope and variable reuse (Phase 29 SESSION-19, behemoth #3 `func_8017F510` 1,511 ins, 97 → MATCH, pin-free)  <sub>L6028</sub>
- **§3-The** — mechanism, with citations  <sub>L6037</sub>
- **§3-The** — attribution primitive (use this before calling anything a scheduling residual)  <sub>L6057</sub>
- **§3-Two** — diagnosis traps this function proved  <sub>L6064</sub>
- **Practice** — Practice  <sub>L6074</sub>
- **§77** — Every extraction tool carries a NARROW hard-coded preamble set; anything outside it silently caps the body's reach. Diff the preamble before you gate. (Phase 29 SESSION-19 — three variants in one session, two different tools)  <sub>L6084</sub>
- **§3-The** — CANDIDATE gate and the REAL gate need DIFFERENT preambles — keep the difference out of the bank  <sub>L6124</sub>
- **§78** — A LENGTH drift can be a register grant in disguise; and `fold` never leaves a literal first in an `|` chain (Phase 29 SESSION-19, behemoth #2 `func_8017D960` 3,338 ins, 1806 → 0, pin-free)  <sub>L6174</sub>
- **§3-The** — drift was an allocation decision, not missing code  <sub>L6180</sub>
- **§3-The** — economics  <sub>L6211</sub>
- **§79** — For a 0-callee giant, fingerprint by DATA symbols (§71 cannot fire); and the STACK-SLOT ORDER is a declaration-order oracle (Phase 29 SESSION-19, `func_8017BF14` 4,763 ins, cold start → 45/4763)  <sub>L6220</sub>
- **§71** — has a blind spot, and this is it  <sub>L6226</sub>
- **§3-NEW** — LEVER — the frame layout reads back the original declaration order  <sub>L6240</sub>
- **§76** — confirmed at scale, and a pin nuance  <sub>L6251</sub>
- **§3-The** — residual, and the honest read  <sub>L6260</sub>
- **§80** — A do-not-re-buy entry is scoped to its BASE, not to the function; and the pin's hidden cost is an unconditional `qty_phys_sugg` (Phase 29 SESSION-19, `func_8017BF14` 45 → 0)  <sub>L6270</sub>
- **§3-The** — pin's hidden cost, with the citation  <sub>L6291</sub>
- **§3-The** — flagged "#1 move" LOST — and why the failure is informative  <sub>L6302</sub>
- **§78** — 's attribution primitive, run and reproduced  <sub>L6312</sub>
- **Cold-start** — economics, now complete  <sub>L6318</sub>
- **§81** — Banking a jr (jump-table) function: the 3-step carve chain, and why `match_one` cannot see the problem (Phase 29 SESSION-19, `func_8017C954`)  <sub>L6323</sub>
- **§3-The** — defect this exposed: a shared type that is present but invisible  <sub>L6354</sub>
- **§82** — Two source-shape oracles from behemoth #6: a duplicated `addiu $aN,$sp,K` across a `jal` means the block was INLINED, and scalar-vs-aggregate decides WHEN a stack slot is allocated (Phase 29 SESSION-19, `func_8017C730` 1,061 ins)  <sub>L6370</sub>
- **§3-1.** — The inlined-helper signature  <sub>L6375</sub>
- **§3-2.** — Scalar vs aggregate decides *when* the slot is allocated  <sub>L6389</sub>
- **Also** — reproduced on this function  <sub>L6401</sub>
- **§3-And** — the banking footnote (§75a class A, one line)  <sub>L6405</sub>
- **§83** — The parameterised-repeat law, the spill-area trap, and why a per-case edit cannot move a per-case symptom (Phase 29 SESSION-20, `func_80183814` 5,122 ins, cold-ish start → 36 structural / 99.3%)  <sub>L6412</sub>
- **§83a** — READ THE HEADLINE NUMBER CORRECTLY: a LENGTH drift makes `match_one`'s count meaningless  <sub>L6419</sub>
- **§83b** — THE LEVER: find the parameterised REPEAT before decoding case-by-case  <sub>L6427</sub>
- **§83c** — TRAP: a "dead local" in a prior draft may be gcc's OWN spill area  <sub>L6445</sub>
- **§83d** — CSE's quantity budget is WHOLE-FUNCTION, so a local rewrite cannot fix a local symptom  <sub>L6452</sub>
- **§83e** — §80 vindicated again, on the same day it was written  <sub>L6466</sub>
- **§84** — The DERIVED-OFFSET remap bug: a hand-computed literal that encodes the DISTANCE between two per-overlay symbols, and why `match_one` is structurally blind to it (Phase 29 SESSION-20, `func_8013D53C`)  <sub>L6472</sub>
- **§3-The** — construct  <sub>L6477</sub>
- **§3-Why** — it survived every candidate gate  <sub>L6495</sub>
- **§3-THE** — FIX IS MECHANICAL — the tool already holds the answer  <sub>L6510</sub>
- **Scope** — , measured (do not over-generalise — §80)  <sub>L6518</sub>
- **§3-Two** — ladder lessons banked with it  <sub>L6528</sub>
- **§85** — The RETURN-axis fleet widen is ALL-OR-NOTHING: widening the shared header alone guarantees a conflict in the source overlay (Phase 29 SESSION-20, `func_8012CC88` / `func_8014D12C`)  <sub>L6539</sub>
- **§3-The** — conflict  <sub>L6544</sub>
- **§3-THE** — FAILURE MODE — widening only `engine_core.h` is worse than not starting  <sub>L6552</sub>
- **§3-The** — precondition, and how to check it in one grep  <sub>L6559</sub>
- **§3-Do** — the WHOLE axis in one edit, then R22 once  <sub>L6567</sub>
- **Reading** — , for the next person  <sub>L6576</sub>
- **§86** — Pinned-exemplar templatability is a PER-FAMILY property, not a per-member rate; and the §42e pin guard is now over-conservative (Phase 29 SESSION-20)  <sub>L6589</sub>
- **§3-The** — guard refuses a class that largely works  <sub>L6591</sub>
- **§3-THE** — LAW: all-or-nothing PER FAMILY  <sub>L6600</sub>
- **§3-Why** — the two live families differ from the three dead ones — the open question  <sub>L6627</sub>
- **§87** — `match_one` COMPILES but never LINKS, so an unresolvable data symbol reads as MATCH; and stored drafts go STALE against the tree (Phase 29 SESSION-20)  <sub>L6633</sub>
- **§3-The** — blindness ladder, now complete — FOUR classes `match_one` cannot see  <sub>L6649</sub>
- **Consequence** — for the backlog ledger  <sub>L6659</sub>
- **§3-The** — cheap discriminator, before spending a sweep  <sub>L6666</sub>
- **§88** — `cross_jump` will not merge a common suffix containing a CALL; and §78 is scoped to ORDERED comparisons only (Phase 29 SESSION-20, the behemoth close-out)  <sub>L6674</sub>
- **§88a** — repeated CALL-shaped blocks are left UNMERGED; call-free tails are merged for you  <sub>L6679</sub>
- **§88b** — the `slti` literal-position law (extends §78 to comparisons)  <sub>L6686</sub>
- **§88d** — BANKING ORDER: run the §81 carve chain BEFORE banking, never after  <sub>L6701</sub>
- **§88e** — a wrong diagnosis, refuted properly (the model for how to treat an inherited lever)  <sub>L6708</sub>
- **§88f** — the missing rung: a RELOCATION gate between `match_one` and the binary  <sub>L6718</sub>
- **§89** — Two throughput rules the project already had written down and was not following (Phase 29 SESSION-20)  <sub>L6726</sub>
- **§89a** — MEASURE the write set; do not assert its tier (`tools/blast_radius.py`)  <sub>L6732</sub>
- **§89b** — the parallel gate farm existed; the family path could not reach it (`tools/sweep_parallel.py`)  <sub>L6745</sub>
- **§3-The** — standing sequence  <sub>L6758</sub>
- **§90** — Five tool-integrity laws from one session, each of which changed an answer (Phase 29 SESSION-21, 2026-07-27)  <sub>L6766</sub>
- **§90a** — A comparison tool MUST share its reference oracle's index space, exactly  <sub>L6772</sub>
- **§90b** — "Byte-neutral" is not "wanted": undo on the SUCCESS path too  <sub>L6789</sub>
- **§90c** — A library-callable function must FAIL CLOSED on an unconfigured module  <sub>L6799</sub>
- **§90d** — Do not measure a live wave's drafts (§87 in real time)  <sub>L6810</sub>
- **§90e** — An agent's CONCLUSION and its EVIDENCE fail independently — re-derive the premise, not the fix  <sub>L6818</sub>
- **§91** — A structure-TRANSFER is only valid where the structure corresponds: the `--like` role trap (Phase 29 SESSION-21, `func_8012AAAC` ×137)  <sub>L6842</sub>
- **§3-The** — three-hypothesis trail, because two of them were wrong and the wrongness is instructive  <sub>L6871</sub>
- **§92** — Conforming a DECLARATION: pointer changes are caller-neutral, scalar-WIDTH changes are not (Phase 29 SESSION-21, `tools/conform_decls.py`)  <sub>L6891</sub>
- **§93** — `set -o pipefail` attributes a pipeline failure to the LAST stage, not the failing one (Phase 29 SESSION-21, `func_8014D820`)  <sub>L6922</sub>
- **§94** — A family sweep's `0/N` is a TYPE-CARRY failure until proven otherwise: lift the exemplar's local types, ALL of them, transitively (Phase 29 SESSION-21, `func_8016B6BC` 0/137 → 137/137)  <sub>L6947</sub>
- **§95** — `reconcile_tu` dropped the SIBLING declarators of a multi-symbol `extern` line (Phase 29 SESSION-21, `func_80176218`)  <sub>L6983</sub>
- **§96** — The same rewrite, one shape down: `reconcile_tu` matched statements to lines by TEXT, so every COMMENTED declaration was silently skipped (Phase 29 SESSION-22, `func_80176218` banked)  <sub>L7018</sub>
- **§97** — The gate's own tree hygiene: a refused carve, an unchecked recovery, and a snapshot that captured a dirty tree (Phase 29 SESSION-22)  <sub>L7066</sub>
- **§98** — `conform_decls` had three defects, and only the third needed R22 to find (Phase 29 SESSION-22, `func_8014CF04`)  <sub>L7113</sub>
- **§99** — The narrow-param wall is a DEF-side problem with a ZERO-blast-radius fix: convert the definition to K&R (Phase 29 SESSION-22, `func_80175AB8` + `func_80175DA8`)  <sub>L7167</sub>
- **§3-Two** — `reconcile_tu` bugs found underneath, one introduced while fixing the other  <sub>L7190</sub>
- **§100** — Prefer the DRAFT-LOCAL fix: a type only one function uses belongs in its BODY, not in a shared header (Phase 29 SESSION-22, `func_80175DA8` 0/137 → 137/137)  <sub>L7216</sub>
- **§101** — The STALE DEFAULT class: a guard whose cause was removed is a silent skip wearing a safety label (Phase 29 SESSION-22, three instances in one session)  <sub>L7245</sub>
- **§102** — A PLUMBING verdict can MASK a DIFF; and K&R is not always a codegen change (Phase 29 SESSION-22, `func_8016EC0C`)  <sub>L7272</sub>
- **§103** — A FILE-scope `extern` in a shared overlay TU is a GLOBAL constraint on every LATER function; move the DECL, not the draft (Phase 29 T48/T51, `func_80135260` — the fleet-wide half)  <sub>L7307</sub>
- **§104** — Two silent-skip defects in one scan: match on MASKED text, emit from the ORIGINAL (Phase 29 T53, `gather_externs`)  <sub>L7395</sub>
- **§105** — A gate's revert must survive an EXCEPTION, not just a failure (Phase 29 T53, `jtbl_family_bank`)  <sub>L7431</sub>
- **§106** — Persist the MEASUREMENT, derive the POLICY: a stored route let a stale file out-vote the live table (Phase 29 T54, `residual_class._ROUTE`)  <sub>L7470</sub>
- **§107** — A lever wired into ONE gate path is a lever most families cannot reach (Phase 29 T56, `func_80144090` 0/136 → 136/136)  <sub>L7513</sub>
- **§108** — Diagnosing a family `0/N`: the four causes, and the third opt-in lever (Phase 29 T59)  <sub>L7551</sub>
- **§109** — Conforming a definition to a shared header: fix the NAMES, then check the RETURN precondition (Phase 29 T60)  <sub>L7600</sub>
- **§110** — A unit must define exactly ONE function, and "ends in `;`" does not tell you which line defines it (Phase 29 T65)  <sub>L7640</sub>
- **§111** — The distinct-code metric is not noisy: a family pays it only if its members are byte-VARIANT (Phase 29 T66)  <sub>L7681</sub>
- **§112** — A macro-scoped declaration only collides where the macro is INSTANTIATED (Phase 29 T67/T69, `audit_header_sigs.py`)  <sub>L7730</sub>
- **§113** — An ARITY blocker only exists if the macro CALLS the function; an address-taken use has no call site (Phase 29 T72)  <sub>L7776</sub>
- **§114** — The THIRD decl axis: a CALLEE the draft declares differently from the target TU (Phase 29 T76/T77)  <sub>L7808</sub>
- **§115** — A `func_XXXXXXXX` predicate rots by design: the same name-form assumption in THREE places (Phase 29 T78)  <sub>L7850</sub>
- **§116** — Optimization level is a property of the FILE, not the function: read a family 0/N against the member's stub HOME (Phase 29 T79)  <sub>L7877</sub>
- **§3-The** — fix moves the DEFINITION, not the stub — and here is why the obvious shortcut fails  <sub>L7895</sub>
- **§117** — Spell the sibling's symbol from the SIBLING's address, not the exemplar's kind (Phase 29 T82)  <sub>L7926</sub>
- **§3-Why** — it survived so long: a MASKED oracle will MATCH a wrong symbol  <sub>L7953</sub>
- **§118** — Ordinal (positional) immediate resolution: compare C tokens to the DIFFERING asm uses (Phase 29 T87)  <sub>L7963</sub>
- **§119** — Two levers on the SAME axis, opposite directions: test the off-diagonal (Phase 29 T89)  <sub>L7997</sub>
- **§120** — Uniquify draft-defined TYPE names; and check which of N staging sites you actually patched (Phase 29 T93)  <sub>L8026</sub>
- **§3-Do** — NOT "strip the duplicate typedef" — it breaks the extern that uses it  <sub>L8037</sub>
- **§3-The** — wiring trap that cost two attempts  <sub>L8043</sub>
- **§121** — Synthesise externs for macro-DEFINED callees from the macro's own definition head (Phase 29 T95)  <sub>L8061</sub>
- **§122** — GATE RAW BEFORE TRANSFORMING; the undo belongs to the WRITER, as a per-edit journal (P30 T0a, 2026-07-30)  <sub>L8084</sub>
- **§123** — PROPAGATE A FAMILY WITH THE TOOL ITS TIER NEEDS: `dedup_propagate` is h_exact-only; its refusals are statements about the TOOL (P30 wave 1, 2026-07-30)  <sub>L8116</sub>
- **§124** — A "not matched" verdict can mean the definition is there under a DIFFERENT C NAME: the asm-label alias blind spot (P30 SESSION-28, `func_8016191C` ×137)  <sub>L8153</sub>
- **§124a** — a family sweep's `0 matched-exemplar families` may be a FILTER, not a wall  <sub>L8198</sub>
- **§125** — Split the CARVE from the BODY before calling a jr residue a wall — and measure it by SHA from a CLEAN tree (P30 SESSION-28; **this section's first draft was WRONG and the method caught it**)  <sub>L8206</sub>
- **§3-The** — method (keep this)  <sub>L8213</sub>
- **§3-The** — instrument rules that make its answer trustworthy (this is where I failed)  <sub>L8223</sub>
- **§3-The** — corrected results (each SHA-verified, from a clean tree, restore re-verified)  <sub>L8238</sub>
- **§3-Two** — further notes worth keeping  <sub>L8248</sub>
- **§3-The** — meta-lesson  <sub>L8257</sub>
- **§126** — The carve-within-a-carve: an ADDRESS RANGE is not an OPTIMIZATION REGION (P30 T2, byte-proven end-to-end)  <sub>L8264</sub>
- **§3-The** — finding: opt level is per FILE, so the file's contents must be opt-HOMOGENEOUS  <sub>L8279</sub>
- **§3-The** — instrument trap that hid it (and it is §124's shape again)  <sub>L8290</sub>
- **§3-The** — mechanics  <sub>L8299</sub>
- **§126a** — a bare `except: continue` around a coverage-asserting oracle re-creates the silent skip (P30 S28)  <sub>L8317</sub>
- **§127** — The `-O0` regime: the CONSTANT-OFFSET FOLD, and why `-O0` needs its own idiom set (P30 T3 wave, 15 targets)  <sub>L8349</sub>
- **§3-The** — idiom they kept re-deriving: the constant-offset fold  <sub>L8356</sub>
- **§3-The** — rest of the `-O0` regime (write PLAIN C, and mean it)  <sub>L8367</sub>
- **§127a** — §71 (sibling-first) is the strongest `-O0` lever, and it beats the index  <sub>L8377</sub>
- **§127b** — the knowledge was in a SOURCE COMMENT, not the cookbook  <sub>L8386</sub>
- **§128** — A raw NUL in C source makes grep SILENTLY SKIP the file (P30 S28, 137 files)  <sub>L8393</sub>
- **§128a** — a negative control must corrupt a SCRATCH COPY, never the tracked file  <sub>L8425</sub>
- **§129** — Post-carve, `rtu_match`/`match_one` COUNT THE JUMP TABLE AS INSTRUCTIONS; and a carve must never be committed without its owner (P30 S28, `func_8013BD74`)  <sub>L8437</sub>
- **§129a** — the target instruction count is INFLATED after a carve  <sub>L8441</sub>
- **§129b** — never commit a carve whose owner is still a stub (it strands the carve)  <sub>L8462</sub>
- **§3-The** — real blocker underneath, for the record  <sub>L8477</sub>
- **§130** — An INCREMENTAL build can report BYTE-IDENTICAL for a change the CLEAN build cannot even LINK (P30 S28, the jr pair)  <sub>L8486</sub>
- **§3-The** — diagnostic ladder that finally located it (reusable)  <sub>L8523</sub>
- **§131** — The jtbl OVER-SPAN: `sltiu N` is ground truth in BOTH directions, and the zero-word rule only guards one (P30 S28, `func_80191C50`)  <sub>L8533</sub>
- **§132** — The `JR-PAIR-IN-ONE-O0-OBJECT` "wall" was TWO instrument defects: a merged-double span the carve could not see, and a truncated object no rule deleted (P30 S29, `func_8013B83C` + `func_8013BD74`)  <sub>L8570</sub>
- **Defect** — 1 — a pre-§8e MERGED DOUBLE is not a single-table predecessor  <sub>L8579</sub>
- **Defect** — 2 — `as` writes a corpse and nothing deletes it  <sub>L8597</sub>
- **§3-The** — fingerprint, and the 30-second ladder that found it  <sub>L8607</sub>
- **§3-The** — transferable rule  <sub>L8628</sub>
- **§132a** — `--like` is for a sibling with NO record; against one that HAS a record it over-derives (P30 S29, `ov_SC07_010`)  <sub>L8635</sub>
- **§132b** — When the span's already-matched owner is ITSELF multi-switch: `--span-rel` (P30 S1, `func_8014032C`)  <sub>L8660</sub>
- **§133** — The DEFAULT-FILTER class: three times in one session, a tool silently answered a narrower question than the one asked (P30 S1–S3)  <sub>L8685</sub>
- **§134** — MULTI-LINE BLINDNESS: one root cause, four faces, in `family_remap`'s preamble scanner (P30 S6, 190 zero-crack families)  <sub>L8711</sub>
- **§135** — Six byte-verified gcc-2.7.2 idioms from the P30 S6f-h waves (and the two-lane wave shape that found them)  <sub>L8765</sub>
- **§3-The** — codegen idioms  <sub>L8770</sub>
- **§3-The** — integration idioms (these decide whether a byte-correct draft BANKS)  <sub>L8804</sub>
- **§3-The** — wave shape that produced these  <sub>L8819</sub>
- **§136** — The LOCAL-VARIABLE lever: how many C locals, at what scope (P30 wave 4a, 25 byte-verified banks)  <sub>L8837</sub>
- **§3-The** — splitting/merging rules (each closed a residual, byte-gated)  <sub>L8853</sub>
- **§3-The** — type-form rules  <sub>L8879</sub>
- **§3-The** — scheduling rules (refining §135-2 and §135-4)  <sub>L8909</sub>
- **§3-The** — declaration surface (integration, not codegen)  <sub>L8938</sub>
- **Wave** — economics (measured, for the next batch's sizing)  <sub>L8950</sub>
- **§136a** — Blocker capture: classify on the OUTPUT, never on the exit status  <sub>L8973</sub>
- **§136b** — A prior wave's "genuine byte-DIFF" verdict is NOT reliable evidence (4 of 4 refuted)  <sub>L9030</sub>
- **§136c** — SIBLING-FIRST is a DERIVATION shortcut, not just a conflict fix (the fastest route in a family wave)  <sub>L9066</sub>
- **§136d** — Four gcc-2.7.2 levers the redraft lane found (each closed a residual no other lever moved)  <sub>L9091</sub>
- **§136e** — §136c's PRECONDITION, and two more symptom keys (wave 4b batch 3)  <sub>L9135</sub>
- **§136f** — Two declaration sub-cases the reconcile lane surfaced (lane now 15/15 lifetime)  <sub>L9183</sub>
- **§136g** — When the index points at the WRONG lever: two byte-refuted routings (func_801863B4)  <sub>L9208</sub>
- **§136h** — CORRECTION: the zero-crack pool does NOT "refill with cheap work" (my error, byte-measured)  <sub>L9237</sub>
- **§136i** — The drafter model LADDER: Haiku → **Sonnet** → Opus → Fable5 (Drew, 2026-08-03)  <sub>L9267</sub>
- **§136j** — The failure MIX flips with function size (measured across four bands, one session)  <sub>L9299</sub>
- **§137** — REGALLOC-PERM is a TWO-COMPILE ARITHMETIC PROBLEM, not a permuter job  <sub>L9341</sub>
- **§137a** — A gate verdict has a TIMESTAMP; re-check it against the draft's mtime  <sub>L9383</sub>
- **§138** — The propagation lanes: a gate refusal is a DECLARATION, and which lever you owe depends on blast radius  <sub>L9412</sub>
- **§3-The** — triage, cheapest first  <sub>L9418</sub>
- **§3-The** — DEFINITION-side alias is the only escape when the fleet canon disagrees on a promoting param  <sub>L9470</sub>
- **Rank** — the lane by measured concentration, not by class count  <sub>L9478</sub>
- **THREE** — carry variants hide in one "CARRY-FIXABLE" bucket — and they need different fixes  <sub>L9487</sub>
- **§134** — again, in a second tool — and the waiter rule corrected  <sub>L9523</sub>
- **STEP** — 0 of sibling-first: grep `src/` for a distinctive LITERAL from the `.s`  <sub>L9543</sub>
- **Reconciling** — a gate-refused draft: which way you edit depends on WHERE the TU's decl is  <sub>L9562</sub>
- **§139** — A GATE THAT GREPS FOR VERDICTS MUST ASSERT 1:1 ACCOUNTING; and a `--src` filter must not survive a carve (P30 S38, wave 6: 10 of 16 drafts vanished)  <sub>L9591</sub>
- **§3-The** — generalisation — three corollaries worth more than the bug  <sub>L9623</sub>
- **§3-And** — the inverse-lookup trap, same session  <sub>L9640</sub>
- **§140** — A METRIC IS NOT A MEASUREMENT UNTIL IT IS REPRODUCIBLE FROM THE COMMITTED TREE (P30 S1e: a phantom regression that gated the session's best lever)  <sub>L9657</sub>
- **§3-The** — three-line proof (do this before diagnosing any metric movement)  <sub>L9677</sub>
- **§3-The** — two instrument defects it exposed  <sub>L9686</sub>
- **§3-The** — same swallow, twice more, in the integration spine  <sub>L9706</sub>
- **§3-Two** — wrong mechanisms I chased first, and why they were wrong  <sub>L9717</sub>
- **§141** — The §134 class is CLOSED: every line-shape decision now routes through `cdecl._mask` (P30 S39)  <sub>L9737</sub>
- **§142** — An open stub whose `h_exact` class is MATCHED elsewhere is FREE. Propagate the body; do not gate a draft. (P30 S39, +7,710 ins in two commands)  <sub>L9775</sub>
- **§3-The** — measurement (do this before any wave; it is ~20 lines and needs no builds)  <sub>L9787</sub>
- **§3-The** — trap that hid it — SAME FUNCTION, TWO ROUTES, ONLY ONE IS FREE  <sub>L9802</sub>
- **Route** — selection (why `--addr` sometimes says "nothing changed")  <sub>L9811</sub>
- **§3-And** — the report-vs-bytes lesson attached to it  <sub>L9819</sub>
- **§143** — `cast_call_sites` read a RETURN STATEMENT as a prototype and deleted it. A 0/39 sweep became 18/39. (P30 S40)  <sub>L9831</sub>
- **§144** — THE LITERAL'S SPELLING PICKS THE IMMEDIATE ENCODING (P30 S40 wave 1, `func_801822E0`)  <sub>L9885</sub>
- **§145** — Three loop/combine levers from the S40 wave-2 drafters (16/16 match_one)  <sub>L9924</sub>
- **§146** — RE-MEASURE A WALL BEFORE YOU RESPECT IT. Both "permanent" giants fell to drafts already on disk. (P30 S6, +50,094 ins)  <sub>L9974</sub>
- **§3-Why** — a correct draft can read as an intrinsic wall  <sub>L9985</sub>
- **Then** — propagation returned 0/137 TWICE — both times a missing TYPE  <sub>L9997</sub>
- **§3-Two** — errors of mine, both instructive  <sub>L10006</sub>
- **§3-The** — rule  <sub>L10020</sub>
- **§147** — The three-stratum FRAME LAW, and four "stop searching" verdicts (P30 S42, `func_8017C294`, serial run)  <sub>L10033</sub>
- **§3-A.** — The frame has THREE strata, and stratum 3 is unreachable from C  <sub>L10038</sub>
- **§3-B.** — A `?:` on MEMORY operands costs ~16 bytes of invisible frame; on REGISTER operands, zero  <sub>L10054</sub>
- **§3-C.** — Inner-block declaration does NOT delay slot allocation — BYTE-REFUTED  <sub>L10065</sub>
- **§3-D.** — A lone `$t8`/`$t9` in the target is RELOAD SCRATCH — reproduce the spill, don't pin the register  <sub>L10070</sub>
- **§3-E.** — A `qty_compare` TIE is not spelling-reachable — recognise it and stop  <sub>L10080</sub>
- **Consequence** — for the family (a real scheduling decision)  <sub>L10092</sub>
- **§148** — The loop.c hoisting THRESHOLD is arithmetic you can compute, and the `?:` clamp that folds to MIN_EXPR (P30 S42, `func_8017C6F4`, 947 ins)  <sub>L10142</sub>
- **§3-A.** — `move_movables` hoists iff `threshold × savings × lifetime ≥ insn_count` — and you can read it  <sub>L10148</sub>
- **§3-B.** — `(v < 0x40) ? v : 0x3F` is folded to `MIN_EXPR` and expands to the WRONG SHAPE  <sub>L10167</sub>
- **§3-C.** — A zero-byte ALLOCNO-PRIORITY slider  <sub>L10181</sub>
- **§3-D.** — Reproduce the original's BUGS verbatim  <sub>L10191</sub>
- **§149** — Four instrument defects in one session, and the two questions they were hiding (P30 S43)  <sub>L10240</sub>
- **§3-A.** — A prep step that returns its input on failure is indistinguishable from a search that found nothing  <sub>L10246</sub>
- **§3-B.** — Same address + same name ≠ same body — and the ledger keys on address  <sub>L10264</sub>
- **§3-C.** — `make: *** [...] Error N` is a summary, never a diagnosis  <sub>L10280</sub>
- **§3-D.** — "Cheap fuel" that was never probed: 0 of 31 templatable  <sub>L10290</sub>
- **§150** — A register ROTATION across symmetric blocks is VARIABLE-IDENTITY evidence, not an allocator tie (P30 S43, `func_8017C6F4`, 947 ins ×4)  <sub>L10309</sub>
- **§3-The** — fix  <sub>L10316</sub>
- **§3-The** — method that found it (this is the transferable part)  <sub>L10326</sub>
- **§3-Two** — corrections to the record  <sub>L10342</sub>
- **Diagnostic** — order (adopt this)  <sub>L10353</sub>
- **§151** — THE GHOST WEDGE: when a load-before-store transposition is unreachable by ANY statement order (P30 S43, `func_8017EF68`, 969 ins)  <sub>L10364</sub>
- **§3-The** — mechanism (read from cc1's own `-dR` trace, not inferred)  <sub>L10369</sub>
- **§3-The** — lever — a zero-emission insn that absorbs the blocked tick  <sub>L10384</sub>
- **§3-The** — two fallouts, and how to close them (both measured, in order)  <sub>L10392</sub>
- **When** — to reach for it  <sub>L10404</sub>
- **§152** — BYTE SIZE is the family key that name- and h_seq-grouping both miss (P30 S43, the 0xECC family: 1 crack → 12 overlays → 11,364 ins)  <sub>L10415</sub>
- **§3-The** — finding  <sub>L10420</sub>
- **§3-The** — key  <sub>L10429</sub>
- **§3-Two** — cautions that must travel with this technique  <sub>L10440</sub>
- **§3-The** — companion defect (open)  <sub>L10451</sub>
- **§153** — THE ADDRESS-REMATERIALISATION LAUNDER: a third zero-emission asm lever (P30 S43, `func_8018D98C`, 710 ins)  <sub>L10463</sub>
- **Symptom** — Symptom  <sub>L10471</sub>
- **Mechanism** — (gcc source + RTL dumps, not inferred)  <sub>L10476</sub>
- **What** — does NOT work (14 byte-measured probes)  <sub>L10483</sub>
- **§3-The** — cure — a fresh launder per site, each in its own block  <sub>L10489</sub>
- **Companion** — levers from the same function  <sub>L10499</sub>
- **§154** — Reading a disc payload: the module-id word, static base derivation, and "type 1 = uncompressed overlay" (P30 S44)  <sub>L10519</sub>
- **§3-A.** — Payload word0 is a global MODULE ID; code starts after the header  <sub>L10524</sub>
- **§3-B.** — Two static base-derivation methods that must AGREE (use both)  <sub>L10533</sub>
- **§3-C.** — PAC type 1 = the same payload class as type 4, just NOT compressed  <sub>L10545</sub>
- **§155** — hi/lo literal scanning MUST track base registers (S45)  <sub>L10555</sub>
- **§155a** — the same failure class, one level up: SHAPE-blind table scanning (S45 p5)  <sub>L10564</sub>
- **§155b** — check the TYPE your oracle returns before comparing against it (S45 p5)  <sub>L10584</sub>
- **§155c** — the ZERO-REFERENCE trap: gcc splits a global-array address across the `lui` and the LOAD (S46)  <sub>L10607</sub>
- **§156** — an ORPHANED reconcile poisons the fleet: `dedup_propagate`'s kept edit (S45 p6/p7)  <sub>L10637</sub>
- **§157** — the cheap-tier size cliff, measured (S45 p6)  <sub>L10684</sub>
- **§158** — The RANGE-EXTENDER: a fourth zero-emission asm lever completes the allocno toolkit (P30 S46 tier-3, `func_8017CE58`, 733 ins)  <sub>L10709</sub>
- **Symptom** — Symptom  <sub>L10718</sub>
- **§3-Why** — the fork is chained (gcc source, validated insn-by-insn against -dS/-dR dumps)  <sub>L10724</sub>
- **§3-The** — method (dump-arithmetic first, then place — no probing)  <sub>L10737</sub>
- **Bonus** — facts worth keeping  <sub>L10754</sub>
- **§156** — THE PREFERENCE-DONOR MERGE: cross-region variable reuse is what fills a0-a3, and a call-arg use in ONE region steers the fill in ALL of them (P30 S46 tier-3, `func_80186E24`, 611 ins: 236-off "S11 regalloc-order" → MATCH, zero new pins)  <sub>L10769</sub>
- **§159** — THE DECLARATION AXIS: conform to byte-truth, and make every guard state its COVERAGE (P30 S47; ~10,930 sites across 8 axes, fleet byte-identical)  <sub>L10825</sub>
- **§160** — THE REACH-15 WAVE HARVEST: an align-1 block move, and the instrument that called it a wall (P30 S47)  <sub>L10890</sub>
- **§161** — THE RETRY-WAVE HARVEST: a jump table indexed from zero, and two allocator traps (P30 S47)  <sub>L10966</sub>
- **§162** — S48 WAVE-1 HARVEST (P30, 2026-08-11): the reach-ordered sibling campaign's first 12 targets  <sub>L11011</sub>
- **§162a** — SHARPENS  *(sharpens §161a, §131, §8a-pad, §129a)*  <sub>L11038</sub>
- **§162b** — SHARPENS  *(sharpens §48-A3, §156, §150, §76)*  <sub>L11067</sub>
- **§162d** — SHARPENS  *(sharpens §31, §21, §30, §55a)*  <sub>L11096</sub>
- **§162e** — NEW  <sub>L11138</sub>
- **§162** — THE LICM PAIR: what makes an address a movable AT ALL, and why the preheader order is the body order (P30 S47, `ov_MAIN_012`)  <sub>L11140</sub>
- **§162f** — SHARPENS  *(sharpens §42d, §41d, §73, §10)*  <sub>L11206</sub>
- **§162g** — NEW  <sub>L11261</sub>
- **§162** — CROSS-JUMP DIRECTION: the surviving copy is always the LATER one, so a BACKWARD `j` into a sibling arm is a source `goto` (P30 S48)  <sub>L11263</sub>
- **§162h** — SHARPENS  *(sharpens §88, §88a, §50-B, §8)*  <sub>L11297</sub>
- **§162** — The cross-jump "CALL veto" is a COUNT law, not a CALL law (BOUNDS §88a; P30 S48, `func_80189540`)  <sub>L11299</sub>
- **§162i** — SHARPENS  *(sharpens §135, §21, §42, §32)*  <sub>L11365</sub>
- **§162j** — SHARPENS  *(sharpens §25, §136d-1, §48-B, §46-L2)*  <sub>L11390</sub>
- **§162k** — SHARPENS  *(sharpens §1-I2, §12, §160d, §21)*  <sub>L11419</sub>
- **§162l** — SHARPENS  *(sharpens §48-B, §48-C1, §20, §21)*  <sub>L11482</sub>
- **§162m** — SHARPENS  *(sharpens §36, §158, §148, §153)*  <sub>L11537</sub>
- **§158a** — THE FIFTH LEVER IS NOT AN ASM: `do { } while (0)` is a REGION ref-multiplier you MINT (P30 S48, `func_8017CBC8`, ov_MAIN_012 / jr_801789AC, 188 ins → MATCH)  <sub>L11539</sub>
- **§3-The** — law  <sub>L11546</sub>
- **Size** — it before you write it (§158 step 1-2, applied)  <sub>L11557</sub>
- **§3-The** — wrap BOUNDARY is the dial — and it is indiscriminate  <sub>L11578</sub>
- **§3-Not** — a pure dial  <sub>L11584</sub>
- **DIAGNOSTIC** — TELL — two faces, one law  <sub>L11588</sub>
- **§162n** — NEW  <sub>L11610</sub>
- **§162o** — SHARPENS  *(sharpens §158, §136-1, §136-6, §79)*  <sub>L11643</sub>
- **§162p** — SHARPENS  *(sharpens §48-B, §46-L2, §156, §136d-1)*  <sub>L11702</sub>
- **§162q** — SHARPENS  *(sharpens §30, §30a, §135-2, §136-13)*  <sub>L11719</sub>
- **§163** — S48 WAVES 2-3 HARVEST (P30, 2026-08-11/12): the five that were byte-probed and are actionable  <sub>L11757</sub>
- **§163z** — THE UNVETTED REMAINDER (do not cite as law; each needs a dedupe pass)  <sub>L11825</sub>
- **§164** — S48 §163z SKEPTIC PASS (P30, 2026-08-12): 190 claims vetted, 82 banked  <sub>L11845</sub>
- **§16Xy** — SHARPENS  *(sharpens §136d-3, §37 /s-DEP LATTICE, §135-2, §136-13, §162q)*  <sub>L12297</sub>
- **§3-The** — `/s` drop clause is in ALL THREE dependence predicates, so the FIXED-ADDRESS STORE is what floats (P30 S48, `func_8017CA18`, ov_MAIN_012)  <sub>L12299</sub>
- **§164z** — REFUTED CLAIMS: do NOT re-derive these  <sub>L13620</sub>
- **§165** — S48 WAVE-4 HARVEST (P30, 2026-08-12): banked the same day the wave landed  <sub>L13686</sub>
- **§16Z** — SHARPENS  *(sharpens §37 "the /s-DEP LATTICE", §136-13, §136-14, §16Xy, §162q, `gcc-2.7.2-map/sched.md` §64)*  <sub>L14337</sub>
- **§3-The** — ADDRESS-CLASS TABLE: which load/store pairs even REACH the `/s` clause (P30 S48 wave 4, `func_80185B44`, ov_SC03_014)  <sub>L14339</sub>
- **§165z** — REFUTED THIS WAVE: do NOT re-derive  <sub>L14855</sub>
- **§166** — THE DESTINATION-TU ORACLE (P30 S48): the seven-attempt bug that was never codegen  <sub>L14899</sub>
- **§167** — S48 WAVE-5/6 HARVEST (P30, 2026-08-12): the saturation point  <sub>L14955</sub>
- **§167z** — REFUTED IN WAVES 5/6: do NOT re-derive  <sub>L16151</sub>
- **§168** — THE COUSIN TIER (P30 S49, 2026-08-12): h_seq's exact-hash brittleness, measured — and the similarity map above it  <sub>L16208</sub>
- **§169** — THE MICRO-ADAPT LANE (P30 S49): edit a proven body, don't crack a new one  <sub>L16260</sub>
- **§170** — THE A-PROP WORD-DIFF CARD (P30 S49): the lane that had no owner  <sub>L16307</sub>
- **§171** — THE STALE SEED SYMBOL (P30 S50, 2026-08-13): why §170's 91%→57% was never codegen  <sub>L16351</sub>