Every stage passes. On all four boots POST is 990/990, with one
dict_hash (0x627f392ccbaa96f1). test-32, the proofs, the images and the
docs pass. The docs stage ran in the background at Captain Bob's ruling.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Every stage passes. On all four boots POST is 990/990, with one
dict_hash (0x627f392ccbaa96f1). aarch64 is inside its 570 s; before
mechanism A it stopped in POST. test-32, the proofs, the images and the
docs pass.
The docs stage runs longer than one foreground command; it ran in the
background at Captain Bob's ruling.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
The fabric keeps its occupied places as a list in ascending order, and
the step's nine passes walk that list in the same order as before. A
place removed mid-walk keeps its link, so the walk goes on to the next
node.
Test first: test_fabric_walk.c, 13 checks.
Every existing fabric and mesh test passes with the same checks, and so
do test-32 and ASan/UBSan. The host mesh boot runs POST 990/990. Over
five alternating runs its median falls from 11.695 s to 7.999 s.
Logs are in logs/20261117-mechanism-a.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Captain Bob's "Find the TCG cost first", continued. aarch64 still ran out of its 570 s at 281972ab
(logs/20261114-133155, STOPPED); amd64 passed in 253 s (logs/20261114-132732).
- (LOOKUP): the count-and-first-character key in register B while the entries are walked, and the
xt kept on the stack until an entry passes it -- about 19 opcodes for an entry that cannot be the
word, from 28. The answer is the same.
- POST: a suite begins with one line, DECIMAL FORTH DEFINITIONS FORGET (POST); its mark is made by
the state line of its first case run (a skipped kernel case passes it on).
mesh-boot: 15.7 s -> 12.2 s (37.7 s before this work), POST 990 of 990. make -C v4 test: all passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Ruled (Captain Bob, 2026-11-14): "Find the TCG cost first". v3's whole POST took the amd64
accept from about 3 m 15 s to 8 m 45 s, and aarch64 past its 570 s (logs/20261114-091651,
STOPPED, committed). Profiled on the host (gprof, mesh-boot): 87% of the time in POST's
lines, about 280,000 opcodes a line -- the interpreter's dictionary lookup.
- POST puts the dictionary back before each of v3's suites, as v3's run_test_suite does
("to remove test-created words"): a mark, (POST), forgotten and made again
(V4_POST_SUITE). Only the dictionary: SCR, BASE and the rest stay as the cases left them.
- (LOOKUP) tries each entry by its count and first character first -- one fetch of the
name's first cell, masked -- and only then its hidden flag and the whole name. The answer
is the same; an entry that cannot be it costs a few opcodes, not some forty.
mesh-boot: 37.7 s -> 15.7 s, POST 990 of 990. make -C v4 test: all v4 tests passed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
capsules/v4/physics.4th, blocks 8202-8204: ENTROPY@ ENTROPY! RESET-ENTROPY as kernel requests -35..-37 (a
word's heat by the address ' leaves, a data word's too; an error for what is not a word, as v3's);
TOP-WORDS and WORD-ENTROPY in StarForth, printing in v3's form what the kernel gives them through (HOT#)
(HOT) (XT-HEAT), -38..-40 (the i-th hottest in v3's stable order); (- in StarForth, skipping to the
matching close paren with nesting, as v3's code does. Tests first: test_mesh_boot.c 101 checks, 110 on, 98
without opcode physics. Listing: 403 live words. amd64 PASS, word_count=403. Not pushed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Captain Bob, 2026-11-14: "v3" ... "we can always change it late". v3 counts an execution per token its
(C) interpreter takes and per word its definitions run. Measured on 8 lines, counting every call on v4 gave
~700 to v3's one for typed lines (the F18 interpreter's own calls) but 1:1 in compiled code, so no single
HEARTBEAT_CHECK_FREQUENCY could match v3. Now INTERPRET counts its tokens in (TOKENS) -- one F18 store, the
map gains the cell -- and the kernel adds them at each call; a call counts, and gets physics, only from
dictionary space or as INTERPRET's or EXECUTE's dispatch (the image gains their addresses). The knob stays
at v3's 256. test_inference.c out of the per-width tests (its own rule). Tests first: test_mesh_boot.c 90,
99 on, 87 without opcode physics; v4 test all pass. amd64 PASS, 181 s, dict_hash 0x5c53b2e18343a9b6.
Not pushed (held to the end of Phase 7).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
capsules/v4/physics.4th, a kernel capsule on Hera (blocks 8200-8201): FREEZE-WORD UNFREEZE-WORD FROZEN?
HEAT! HEAT@ DECAY-RATE@ Q.VARIANCE INFER-DECAY-SLOPE INFER-WINDOW-WIDTH INFER-RUN INFER-WINDOW@ INFER-DECAY@
INFER-VARIANCE@ INFER-FIT@ INFER-EARLY-EXIT@ -- each a KERNEL-WORD, requests -21..-34 (v4/mesh.h), served by
the kernel on any node: "The compudynamics is stadium/fabric" (Captain Bob, 2026-11-14). A word is named as
v3's, the newest visible entry of CONTEXT then FORTH (the image gains CONTEXT's address); the arithmetic is
v4/system/inference.c, v3's to the bit. Q48.16 results are v4's doubles (low cell, then high). HEAT! starts
the word's decay from now. Test first with v3's POST cases in v4's address units: test_mesh_boot.c 86
checks, 95 on, 83 without opcode physics. docs/reference/v4-dictionary.tsv: the 15 words (392 live).
amd64 PASS, word_count=392. Not pushed (held to the end of Phase 7).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Captain Bob, 2026-11-14: "we're supposed to be working on f18 opcode as well as our composed f18
dictionary"; ruled every loop on both, "and an easy revert if opcodes are too granular".
- The engine: a hook at retire (exec.h on_retire), as Part A's at call; NULL unless set. test_exec.c.
- The mesh: 32 opcode records per node, an opcode window, opcode transitions under L4; at every retired
instruction decay at the opcodes' own slope, heat, temperature, the window, the transition; in each cycle
all 32 cooled, the opcode window's service, inference on its heat. L8 scores from the words and its gates
govern both; the tick stays paced by word calls. The window is now one type for words and opcodes.
- The revert: OPCODE_PHYSICS (Kconfig.physics, default y; the kernel's flags) -- 0 and the opcodes are
counted and charged to the Stadium as in Phase 1. make -C v4 mesh-boot-noop builds and tests it.
- Tests first: test_mesh_boot.c 77 checks as built, 86 with pipelining on, 74 with the revert.
- Measured on amd64, back to back: 219 s on, 196 s off (+12% under TCG). Both boots' logs labelled.
Not pushed (held to the end of Phase 7).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
v4/system/inference.c: v3's inference_engine.c and its Q48.16 arithmetic ported to the bit, no C library --
the variance, the window width (Levene), the decay slope (log-linear regression), the run with its 5%
early exit. v4/tests/test_inference.c links v3's own and compares them: POST's fixed cases, 20000 random
multiply/log pairs, 400 random trajectories (make -C v4 inference; in make -C v4 test). In each node's
cycle every 1000 ticks (HEARTBEAT_INFERENCE_FREQUENCY), once the window is warm: the trajectory from the
window's words' heat, the width and slope applied unless it exits early, then Loop #7 -- the heartrate
slower when stable, faster when not, within 1/4..4x; Hera's re-arms the kernel's heartbeat
(heartbeat_set_adaptive_period_ns), as v3's Hera. L8 reads its outcome. Test first: test_mesh_boot.c 72
checks, 80 on. amd64 PASS, 163 s. Not pushed (held to the end of Phase 7).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
v3's vm_heartbeat_update_l8 with its table (ssm_jacquard.c, compudynamics.c), rewritten in integers: the
regime from the window's entropy, the prediction CV and the locality of the last 32 calls; the current card
scored (ANOVA from inference, joint convergence under L5/L6), decayed 0.99, the next chosen by UCB1 over 64
cards; its gates L1-L7 applied to the loops, L1 and L4 within the build's ceilings. Every gate off and card 8
at the start, as v3's kernel boot. Applied through card_bits: v3 applies the card's number, one bit off --
filed #196, ruled "v4 correct, v3 as is". v4_mesh_l8_force as v3's ssm_l8_force_config; the exact decay
and window tests hold a card. Test first: test_mesh_boot.c 67 checks, 75 on. amd64 PASS, 156 s.
Not pushed (held to the end of Phase 7).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Commit b880a6ea (Task 2a) checked out detached and booted on amd64: 175 s, as slow as Task 2b's 174 s and
181 s, against 2a's 146 s an hour before. 2b has no measurable cost; on the host the mesh boot was no
slower either. Not pushed (held to the Phase 7 boundary).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Rulings, Captain Bob 2026-11-13: "Build a hot-words cache"; then, v3 being found to have both off
(ENABLE_PIPELINING, ENABLE_HOTWORDS_CACHE: n in Kconfig.physics and kernel/Makefile), "Built, off as v3".
- The kernel, at every call (mesh.c): v3's hit check, the transition, the likeliest next word (kept as each
count rises; v3 scans 1024 a call -- the same word but in a tie v3's Q48.16 rounding makes), speculation
at v3's thresholds into the node's cache; a word past heat 50 promoted at its call (v3 promotes at a
lookup; v4's heat is the kernel's). Counters from the owner's heap, given back when a word goes.
- The node (dict.f18): (LOOKUP) tries (HOT)'s xts first, in F18, calling only C@; the kernel keeps it true
(entries that go, or whose name is defined again, taken out). Empty unless the build has the cache.
- Tests: test_mesh_boot.c as built (57 checks) and with both on (mesh-boot-on, 65). v4 test: all pass.
- Reported, v3's own: when full, its promote takes the next slot without looking for the word there (a few
hot words churn all 32 slots); its prefetch misses are never counted. Both reproduced.
- amd64 boots, RESULT: PASS, but 174 s and 181 s against 2a's 146 s, while on the host the mesh boot is no
slower: being measured. Not pushed (held to the Phase 7 boundary).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
As v3's physics_pre_execute, in its order (Captain Bob, 2026-11-13: "Every call, as v3"): a word's heat
decays by the ticks since its last touch at the node's decay slope (Q48.16, v3's arithmetic), unless
frozen or L8 has Loop #3 off; +1; its temperature smoothed 3:1; its ID into the node's Rolling Window of
Truth (4096, warm at 1024, a diversity check due every 512). The kernel's heartbeat is read at each call.
Pipelining and the hot-words cache ruled "Built, off as v3" (Task 2b). Test first: test_mesh_boot.c, 56
checks. amd64 boots, RESULT: PASS, 146 s against Phase 2's 142 s. Not pushed (held).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
All eight stages passed: the four boots (POST 539/539, word_count=377, dict_hash=0xa854ba15b437d288 on all
three architectures, Hera's words admitted to the Stadium after POST on each), test-32 (39 programs, 0
failures), the proofs, the images, the documents. logs/20261113-191247/PROTOCOL.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
ENGINE step 4 Phase 2 (plan docs/v4.0.0/plans/2026-11-13-engine-step4-phase2-word-accounts.md). Rulings,
Captain Bob 2026-11-13: "ID field in the header"; "Per heartbeat tick"; the ACL "All of it in Phase 4".
- The word ID, at xt-4 before the flags (ENGINE.md 3b). The system's words are numbered as they are
assembled, by a new directive, #id, (and #ids, the count), in the C assembler (text.c) and in the
machine ROM's (capsules/v4/rom), which counts in the system's own (NEXT-ID). (HEADER) numbers later
words; FORGET gives back the IDs it frees, always the newest; COLD resets to (BOOT-ID); (SEAL) records
it. All 313 system headers carry the cell.
- The kernel's records (v4/mesh.h): (TELL-DEFINED) and (TELL-FORGOTTEN), requests -19/-20 on port 0,
which the kernel puts in (WORD-DEFINED)/(WORD-FORGOTTEN) when it seals a node, and reads the system's
own words once. A bit per address marks a word's xt (calls into the ROM or headerless code are not
words); at call, the bit test and a count. v4_mesh_charge hands the owner the calls since.
- The Stadium: stadium_words_charge -- N calls as N of stadium_word_dispatch's dispatches, shared by calls
when the reservoir is short; sk_v4.c charges every node's words at the tick beside its opcodes, and
stadium_word_forget before an ID is used again.
- Tests first: test_text.c, test_host_dict.c, test_host_quit.c (numbering, FORGET, COLD),
test_mesh_boot.c (44 checks), kernel/tests/test_stadium_words.c; qemu_session.py: Hera's words
admitted after POST.
- amd64: Stadium words: promotions=141 after POST, CONSERVED; POST 539/539; word_count=377;
dict_hash 0xa854ba15b437d288 (the new cell); the boot 142 s against Phase 1's 153 s.
- v4 test, test-32: all pass; v4-checks: 80 programs under the sanitizers, 0 failures; the hosted
builds alike. docs/reference/v4-dictionary.tsv: the two new words; FABRIC-3 XLVIII.24.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
All eight stages passed: the four boots (POST 539/539, word_count=375, dict_hash=0xbbf72b1f752a1a51 on
all three architectures, the edge cases on amd64), test-32 (39 programs, 0 failures), the proofs, the
images, the documents. logs/20261113-170011/PROTOCOL.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
ENGINE step 4 Phase 1, Tasks 5b-5d, and the edge cases (Captain Bob, 2026-11-13: "Build it to proper
spec"; "Check for the edge cases"; "Yes, add it to v4-checks").
- v4/system/mesh.c, v4/include/v4/mesh.h: the nucleus assembled by the ROM and given to Hera as F18
cells; Hera born empty at place 0, the kernel on port 0, the console on port 1; told she is 10, in the
registry, the kernel's words, FORTH-79, sealed, POST, the kernel's capsules. NODE-BORN / NODE-KILL in
the owner's room.
- No scheduler: v4_mesh_tell returns once the node told says how its text ended, and
v4_mesh_step / v4_mesh_busy let the host step the fabric while it waits, so nodes run between
lines as during them. A text left "waiting" is owed its end: the next line ends it "Interrupted"
and is then run (MESH.md 7b.6).
- kernel: sk_v4.c boots through the mesh; its idle steps the fabric while a node can run, and halts
only when none can (#151 kept). stadium_return_quota and stadium_ops_forget on a kill.
- Tests: v4/tests/test_mesh_boot.c (32 checks: the boot, births, kills, rebirth, a live number, a
fault, no such node, Hera not killing herself, no room, a line too long, a node running for ever,
a wait broken by the next line), run by `make -C v4 test` and under ASan+UBSan by `sanitize`
(v4-checks); kernel/tests quota return; scripts/qemu_session.py births() and the edges on bare
metal.
- Docs: ENGINE.md 6 "who runs next" struck (settled: no scheduler); the Phase 1 plan; FABRIC-3
XLVIII.23; .claude/memory/rules.md: alpha.5 closed to new work.
- Evidence: v4-checks 80 programs under the sanitizers, 0 failures; the hosted builds alike; two
stopped v4-checks runs labelled STOPPED; three development amd64 boots labelled.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
#189 Phase 1 Task 5a: Hera in the kernel's fabric and registry. On the
four boots: NODE-ME . gives 0 ok, the physics lines as before, POST
539/539; word_count 375 and dict_hash 0x06531687cc075aee on all three
architectures (centre.4th and hera.4th loaded); test-32 39 programs 0
failures; proofs; images; documents, the dictionary covering all 375
live words. Every boot under timeout 570. logs/20261113-142432/PROTOCOL.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
#189, ENGINE step 4 Phase 1, Task 5a. Captain Bob, 2026-11-13: "Full
births now"; "Hera as now, in a fabric".
- sk_v4.c: the kernel's fabric (v4/fabric.h), Hera's node at its one
place, entered in its registry (v4/manage.h) as node 10. The registry's
answers are real: a birth is "no room" -- the fabric has one place
until 5b; CAPSULE-OPEN/-LINE read the kernel's capsule directory,
checked as the boot checks (no capsule of F18 code is there until 5b);
a born node's parity line. She is not told her number: the boot's
console addresses node 0 (told 10, the boot stopped after POST -- kept,
labelled, logs/20261113-132712).
- boot.h/boot.c: v4_boot's `request` hook (the registry's requests) and
`capsules` (the kernel's own, loaded before the seal, so POST runs on
them). hosted.c sets neither: the hosted product's boot is as it was.
- kernel/Makefile: manage.c in the kernel's v4 engine.
- qemu_session.py: a v4 acceptance boot asks NODE-ME . and wants 0.
- v4-dictionary.tsv: the 55 words of centre.4th and hera.4th, written
from their source and manage.h; mkdictionary.py loads the kernel's
capsules into the hosted system before WORDS, so the live list is the
kernel's -- 375 words.
- .claude/memory/rules.md, rule 6: every boot command under timeout 570.
Seen to fail first (logs/20261113-132410): NODE-ME was an unknown word.
After: amd64 NODE-ME . gives 0 ok; POST 539/539; word_count 375. v4
suite: test, test-32 99 programs 0 failures; v4-checks 79 under the
sanitizers 0 failures, the hosted builds alike.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
#189 Phase 1, Tasks 1-4: v4 in the kernel's boot. On the four boots: the
Stadium made ("Stadium: 83886 cells"), POST 539/539, "Stadium opcodes: 30
resident, hottest opcode 8", "Stadium conservation: CONSERVED", the quota
self-test and the heartbeat's ticks -- the new physics check passing;
word_count 320 and dict_hash 0x47570d4ced6ce255 on all three
architectures, unchanged; test-32 39 programs 0 failures; proofs; images;
documents. logs/20261113-125011/PROTOCOL.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
#189, ENGINE step 4, Phase 1 (plan 2026-11-13-engine-step4-phase1), Tasks
1-4. Captain Bob, 2026-11-13: Q1 "A new node key", Q2 "Per heartbeat
tick", Q3 "All of v3's", Q4 "In Phase 1" / "Full births now", who runs
next "Hera by compudynamics now" (Task 5, to come).
- stadium.c/.h: stadium_node_key(node) ("NODE" | number),
stadium_boot_init_for(first), stadium_birth_node(key, name);
stadium_boot_init() and stadium_birth_hera() are them for Hera's key, so
v3's path is as it was (only an impossible panic's words changed).
- kernel_main.c: the system is chosen and said first, then the Stadium,
sessions, switch-signal and kernel-Hermes tables for both systems in
v3's order; v4: Hera, node 10, patron zero under her node key, the word,
block and opcode layers. The quota self-test and the heartbeat's line
are functions now, called by both systems.
- sk_v4.c/.h: sk_v4_run() goes -- sk_v4_boot() returns, the kernel's
checks after POST (opcodes, words, conservation, quota, heartbeat), then
sk_v4_prompt(). sk_v4_charge_opcodes() at the heartbeat's service points
(the key wait, after each line) and once after POST.
- stadium_ops.c/.h: a node's 32 opcodes as Stadium patrons (HEAT-OP's
feed, "Both"): retirements charged as the word layer's dispatches, a
quantum each, the reservoir's floor kept, shared by retirements when
short; cooling per tick as words. ~0.5 us a charge on the host: kept.
- qemu_session.py: a v4 acceptance boot fails without "Stadium: N cells",
"Stadium conservation: CONSERVED" and the heartbeat's ticks (#189).
Tests first (logs/20261113-122142): the physics check failed on the v4
boot; test_stadium_node, test_stadium_ops (and its share case, after a
real boot showed ";" taking all the heat) failed before. Now: kernel host
tests all pass; amd64 v4 boot passes, POST 539/539, dict_hash unchanged,
"Stadium opcodes: 30 resident, hottest opcode 8", CONSERVED; a system=v3
boot prints its Stadium lines as on 2026-11-06.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
A2 (#189): every word but the ten called. POST 539/539 on the four
boots; word_count 320 and dict_hash 0x47570d4ced6ce255 on all three
architectures; test-32 39 programs 0 failures; proofs; images; documents.
Boot times (stage, with its build) against be456d44's run: amd64 123 s ->
118 s, aarch64 208 -> 197, riscv64 96 -> 93 -- no slower under QEMU.
logs/20261113-053019/PROTOCOL.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Plan 2026-10-20 (ENGINE step 4), Task A2, for #189. Captain Bob,
2026-11-13: "The ten"; "Accept it".
- compile.f18 (COMPILE,): a word is copied in line only when it is flagged
in-line and compile-only (24) -- the ten that work on the return stack,
>R R> R@ I J LEAVE and the later UNLOOP 2>R 2R> 2R@. Every other word,
in-line ones too, is a call the hook at `call` sees.
- EXECUTE calls the xt: `call xt` (opcode 3, slot 0) and a ; in (EXEC),
run. INTERPRET runs a word by (GO): `call xt` and `jump (IL)` in
(RUNX), entered by a jump, at the old return-stack depth.
- host_map.h: (EXEC) BUF0_W-22, (RUNX) BUF0_W-115, from what was free.
- Its cost, accepted: one return entry at the bottom of a chain (30 deep
from the prompt, not 31) and one per explicit EXECUTE while its word
runs; the hosted boot +2.4% (2.917 s to 2.987 s), instructions +0.4%.
- Tests first (logs/20261113-045122/a2): inline_words_are_called,
residue_stays_inline, execute_passes_the_hook, rstack_kept_across_call,
abandoned_definition_not_called -- 3 of 1096 failed before. The depth
checks moved to the new figures; SEE shows the calls; two layout checks
use in-line test words; 2>R 2R@ 2R> checked in test_host_quit.
- DECOMPOSITION.md EXECUTE, the plan's A2 and rulings, FABRIC-3 XLVIII.22.
Suite: test, test-32 99 programs 0 failures; v4-checks 79 under the
sanitizers 0 failures, the hosted builds alike; hosted POST 539/539.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
#190 (POST regenerated, 539 cases) and #16 (v4-checks nightly). POST
539/539 on the four boots, at the new floor; word_count 320 and dict_hash
0xcac4cf58ed493172 on all three architectures; test-32 39 programs 0
failures; proofs; images; documents, the dictionary covering all 320
live words. logs/20261113-031936/PROTOCOL.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Captain Bob, 2026-11-13: "Create the issues then fix them before the long
acceptance test runs"; #16: "Nightly".
#190 -- make -C v4 post79 did not reproduce POST:
- mkpost.py runs v3 with address randomisation off (setarch -R): its
printed addresses (>IN, FIND) are the same each run, so POST79.md is.
- v4/Makefile post79 builds hosted v3 from this tree's source first and
uses it, not the lfs/ copy, which can be older.
- post_cases.c and POST79.md regenerated: 539 cases -- LOAD.rest_of_line,
which v3 gets through since it changed after 2026-10-07. Two runs the
same. POST_FLOOR 538 -> 539 (#173).
- test_host_unit.c takes the table's count, not 538 typed in the test.
#16 -- scripts/nightly.sh v4-checks: make -C v4 sanitize hosted-check,
logged under logs/<TS>/v4-checks/, refused without qemu-user; nightly.yml
runs it after publishing, before the proofs, installing qemu-user if the
image lacks it.
Evidence (logs/20261113-014927/post79, -024314): before, regenerated with
nothing changed: 539 cases and addresses differing run to run; after: the
same twice, hosted POST 539/539; test, test-32 99 programs 0 failures;
v4-checks 79 programs under the sanitizers 0 failures, the hosted builds
alike. The first verification failed at test_host_unit.c:442 (kept); a
stray v4-checks run stopped at once is kept, labelled (-014453).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
A1 (the hook at call), the merge of master (the six runner.capacity
comments) and HEAT-OP/ANTICLOCK (#189). POST 538/538 on the four boots;
word_count 320 and dict_hash 0xd148437fb39a1afd on all three
architectures (318 and 0x8c92508f748f99cf before: the two new words);
test-32 39 programs 0 failures; proofs; images; documents, the
dictionary covering all 320 live words. logs/20261113-005100/PROTOCOL.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
#189 -- Captain Bob, 2026-11-12: "Add HEAT-OP. If it's a huge performance
hit, we'll revert it. If not ... it stands"; a full 33-cell array, as
DECOMPOSITION.md section 7 draws it; the Stadium feed is Phase 1's.
- node.h/node.c: v4_node_heat_regs_attach; a data fetch from base+k gives
opcode k's count, base+32 the anti-clock (the low cell at 32 bits); a
store changes nothing. Reset detaches them.
- host_map.h: HEAT_REG at word 13824, in what was free between the
dictionary space and SEND_TEXT; the compile-time check covers it.
Nothing moved. map-*.f18: HEAT-OP 13824, ANTICLOCK 13856.
- image.h/image.c/mkmap.c: the image carries the address; v4_image_born
attaches the node's own counts.
- system.f18: HEAT-OP@ ( op -- n ), op mod 32; ANTICLOCK@ ( -- n ).
word_count 318 -> 320 (hosted 317 -> 319): dict_hash changes.
- v4-dictionary.tsv: both words; DECOMPOSITION.md 7 records the build.
Also in this batch:
- Makefile: make init installs qemu-user -- hosted-check needs it
(Captain Bob: "Add to make init now").
- .claude/memory/rules.md, rule 5: stop and ask when something comes up
(Captain Bob: "you are supposed to stop and ask").
Tests first (logs/20261113-005047/heat-op): test_node did not compile;
test_host_quit 3 of 1337 failed. Suite: 199 programs, 7448554 checks, 0
failures; hosted-check, the three alike. Timing: no measurable cost.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Plan 2026-10-20 (ENGINE step 4), Task A1, for #189 -- Captain Bob,
2026-11-12: the physics blocks alpha.5; order A, 1, 2, then 7.
- exec.h/exec.c: v4_call_hook, es->on_call/on_call_ctx; called at `call`
once the return address is pushed, before P moves. NULL: as before.
- heat.h/heat.c: the per-call-target table (4096 slots, no capsule word
counted), its freeze mask and v4_heat_call_* go (ENGINE.md step 3: "The
per-call-target count goes"); the kernel will keep a word's accounts.
- tests: test_call_hook_sees_every_call, seen to fail first (no on_call);
test_heat's per-call tests go with the table.
- the plan records today's rulings (order; HEAT-OP: both, a full array).
Suite (logs/20261112-225951/v4-tests): test, test-32, sanitize, 0
failures; hosted-check, the three alike. amd64 boot: POST 538/538,
dict_hash 0x8c92508f748f99cf, unchanged -- nothing sets the hook yet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
#188 (the Jacquard selector) and 7e0728e4 (a release refuses an unsigned
build). POST 538/538 on all four boots, dict_hash 0x8c92508f748f99cf on the
three, test-32 39 programs 0 failures, proofs, images, documents.
logs/20261112-210531/PROTOCOL.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Captain Bob, 2026-11-12: "Wire L2, keep L7 on". The table held the 128
cards of the 7-bit L1-L7 space; nothing read L7, and the rolling window
recorded whatever the card said, so the 128 were 32 behaviours.
- compudynamics.c: cd_tuning_word() sizes the table to 64 (L1-L6).
- ssm_jacquard.c: card_bits() makes a card's 7-bit mask, L7 set, where a
card is read, seeded and started (C4: 17 -> card 8);
ssm_l8_force_config() still takes the 7-bit index, its L7 bit implied.
- physics_execution_hooks.c, kernel vm_core.c: the rolling window records
only while L2 is on (or before there is a config).
- ssm_jacquard.h says what is wired, bit by bit (it said L7 was);
v3/CLAUDE.md and the L8 reference corrected in place; FABRIC-3 XLVIII.21.
Tests first, seen to fail against HEAD (logs/20261112-210510/kernel-tests):
test_ssm_jacquard 4 of 7, test_physics_l2 1 of 4; all pass now. v4 reads
no rolling window: acceptance is unchanged; the protocol runs next.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Captain Bob, 2026-11-12: "release MUST refuse unsigned builds". release.yml's
"The capsule-signing key" step warned when CAPSULE_SIGN_KEY was missing and the
release went on, unsigned. Now it is an ::error:: and the step fails, as the
gate into master's does (#174). build.yml and nightly.yml still warn.
scripts/test_acceptance_signing.sh now takes the step from both acceptance.yml
and release.yml as CI runs them. Against release.yml as it was: the release's
unsigned case passed, no error -- 2 failures (acceptance's 3 checks passed).
Now 6 checks, all pass. logs/20261112-205446/.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
logs/20261112-185800/PROTOCOL.md, for #142, #173, #174 and the notes: the four
boots 538/538, word_count=318, dict_hash 0x8c92508f748f99cf on all three; the
new post-count stage (#173) on the real boots: all four at or above the floor,
all passed; test-32 39 programs, 1,869,223 checks, 0 failures; the proofs
(40 s); the images; the documents. The tree clean before and after.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Captain Bob, 2026-11-12: "Yes, reject unsigned builds". acceptance.yml's
"The capsule-signing key" step warned when CAPSULE_SIGN_KEY was missing and
the PR's acceptance went on, unsigned. Now it is an ::error:: and the step
fails, so the gate does too. A local or protocol build may still be unsigned
(kernel/Makefile, v4/Makefile): the refusal is the gate's. build.yml,
nightly.yml and release.yml still warn -- release.yml is Captain Bob's to rule.
scripts/test_acceptance_signing.sh takes the step from acceptance.yml as CI
runs it and runs it with and without a key. Against the workflow as it was:
the unsigned case passed, no error -- 2 failures. Now: no key fails with an
::error::, a key is handed to the build as SIGN_KEY. logs/20261112-185739/.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
4738aee4's log has a placeholder where the run against the code before #173
belongs ("see the session") -- not evidence. Rerun against
scripts/nightly.sh as it was (4738aee4~1): every one of the 7 cases fails,
the stage not existing. post-count-before-rerun.log; the first log is kept.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Captain Bob, 2026-11-12: "Yes". Acceptance checked POST: PASSED, so a POST
that lost cases still passed. Now scripts/nightly.sh post-count, which the
stick-high stage runs after dict-hash: each of the four boots must report
PARITY:V4_POST with every test passed and at least the floor's tests --
v4/system/POST_FLOOR, 538 today, raised when POST grows and never lowered
without Captain Bob. A boot with no POST line, not run, or no floor file is
refused, not skipped.
scripts/test_nightly_post_count.sh, written first and seen to fail (the stage
not existing), then 7 cases: 538 and 600 pass; fewer than the floor, a
failure, no POST line, a boot not run, no floor -- refused. Run on the boots
of the last protocol (d7b2f6bf): all four 538 of 538. logs/20261112-185620/.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
Captain Bob, 2026-11-12: "correct deficiencies". heartbeat_capture_tick_snapshot
wrote cache_hits_delta and bucket_hits_delta as 0 ("TODO: wire ..."), though
physics_hotwords_cache.c counts both. Now each is read from the VM's hot-words
cache as a delta since the previous tick, the way word_executions_delta is
(and after the cache's stats are reset, the count since then); a VM with no
cache has 0. word_executions_delta was measured already: doe_log.c's "TODO: 0"
for it was stale, and its three column notes are corrected.
kernel/tests/test_heartbeat_export.c, 4 checks: a VM in memory with a cache
whose counters the test sets, captured tick by tick. Against the code as it
was: 3 failures (the hits 0 where the cache counted 5 and 3, then 7 and 7).
Now: 0. Built for amd64, aarch64, riscv64 and the hosted v3. Shared v3 code,
under the ruling. v3 is not part of acceptance; no DoE has run since 2026-09-16.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
logs/20261112-182015/PROTOCOL.md, for the kernel host-test chunk: the harness
and #170, #171, #176, #177, #178, #182. The four boots 538/538 -- a data stick
enumerating through the moved USB descriptor walk, genesis through the identity
code's new seed handling -- word_count=318, dict_hash 0x8c92508f748f99cf on all
three; test-32 39 programs, 1,869,223 checks, 0 failures; the proofs (37 s);
the images, no race; the documents. The tree clean before and after.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
sk_v4_genesis never wiped its seed or its read-back; sk_v4_drive_arrived
never wiped the proxy's seed after her login; sk_v4_real_zuse_arrived wiped
the seed only on a match. (The logged-in Zuse's own key stays in g_zuse_seed,
by design.)
Option 1: the three use one buffer the identity module owns, g_seeds (seed,
then its read-back), and sk_v4_drive_arrived -- which every drive goes through,
and which alone calls the other two -- runs the drive's handling (now
handle_drive()) and then wipes it, whichever way it went.
sk_v4_identity_seeds() lets the host tests read it.
kernel/tests/test_sk_v4_identity.c, 10 checks, through the real drive-arrival
path: a fake xHCI controller marks a slot attach-pending, the identity code's
idle beat handles it, the real gpt.c reads the stick's table from the RAM disk;
the block layer, signatures, key mint, marker and console faked. Genesis; a
genesis not read back; the proxy logging in; a real Zuse named, then her stick
logging in. Against the code before the wipe: 5 failures (every case's seeds
left). Now: 0. The harness gains fake_disk_bind. Built for amd64, aarch64,
riscv64. #176 done with 2bac2ac8 (the mint).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
sk_v4_mint_zuse wiped seed and rseed only when the real Zuse was minted; its
refusals and every failure after YES left them in stack memory.
Option 1 (Captain Bob, 2026-11-12: "1 is great"): the seeds are in a 64-byte
buffer the caller owns. sk_v4_mint_zuse_with(seeds) runs the mint (now a
private mint()) and then wipes the buffer, whichever way it returned -- one
place, every exit; sk_v4_mint_zuse() gives it a buffer of its own.
kernel/tests/test_sk_v4_mint.c, 17 checks: the real gpt.c on the RAM disk, the
identity code, key mint, signature checks, marker and keyboard faked; each of
the mint's eight exits run with the buffer filled with 0xEE first -- no stick,
under 64 MiB, not YES, the key mint failing, the read-back not hers, no
marker, the marker write failing, and minted. Against the mint as it was: 7
failures (every exit but success). Now: 0. Both runs in logs/20261112-181719/. Built for
amd64, aarch64, riscv64. The identity code's two (sk_v4_genesis,
sk_v4_real_zuse_arrived) follow.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
sk_v4_grow.c's dev_rw4k computes 1 KiB block numbers as (uint32_t)devblock * 4,
which wraps past 4 TiB; Artemis is capped at 256 GiB (STICK_ARTEMIS_CAP_GIB,
kernel/Makefile), so it is latent -- but nothing said so where the cast is. Now
a compile-time check there (the C99 negative-array pattern the tree uses): a
cap past 4 TiB stops the build, and a cap not defined at all is an #error, not
a skipped check. gpt.c's casts are bounded by the disk's own size, itself a
32-bit count of 1 KiB blocks, checked before each: each now says so.
kernel/tests' check_cap, seen to fail first (a 4096 GiB cap built); now 3
checks, 0 failures -- 256 GiB builds, 4096 GiB and none refused. Built for
amd64, aarch64, riscv64. The plan's increments 2 and 3 marked.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
sk_v4_idle allocated a controller's slot array again whenever max_slots + 1
was more than it held -- the old array dropped, neither copied nor freed --
while the identity code holds pointers into it (g_offered, g_spare,
g_pending_real, g_real_drive, g_zuse_dev). Its two slot loops then ran to
max_slots whatever the array held.
sk_v4_slots_get() (kernel/src/v4/sk_v4_slots.c): allocated once, zeroed, at the
size first asked for; never replaced; the slots that may be used are those it
has. sk_v4_idle uses it, and its loops stop at the array's end.
kernel/tests/test_sk_v4_slots.c, kmalloc_aligned faked, seen to fail first
against the allocation as it was (4 failures: replaced when more were asked
for, 17 said usable where 9 existed, a held pointer left behind, an allocation
for 0); now 6 checks, 0 failures. Built for amd64, aarch64, riscv64; latent
today (max_slots does not grow), so the boots show no change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm
sk_v4_run's three stops -- "not started", "did not let go of POST's
blocks", "stopped on a fault" -- ended in for (;;) { }: a core spinning at
100%, the thing #151 removed from the prompt. Each now calls sk_v4_stop()
(kernel/src/v4/sk_v4_stop.c): interrupts off, then arch_halt() for ever. Each
still says why first. sk_v4_stop.c added to the loader's and the kernel's v4
sources.
kernel/tests/test_sk_v4_stop.c, with arch_halt and arch_disable_interrupts
faked (the halt jumps back out after three calls; an alarm fails a spin):
against the stop as it was -- "spun for 2 s without halting"; now -- 2
checks, 0 failures: halts in a loop, interrupts off before the first halt.
Built for amd64, aarch64 and riscv64. #177's "seen on the three
architectures": no boot reaches these stops, so it is shown on the host only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017RzZgrf7V7Zk2iczBXDRbm