Nishi Family › Compare › Image Generation

Nishi Compare · capabilities, evidence and delivery

Image Generation

Compare capabilities and delivery evidence. Source discovery, qualification, deployment and accepted outcomes are separate; an open contract names work still to verify.

The sovereign Elder AI gen stack (nishifamily.com/gen: a NAS daemon dispatching to a 5080 worker, content-addressed gallery, companion identity and judges) measured against the local-UI field -- ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus -- and Midjourney. The June-2026 look is the bar this board works back to: the trigger-first slot order landed 2026-08-27, the grammar port and the negative-capable model path are the next rungs, and every gap is a watched contract.

Overview

The purpose, declared position and evidence coverage of this domain. Source presence and completed acceptance are different measures.

Where we are. Measured and working on 2026-08-27: the sovereign gen chain end to end (NAS gateway and orchestrator, OpenAI-shape dispatch to the 5080 worker, async batch, img2img, chat with photo-on-request, content-addressed gallery ingest with GENREC provenance, three sovereign judges, native GPU residency guard and self-healing keepers). The June-2026 look was traced to its record (seed 1480332350, 14 steps, 832x1088, the old Director's Shot-Sheet prompt) and its dominant missing term -- the realism trigger prefix in the leading slot -- was restored as data and shipped through ONE shared composer that the GPU-free probe now proves. Still open: the full prompt-grammar port, the negative-capable model path (the measured controllability wall), the batch board, ControlNet, inpaint, upscale, model acquisition, the LoRA training loop, the graph executor, and the external referee run. The counts sit below this line, measured at emit.

Where we need to go. Best of breed against ComfyUI and the local field on what they do (control, workflow, models) while keeping what none of them have (companion identity, pre-spend QC, judges, provenance-named assets, sovereign ELF control plane) -- and the Elder AI studio back to batch-to-gallery with the June realism, proven by rulers not by eye.

The unit. 1 u = one measured session-leg. Calibration: G1 (trigger-first slot order, shared composer, probe re-based, daemon rebuilt on the NAS toolchain) went from record-mining to deployed in one leg on 2026-08-27.

Cost to get the June look back: 3 u. G1 landed; G2 the grammar port (1,640-row vocabulary as planes plus the slot-ordered composer) and G3 the negative-capable model path are what remain -- G3 is the wall every clothed render hits.

Cost to studio parity with the local UIs: 5.5 u. Batch board, inpaint mask, upscale, weighting and the hires pass; each rides the engine surfaces already served (A1111 img2img shape, ESRGAN in-engine).

Cost to a grade-1 progress claim: 2 u. G5 runs HPSv2, PickScore and GenEval on our renders and prints our row beside the papers; until then every realism claim here is grade 2 (banked rulers) at best.

Research bar. Z-Image-Turbo model card is measured on steps and latency envelope of the base model. Theirs: 8 NFEs, sub-second on an H800, fits 16 GB consumer VRAM [zimage]. Ours: measured through the live daemon on the 5080: 8 steps 12.0 s and 30 steps 25.0 s at 512x512 (gen_steps.conf, 2026-08-04); a cold offload render 167 s on 2026-08-27 with the card shared with a game.

Research bar. HPSv2 human preference benchmark is measured on human-preference score of generated images on the HPD v2 prompt set (798,090 choices over 433,760 pairs) [@hpsv2]. Theirs: the paper's own leaderboard, to be read from its tables when G5 runs. Ours: not yet benched -- G5 runs the published protocol and prints our row.

Research bar. PickScore / Pick-a-Pic is measured on CLIP-based preference scorer trained on real users' choices, recommended for model evaluation and ranking [@pickscore]. Theirs: the paper's own model rankings. Ours: not yet benched -- G5; also the candidate ranker beside nx_gen_rank for generate-N-keep-best.

Research bar. GenEval compositional benchmark is measured on object co-occurrence, position, count and color adherence via detectors [@geneval]. Theirs: the paper's open-model table. Ours: not yet benched -- G5; prompt adherence is the axis the companion lane cares least about and the studio lane most.

Operator investment 2026-09-10: reproducible image recipes: REQUESTED / UNFINISHED. Extend existing G2 composition and GENREC provenance into user-configurable recipes binding identity, pose, camera, garment and lighting to model/version, seed, inputs and outputs. Recipe reuse must retain uncertainty and source permissions; no provider training permission inferred. Score actual results against retained aesthetic references and user decisions, not prompt completion. Beach remains primary; gameengine GE61 records accepted results and review cost.

Latest recorded release

2026-10-08 · G19

album clock fixed at the root: go_album_append now stamps sys_now_realtime_sec(); daemon rebuilt over /api/build (sha a5638247, 549,248 B, contentdiff lost 6 runs all shared-library messages, 44 source paths gained), pre-change binary banked by hash at /volume1/ai/gen/banked_genorch_pre_album_c41a6886.elf, deployed via gdeploy@genorch (the first deploy promoted a no-op because the row stages beside the live path -- copied the stage beside live and deployed again; live hash re-read a5638247). The three historical rows were recovered exactly from the monotonic offsets (uptime 70.5 days exceeds every row, so no reboot intervened): elara_album.tsv rewritten 234 -> 243 B with the original banked as elara_album.tsv.pre-clockfix-20261008

Release history and work log

Release entries describe recorded changes; they do not establish that every capability passed evaluation.

44 of 62 capability source symbols found|15 of them authored stronger-claims|18 open|coverage 709/1000|adoption 32 full / 12 partial

Evidence profile — what the gaps on this board actually are

Measured by nx_swcompare_evidence, read back by nx_evprofile_lib. Every figure is a count with its denominator — there is deliberately no score, no grade and no percentage anywhere in this band, because a stored scalar is a field a seat can edit and a counted partition is not.

proven able to fail counts the gates that have a RECORDED RED — nx_gate_bite mutated the gate subject, rebuilt it, watched the gate go red, and that record is inside the shared TTL. never bitten is its complement over the same denominator: those gates ran and were green, and nothing has ever shown them able to detect anything, so their green is a statement about this run and not about the gate. green at 0/0 is a separate and much weaker observation — the gate printed GREEN on a zero denominator, so its own tooth counter says it examined nothing. A gate can be green, non-zero, and still never bitten; that is the common case and it is now visible instead of implied.

partition: grounded + unsupported = 43 vs present 43 · named + unnamed + withheld = 18 vs open 18 · both reconcile

liar-kill conj=GPQN · all four conjuncts held

graded document: BUILDROOT tree, 48763 bytes · gates map: PRIMARY · stamped 0d 13h ago · source ../knowledge/status/evstamp_gen.verdict

Gap class	What it is, and the work it names
NONE	No gap class fires on this board: every published claim is grounded, the declared gates ran and were green, no watch contract is unnamed or already landed, and the stamp is inside its TTL. That is a statement about evidence, not about depth, scale or polish.

This band reports the referee counts. The per-axis worklist rows — which axis is unsupported, which watch is ready to flip, which gap is unnamed — are printed by nx_swcompare_evidence gen itself and are not carried on the stamp, so this page names the classes and the producer names the rows. That split is stated rather than hidden: a count without a worklist is not actionable, and this band is honest about which half of that it is.

Production map

Follow the dependencies, declared acceptance criteria and recorded priorities. Inspect source binding before treating a rank as executable work.

Ranking source binding: PLAN_MATRIX_BOUND_ONLY. Recorded priorities require current acceptance evidence and resource checks before execution.

Ranking matches the captured plan and matrix only. Latest execution outcome, research freshness, accepted delivery and investment return are unverified.

Recorded priority estimates

Order from nx_compare_rank (nx_dr_ocm: (deficit + cost-of-delay + option + enables) x sponsor x self-sufficiency x momentum / cost). FINISH rows are rungs whose symbol is present but whose organ is short of full adoption: listed before new work by this heuristic. Priority is not measured delivery cost or execution readiness. Stamp: # asof=1791486933 domain=gen target_version=0.1 rungs=20 done=9 open=11 finish=3 ranker=nx_dr_ocm

#	Stage	Rung	Priority	Derivation
F	FINISH	Model acquisition (G9) mg_fetch dossier	REGISTERED-DARK	callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)
F	FINISH	The /gen page is emitted, never handcrafted (G19) gu_emit dossier	REGISTERED-DARK	callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)
F	FINISH	Broken-parts census of /gen as the browser sees it, on a beat (G28) gh_album_years dossier	LIB-UNCENSUSED	a library absent from the import-graph census (younger than the last eco-graph build, or unreachable from buildroot/runtime)
#1	0.1	Live unconditional branch, selected by intent (G3) gw_guidance_select dossier	4650	v=31 m=3 c=20
#2	0.1	Prompt-grammar port (nx_photo_grammar) (G2) pg_compose dossier	1000	v=10 m=2 c=20
#3	later	Upscale on render (G8) gw_upscale dossier	5000	v=25 m=1 c=5
#4	later	Hires and refiner pass (G14) gw_hires_pass dossier	4000	v=20 m=1 c=5
#5	later	Batch board (G4) go_batch_list dossier	1300	v=13 m=1 c=10
#6	later	Workflow graph executor (G11) gen_graph_exec dossier	1000	v=30 m=1 c=30
#7	later	Model swap with TTL (G12) mx_swap_ttl dossier	700	v=7 m=1 c=10
#8	later	Workflow ingest and reproduce (G15) wi_reproduce dossier	700	v=7 m=1 c=10
#9	later	ControlNet and pose conditioning (G6) gw_controlnet dossier	500	v=10 m=1 c=20
#10	later	External referee run (G5) gb_bench_run dossier	100	v=2 m=1 c=20
#11	later	Identity LoRA training loop (G10) lt_train_deploy dossier	100	v=1 m=2 c=20

Declared roadmap — contract, acceptance, executor, effort

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Trigger-first slot order and ONE composer (G1)	go_load_trigger	elara_trigger.txt leads the prompt on both photo paths; the probe's T1 asserts trigger-then-camera against the live data; daemon rebuilt, content-diff read, deployed with the pre-change binary banked by hash	Organ	1 u
Prompt-grammar port (nx_photo_grammar) (G2) after G1	pg_compose	content.db vocabulary (zimage_reinforcements, realism_microdetails, photo_styles, lighting, expressions, prompt_fragments, quality_vocabulary) seeded as planes; slot-ordered composer replaces the hand tail; probe-gated; June-repro dhash against the original must not regress from the G1 baseline	Organ	2 u
Live unconditional branch, selected by intent (G3) after G9,G16	gw_guidance_select	SYMBOL CORRECTED 2026-08-30 to match the matrix cell, which had been renamed while this row still said gw_model_select -- the board and the plan were declaring two different contract NAMES for one rung, and both were ABSENT-PROVEN. RE-SCOPED by measurement: the gate is the CFG SETTING not the model (cfg 1.0 with and without a negative produced byte-identical sha256 96de208a; cfg 4.0 produced 667f268b against 7b32b0ff), and the cfg-ladder arm is DEAD -- at the true June factors cfg 1.5 and 2.5 BOTH destroy the render with elf ears and fused fingers, so there is no usable window. The served checkpoint is zimage_turbo_nsfw_2602-Q8_0.gguf, an NSFW FINETUNE. CORRECTED 2026-08-30 BY HASH: a clean base checkpoint DOES exist on disk -- /volume1/ai/models/diffusion/z_image_turbo-Q8_0.gguf measures 6,577,440,704 bytes with sha256 df1c5baa86d1398c979495a6072dbcee79444fdb884a2445582ba0769c44e9a1, byte-identical to the publisher's lfs.oid in leejet/Z-Image-Turbo-GGUF (the stable-diffusion.cpp author's own quant repo), so the earlier no-clean-checkpoint census had not looked in the diffusion store and G9 is no longer this rung's blocker; the next arm is a render experiment on the clean checkpoint (does the negative fire at cfg 1.5 to 2.5 without the finetune's fantasy contamination), which nobody has run. PARTIAL CREDIT ALREADY BANKED: removing a contaminated LoRA (G16) made a clothed-scene instruction render EXACTLY the buttoned white cotton sundress asked for, so some of this wall was LoRA hygiene rather than guidance. DONE-RULE: 20 clothed-scene renders measure clothed under the wardrobe judge, published through the rigor envelope with n, an interval, the seed and the engine-config hash, against a pre-declared negative control that must NOT measure clothed.	Organ	2 u
Batch board (G4)	go_batch_list	named jobs with a scene list, a row appended per render as it lands, list, status with progress, stop and resume; the UI polls the board; an interrupted batch keeps every finished cid	Organ	1 u
External referee run (G5) after G2	gb_bench_run	HPSv2, PickScore and GenEval protocols executed on our renders; our numbers published beside the papers' tables; the run is a gate with a negative control	Local model	2 u
ControlNet and pose conditioning (G6) after G3	gw_controlnet	pose and structure maps dispatched through the worker; the 3D lane's skeletons are the deterministic source	Organ	2 u
Mask inpaint and outpaint (G7)	gw_inpaint_mask	COMPOSITION HALF LANDED 2026-08-30: gw_inpaint_mask (nx_gen_worker.nx) appends the engine's own mask and inpainting_mask_invert fields to the A1111 img2img body and refuses by named code (EMPTY, NOT-BASE64, INVERT-NOT-BOOLEAN); go_handle_img2img reads optional mask and mask_invert and returns a 400 naming the rule before any dispatch; nx_gen_inpaint_gate GREEN 15/15 via /api/gate_run and BITE-PROVEN (killed=1, not_reached=0), promoted, registered, declared in gen.gates; daemon built (sha 932d6141, contentdiff lost_from_live=0), pre-change binary banked by hash (a36b5d27), deployed via gdeploy@genorch with an audited pre-deploy-gate override (evidence-honesty RED on four unrelated domains), live hash re-read and the listen edge witnessed. DONE-RULE (render half OWED, GPU shared with a game at deploy time): mask rides the A1111 img2img shape the engine already serves; edited region judged by the coherence organ	Organ	1 u
Upscale on render (G8)	gw_upscale	ESRGAN in-engine wired as a gen verb; upscaled cid ingested beside the source with a GENREC link	Organ	0.5 u
Model acquisition (G9)	mg_fetch	HF, civitai and civitaiarchive pulls with an Authorization header, SHA256-verified by nx_vecfetch, registered in a model plane; a mismatch refuses	Organ	1 u
Identity LoRA training loop (G10) after G9	lt_train_deploy	favorites to dataset to train to deploy; the tuned model must EXCEED the baseline on the same prompts under the judge panel or it is not wired	Local model	2 u
Workflow graph executor (G11)	gen_graph_exec	the pathway lib gains an executor whose cells are the existing organs; partial re-execution on unchanged upstream cells	Organ	3 u
Model swap with TTL (G12)	mx_swap_ttl	LLM and diffusion time-share the 16 GB card under a residency TTL; VRAM per model measured and published	Organ	1 u
Prompt weighting and wildcards (G13) after G2	cw_prompt_weight	WEIGHTING HALF LANDED 2026-08-30 on the contract knowledge/gen_prompt_weight.conf declared before the code (permil units, envelope [min,max] as the first two ints, out-of-envelope REFUSED by named code never clamped, slot weights as additive ints 3..5 with 1000 the identity so the composed prompt is byte-identical until data changes): cw_prompt_weight, cw_weight_slot (shared composer; a data slot out of envelope falls back to the baseline slot) and cw_weight_admit (go_handle_generate refuses a user group with permil and byte offset named) in nx_companion_wardrobe.nx; the engine honours the syntax (elder-sdcpp conditioner.hpp per-token weights, mean-preserving). nx_gen_promptweight_gate GREEN 32/32 via /api/gate_run, BITE-PROVEN on the third attempt (the first two refused by build admission, exit 6), promoted, registered, declared in gen.gates; live on the deployed daemon (sha 932d6141): a (skin:13) group answered rule=OUT-OF-ENVELOPE permil=13000 before any dispatch. DONE-RULE (wildcard half OWED): weights and wildcard expansion as data; a wildcard row that changes nothing is a defect to rewrite, never delete	Organ	1 u
Hires and refiner pass (G14) after G8	gw_hires_pass	two-pass render through the worker; latency published beside the single pass	Organ	0.5 u
Workflow ingest and reproduce (G15) after G11	wi_reproduce	a rival's PNG metadata or workflow reproduced on our engine; fidelity measured by the June-repro ruler (dhash and MAD)	Organ	1 u
LoRA and checkpoint contamination screen before adoption (G16)	lc_screen	every acquired LoRA or checkpoint renders ONE fixed-seed screening image before it may enter elara_loras.txt or the served model path; a contaminated artifact is refused by name with the tell that fired. FOUND 2026-08-30: amateur_photography_zimage_v1 was LIVE in elara_loras.txt and produced elf ears at cfg 1.0 in a single-LoRA ladder, then a catastrophic identity-loss and facial-artifact render in the live pair -- while the platform registry reported ten UNRELATED LoRAs as NOT YET TRAINED and knew nothing about the artifacts actually loaded. The screen is the acquisition gate G9 needs.	Organ	1 u
Dispatch honesty -- a refusal that names the failing conjunct (G17)	gw_read_full	BUILT AND DEPLOYED 2026-08-30: /api/build (src_path verified, sha ecee75fb, 301306 B) then gdeploy@genorch DEPLOYED-GREEN listen=VERIFIED, pre-change binary banked by hash (banked_genorch_pre_g17_d1e95cb1.elf). The live daemon carries BUSY-TIMEOUT and the conjunct-naming refusals (string-grep flipped matches=0 to present at the live path). DONE-RULE MET 2026-08-30: nx_gen_dispatch_gate GREEN 15/15 against the LIVE daemon (silent worker reads BUSY-TIMEOUT, absent worker reads the health-probe refusal fast, neither reads registry-down; the gate's first run leaked to the real worker through the SSOT's absolute fallback and now writes its own swarm row plus an isolation tooth). Originally: the planted slow-worker fixture -- WRITTEN 2026-08-30 as nx_gen_dispatch_gate (a silent worker that listens and never accepts, and an absent worker, each against the LIVE binary launched on a throwaway port with budget=2 from a scratch cwd; ~80 s, run via nx_job_run; RED by construction against the pre-fix build, so the bite is the live history). ALSO FOUND THE SAME DAY: the CHAT path still carried the false registry-down message -- the generate path was cured and its sibling was not -- fixed with the same clastfail/canyup attribution (five named conjuncts) and riding the queued daemon build. DEPLOY-LANE DEFECT FILED on the way: the first two PROMOTED receipts changed nothing because hc_deploy_one resolves stages beside the LIVE path only -- see the control-plane debt row of 2026-08-30. The identical transformation was applied to the laptop tree and landed on EXACTLY 7632 bytes, byte-count-equal to the NAS artifact, then compiled with the sovereign toolchain: nx_gen_orchestrator_daemon built clean, asm=1451202B, elf 306,603 B, no gcc. CONTROL RUN FIRST AND IT DISCRIMINATED: the pre-fix 6396 B source produced the SAME unrelated error at line 71 with gw_read_full ABSENT from the symbol table, while the fixed source produced it at line 88 with gw_read_full PRESENT -- a 17-line shift exactly equal to the inserted helper. That unrelated error was a LOCAL tree defect, nx_netprobe_lib.nx truncated to 0 bytes on the laptop while the NAS holds it at 2528 B; repairing it from the NAS copy made the build succeed. So the compile risk of this edit is retired: it is not merely written, it is proven to assemble and link. go_worker_dispatch, go_worker_img2img and go_chat_dispatch shared one copy-pasted read loop whose `if r <= 0` collapsed a TIMEOUT and a clean EOF onto one branch, so a 180s timeout that read nothing returned 0, every caller tested only >0, the candidate loop exhausted, and the daemon reported worker ABSENCE. MEASURED: a socket watch showed the worker ESTABLISHED and held 164s while the caller was told no worker was available. Fix EXTRACTED to one gw_read_full rather than re-typed at three sites, returning 0-3 busy-timeout and 0-4 partial-read alongside the existing 0-1 socket and 0-2 connect, preserving the >0 success contract. The orchestrator now names BUSY-TIMEOUT, PARTIAL-READ, CONNECT-REFUSED or no-candidate-passed-its-probe, and only says all-down when anyup==0 -- the old message was false on its own control flow, since that path is reachable only AFTER go_worker_up returned 1. DONE-RULE: a planted slow-worker fixture must produce BUSY-TIMEOUT and must NOT produce worker-absence.	Organ	2 u
Async render dispatch so no render is ever dropped (G18) after G17	go_job_submit	BUILT AND DEPLOYED 2026-08-30 in the same genorch deploy as G17 (DEPLOYED-GREEN listen=VERIFIED, live sha ecee75fb): POST /api/generate_async and /api/img2img_async are live routes. UPDATE 2026-08-30: CHAT rides the async lane too (POST /api/chat_async; GO_JOB_CHAT reuses go_job_submit with the handler unchanged) and the emitted UI polls it -- the measured 17.6 s completion against the 15 s edge window was the whole of 'chat does not work'; Studio already used the batch lane. DONE-RULE MET 2026-08-30 -- nx_gen_dispatch_gate T3 (an answering worker held 6 s behind the submit round-trip) reads 25/25 GREEN against the LIVE daemon a36b5d27, after reading RED 23/25 against 5ef619e9 on exactly the two async teeth: the forked job child inherited the client socket, so the submit reply never reached EOF until the job finished; the child now closes it first. The bite is the live history, both directions. go_job_submit is EXTRACTED from go_handle_batch, which was the daemon's only async route, and go_handle_batch now CALLS it so there is one implementation that cannot drift. The mechanism cost nothing because it already existed: every handler writes its complete HTTP response to cfd, so a forked child opening the job FILE and passing that fd AS cfd makes any synchronous handler async with ZERO changes to the handler. Routes POST /api/generate_async and POST /api/img2img_async added AHEAD of their sync counterparts, because go_starts is a PREFIX match and the specific route is unreachable otherwise -- that ordering is load-bearing and is asserted in the patch. Sync routes unchanged for back-compat and the job reply is additive, so nothing that works today stops working. Compile-verified on the laptop sovereign toolchain before landing: asm 1451202 -> 1454989 B, elf 306,603 -> 307,396 B across the two edits. DONE-RULE: a render whose duration exceeds the dispatch budget must still land a cid, proven by a planted slow-worker fixture.	Organ	2 u
The /gen page is emitted, never handcrafted (G19)	gu_emit	nx_gen_ui_emit emits elara_ui.html from knowledge/gen_ui.conf: two emits byte-identical; every behaviour marker present; a conf missing a key refuses by name with exit 3 and writes nothing; the bytes the daemon serves equal the emitted bytes (nx_gen_ui_gate, 13 teeth). Any UI change is a conf row or an emitter change shipped through the gate -- never an edit to the served file. STATUS 2026-08-30: DONE-RULE MET -- nx_gen_ui_gate GREEN 13/13 on the NAS including served-page-is-the-emitted-page; emitter and gate promoted and registered; the served page is the emitted page.	Organ	1 u

Milestones

Milestone	Rungs	Cumulative
M0 · The June look back	G1,G2,G3,G17,G18	9 u
M1 · Studio parity	G4,G7,G8,G13,G14,G19	14 u
M2 · Control and models	G6,G9,G12,G16	19 u
M3 · Referee and training	G5,G10	23 u
M4 · Graph parity	G11,G15	27 u

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Broken-parts census of /gen as the browser sees it, on a beat (G28)	gh_album_years	nx_galx_health gen probes the orchestrator on loopback every beat -- /health, every /api/recent image fetched and sniffed as an image, the album day labels -- writes gen.rows and gen.census under knowledge/status/galxhealth and the gen.stamp gauge this page renders; the gate drives the classifiers on planted fixtures; a leg that did not run reads BLIND, never zero	Organ	0.5 u

Ladder to SOTA

Two dated targets, and every rung's role toward them. Best in class is the proven, deployed leader today; the frontier is the research edge as of its month. Each target is dated by a sotabar row and judged against the current month by the bar-age ruler (FRESH, ATTESTED with the fallback month named, or refused). Each rung is substrate (what the target's own methods consume), an arm (a sovereign arm that triangulates without competing for the number), a contender (its done-rule is a number against the target) or superseded (kept as capability, retired as a claim), and carries the rungs it stands on.

best-in-class (T-BIC-GEN), dated by SB-GEN-T2I = 2026-10 FRESH. A local studio at ComfyUI-class control on Qwen-Image-2.1 or Z-Image-Turbo class open weights with few-step distills, ControlNet, inpaint, upscale and batch, judged on HPSv2, PickScore and GenEval beside the papers' own tables, with the June realism reproduced by ruler [hf_t2i_202610] ref hf_t2i_202610

frontier (T-FRONTIER-GEN), dated by SB-GEN-T2I = 2026-10 FRESH. Preference-and-rubric post-training on our own judge panel and identity data, with provenance-named assets and pre-spend QC no rival ships; pixel-space and unified multimodal lines tracked as substrates until a sovereign engine can run them [arxiv_t2i_202610] ref arxiv_t2i_202610

Rung	Role	Toward	Stands on	Why
G1	substrate	T-BIC-GEN	-	one composer, trigger-first slot order: the prompt path every render crosses
G2	substrate	T-BIC-GEN	G1	the prompt grammar as planes
G3	contender	T-BIC-GEN	G9,G16	a negative-capable guidance path: the controllability wall every clothed render hits
G4	contender	T-BIC-GEN	-	batch board (studio parity)
G5	arm	T-BIC-GEN	G2	the external referee run prints our row beside the papers
G6	contender	T-BIC-GEN	G3	ControlNet and pose conditioning
G7	contender	T-BIC-GEN	-	mask inpaint and outpaint
G8	contender	T-BIC-GEN	-	upscale on render
G9	substrate	T-BIC-GEN	-	model acquisition with hash verification: the door through which Qwen-Image-2.1 class weights arrive
G10	contender	T-FRONTIER-GEN	G9	identity LoRA training that must exceed the baseline under the judge panel
G11	contender	T-BIC-GEN	-	workflow graph executor (ComfyUI class)
G12	substrate	T-BIC-GEN	-	model swap with TTL on one 16 GB card
G13	contender	T-BIC-GEN	G2	prompt weighting and wildcards
G14	contender	T-BIC-GEN	G8	hires and refiner pass
G15	contender	T-BIC-GEN	G11	reproduce a rival's workflow on our engine
G16	substrate	T-FRONTIER-GEN	-	contamination screen before adoption: pre-spend QC no rival ships
G17	substrate	T-BIC-GEN	-	dispatch honesty
G18	substrate	T-BIC-GEN	G17	async dispatch so no render is dropped
G19	substrate	T-BIC-GEN	-	the page is emitted, never handcrafted
G28	arm	T-FRONTIER-GEN	-	the self-census of the surface, published on a beat

Ladder verdict. LADDERED. Targets, rung roles and bars are counted in the tables below, each partition re-derived beside its total. Bars read FRESH against the current month 2026-10.

Ladder verdict, counted (17 rows, more than the 12-row disclosure bar)

Ladder verdict, counted
Class	Count	Note
targets	2	sotatarget rows on the plan
best in class	1	targets whose class is best-in-class, by code
frontier	1	targets whose class is frontier, by code
undated	0	targets naming no bar that exists
bad class	0	targets whose class is neither word
duplicate	0	target ids an earlier row already carries
rungs	20	rung rows on the plan
placed	20	rungs carrying a rungrole row
unplaced	0	rungs with no role toward any target
bad role	0	rungrole rows naming a role the ladder does not know
bad target	0	rungrole rows naming a target no sotatarget row declares
rungs, summed	20	RECONCILED: the parts sum to the total 20
substrate	8	what the target's own methods consume
arm	2	a sovereign arm that triangulates without competing for the number
contender	10	its done-rule is a number against the target
superseded	0	kept as capability, retired as a claim
placed, summed	20	RECONCILED: the parts sum to the total 20

Bar freshness, counted
Class	Count	Note
bars	1	sotabar rows on the plan
fresh	1	the bar month is the current month
attested	0	an older bar with a this-month barscan reporting no newer entry
stale	0	a this-month scan saw a newer month the bar ignores
unattested	0	nobody looked this month
malformed	0	a sotabar row missing a field
duplicate	0	bar ids an earlier row already carries
bars, summed	1	RECONCILED: the parts sum to the total 1
current month	2026-10	the month every bar is judged against

Treatment verdict. NOT DECLARED. No rung on this board declares its three-step treatment (a treat row naming the MCP row, the API route, the primitive it composes, the workflow that composes it and the agent beat that runs it); the ranker labels every rung TREATMENT-UNDECLARED until it does.

Inspect a rung and its prerequisites

Declared nodes 20. Rank input binding: PLAN_MATRIX_BOUND_ONLY. Dependency order is authored. Implementation, acceptance evidence, authority and resource readiness are unverified. No action is recommended or dispatched here.

Plan SHA-256 b50dba8a7e460c933aa857389a23d673278040e744b59108a38c4ca9ee24d8a5. Target rows 2; role rows 20. Existing risks and release worklog retain their own scope; no node completion is inferred.

Runtime observation coverage: 0 of 20 rungs have current, identity-bound execution observations. Observation may report RED. Accepted completion remains unknown; source presence, qualification and deployment do not establish it.

Use Enter or Space on a rung to inspect its contract. Prerequisite links locate another rung in this list; open its summary to inspect it. Estimates are authored effort, not forecasts.

G1 — Trigger-first slot order and ONE composer
Prerequisites: None declared; this does not establish execution eligibility.
Contract: go_load_trigger
Acceptance: elara_trigger.txt leads the prompt on both photo paths; the probe's T1 asserts trigger-then-camera against the live data; daemon rebuilt, content-diff read, deployed with the pre-change binary banked by hash
Authored effort: 1. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G2 — Prompt-grammar port (nx_photo_grammar)
Prerequisites: G1 (acceptance unverified)
Contract: pg_compose
Acceptance: content.db vocabulary (zimage_reinforcements, realism_microdetails, photo_styles, lighting, expressions, prompt_fragments, quality_vocabulary) seeded as planes; slot-ordered composer replaces the hand tail; probe-gated; June-repro dhash against the original must not regress from the G1 baseline
Authored effort: 2. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G3 — Live unconditional branch, selected by intent
Prerequisites: G9 (acceptance unverified), G16 (acceptance unverified)
Contract: gw_guidance_select
Acceptance: SYMBOL CORRECTED 2026-08-30 to match the matrix cell, which had been renamed while this row still said gw_model_select -- the board and the plan were declaring two different contract NAMES for one rung, and both were ABSENT-PROVEN. RE-SCOPED by measurement: the gate is the CFG SETTING not the model (cfg 1.0 with and without a negative produced byte-identical sha256 96de208a; cfg 4.0 produced 667f268b against 7b32b0ff), and the cfg-ladder arm is DEAD -- at the true June factors cfg 1.5 and 2.5 BOTH destroy the render with elf ears and fused fingers, so there is no usable window. The served checkpoint is zimage_turbo_nsfw_2602-Q8_0.gguf, an NSFW FINETUNE. CORRECTED 2026-08-30 BY HASH: a clean base checkpoint DOES exist on disk -- /volume1/ai/models/diffusion/z_image_turbo-Q8_0.gguf measures 6,577,440,704 bytes with sha256 df1c5baa86d1398c979495a6072dbcee79444fdb884a2445582ba0769c44e9a1, byte-identical to the publisher's lfs.oid in leejet/Z-Image-Turbo-GGUF (the stable-diffusion.cpp author's own quant repo), so the earlier no-clean-checkpoint census had not looked in the diffusion store and G9 is no longer this rung's blocker; the next arm is a render experiment on the clean checkpoint (does the negative fire at cfg 1.5 to 2.5 without the finetune's fantasy contamination), which nobody has run. PARTIAL CREDIT ALREADY BANKED: removing a contaminated LoRA (G16) made a clothed-scene instruction render EXACTLY the buttoned white cotton sundress asked for, so some of this wall was LoRA hygiene rather than guidance. DONE-RULE: 20 clothed-scene renders measure clothed under the wardrobe judge, published through the rigor envelope with n, an interval, the seed and the engine-config hash, against a pre-declared negative control that must NOT measure clothed.
Authored effort: 2. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G4 — Batch board
Prerequisites: None declared; this does not establish execution eligibility.
Contract: go_batch_list
Acceptance: named jobs with a scene list, a row appended per render as it lands, list, status with progress, stop and resume; the UI polls the board; an interrupted batch keeps every finished cid
Authored effort: 1. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G5 — External referee run
Prerequisites: G2 (acceptance unverified)
Contract: gb_bench_run
Acceptance: HPSv2, PickScore and GenEval protocols executed on our renders; our numbers published beside the papers' tables; the run is a gate with a negative control
Authored effort: 2. Executor kind: Local model. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G6 — ControlNet and pose conditioning
Prerequisites: G3 (acceptance unverified)
Contract: gw_controlnet
Acceptance: pose and structure maps dispatched through the worker; the 3D lane's skeletons are the deterministic source
Authored effort: 2. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G7 — Mask inpaint and outpaint
Prerequisites: None declared; this does not establish execution eligibility.
Contract: gw_inpaint_mask
Acceptance: COMPOSITION HALF LANDED 2026-08-30: gw_inpaint_mask (nx_gen_worker.nx) appends the engine's own mask and inpainting_mask_invert fields to the A1111 img2img body and refuses by named code (EMPTY, NOT-BASE64, INVERT-NOT-BOOLEAN); go_handle_img2img reads optional mask and mask_invert and returns a 400 naming the rule before any dispatch; nx_gen_inpaint_gate GREEN 15/15 via /api/gate_run and BITE-PROVEN (killed=1, not_reached=0), promoted, registered, declared in gen.gates; daemon built (sha 932d6141, contentdiff lost_from_live=0), pre-change binary banked by hash (a36b5d27), deployed via gdeploy@genorch with an audited pre-deploy-gate override (evidence-honesty RED on four unrelated domains), live hash re-read and the listen edge witnessed. DONE-RULE (render half OWED, GPU shared with a game at deploy time): mask rides the A1111 img2img shape the engine already serves; edited region judged by the coherence organ
Authored effort: 1. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G8 — Upscale on render
Prerequisites: None declared; this does not establish execution eligibility.
Contract: gw_upscale
Acceptance: ESRGAN in-engine wired as a gen verb; upscaled cid ingested beside the source with a GENREC link
Authored effort: 0.5. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G9 — Model acquisition
Prerequisites: None declared; this does not establish execution eligibility.
Contract: mg_fetch
Acceptance: HF, civitai and civitaiarchive pulls with an Authorization header, SHA256-verified by nx_vecfetch, registered in a model plane; a mismatch refuses
Authored effort: 1. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G10 — Identity LoRA training loop
Prerequisites: G9 (acceptance unverified)
Contract: lt_train_deploy
Acceptance: favorites to dataset to train to deploy; the tuned model must EXCEED the baseline on the same prompts under the judge panel or it is not wired
Authored effort: 2. Executor kind: Local model. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G11 — Workflow graph executor
Prerequisites: None declared; this does not establish execution eligibility.
Contract: gen_graph_exec
Acceptance: the pathway lib gains an executor whose cells are the existing organs; partial re-execution on unchanged upstream cells
Authored effort: 3. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G12 — Model swap with TTL
Prerequisites: None declared; this does not establish execution eligibility.
Contract: mx_swap_ttl
Acceptance: LLM and diffusion time-share the 16 GB card under a residency TTL; VRAM per model measured and published
Authored effort: 1. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G13 — Prompt weighting and wildcards
Prerequisites: G2 (acceptance unverified)
Contract: cw_prompt_weight
Acceptance: WEIGHTING HALF LANDED 2026-08-30 on the contract knowledge/gen_prompt_weight.conf declared before the code (permil units, envelope [min,max] as the first two ints, out-of-envelope REFUSED by named code never clamped, slot weights as additive ints 3..5 with 1000 the identity so the composed prompt is byte-identical until data changes): cw_prompt_weight, cw_weight_slot (shared composer; a data slot out of envelope falls back to the baseline slot) and cw_weight_admit (go_handle_generate refuses a user group with permil and byte offset named) in nx_companion_wardrobe.nx; the engine honours the syntax (elder-sdcpp conditioner.hpp per-token weights, mean-preserving). nx_gen_promptweight_gate GREEN 32/32 via /api/gate_run, BITE-PROVEN on the third attempt (the first two refused by build admission, exit 6), promoted, registered, declared in gen.gates; live on the deployed daemon (sha 932d6141): a (skin:13) group answered rule=OUT-OF-ENVELOPE permil=13000 before any dispatch. DONE-RULE (wildcard half OWED): weights and wildcard expansion as data; a wildcard row that changes nothing is a defect to rewrite, never delete
Authored effort: 1. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G14 — Hires and refiner pass
Prerequisites: G8 (acceptance unverified)
Contract: gw_hires_pass
Acceptance: two-pass render through the worker; latency published beside the single pass
Authored effort: 0.5. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G15 — Workflow ingest and reproduce
Prerequisites: G11 (acceptance unverified)
Contract: wi_reproduce
Acceptance: a rival's PNG metadata or workflow reproduced on our engine; fidelity measured by the June-repro ruler (dhash and MAD)
Authored effort: 1. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G16 — LoRA and checkpoint contamination screen before adoption
Prerequisites: None declared; this does not establish execution eligibility.
Contract: lc_screen
Acceptance: every acquired LoRA or checkpoint renders ONE fixed-seed screening image before it may enter elara_loras.txt or the served model path; a contaminated artifact is refused by name with the tell that fired. FOUND 2026-08-30: amateur_photography_zimage_v1 was LIVE in elara_loras.txt and produced elf ears at cfg 1.0 in a single-LoRA ladder, then a catastrophic identity-loss and facial-artifact render in the live pair -- while the platform registry reported ten UNRELATED LoRAs as NOT YET TRAINED and knew nothing about the artifacts actually loaded. The screen is the acquisition gate G9 needs.
Authored effort: 1. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G17 — Dispatch honesty -- a refusal that names the failing conjunct
Prerequisites: None declared; this does not establish execution eligibility.
Contract: gw_read_full
Acceptance: BUILT AND DEPLOYED 2026-08-30: /api/build (src_path verified, sha ecee75fb, 301306 B) then gdeploy@genorch DEPLOYED-GREEN listen=VERIFIED, pre-change binary banked by hash (banked_genorch_pre_g17_d1e95cb1.elf). The live daemon carries BUSY-TIMEOUT and the conjunct-naming refusals (string-grep flipped matches=0 to present at the live path). DONE-RULE MET 2026-08-30: nx_gen_dispatch_gate GREEN 15/15 against the LIVE daemon (silent worker reads BUSY-TIMEOUT, absent worker reads the health-probe refusal fast, neither reads registry-down; the gate's first run leaked to the real worker through the SSOT's absolute fallback and now writes its own swarm row plus an isolation tooth). Originally: the planted slow-worker fixture -- WRITTEN 2026-08-30 as nx_gen_dispatch_gate (a silent worker that listens and never accepts, and an absent worker, each against the LIVE binary launched on a throwaway port with budget=2 from a scratch cwd; ~80 s, run via nx_job_run; RED by construction against the pre-fix build, so the bite is the live history). ALSO FOUND THE SAME DAY: the CHAT path still carried the false registry-down message -- the generate path was cured and its sibling was not -- fixed with the same clastfail/canyup attribution (five named conjuncts) and riding the queued daemon build. DEPLOY-LANE DEFECT FILED on the way: the first two PROMOTED receipts changed nothing because hc_deploy_one resolves stages beside the LIVE path only -- see the control-plane debt row of 2026-08-30. The identical transformation was applied to the laptop tree and landed on EXACTLY 7632 bytes, byte-count-equal to the NAS artifact, then compiled with the sovereign toolchain: nx_gen_orchestrator_daemon built clean, asm=1451202B, elf 306,603 B, no gcc. CONTROL RUN FIRST AND IT DISCRIMINATED: the pre-fix 6396 B source produced the SAME unrelated error at line 71 with gw_read_full ABSENT from the symbol table, while the fixed source produced it at line 88 with gw_read_full PRESENT -- a 17-line shift exactly equal to the inserted helper. That unrelated error was a LOCAL tree defect, nx_netprobe_lib.nx truncated to 0 bytes on the laptop while the NAS holds it at 2528 B; repairing it from the NAS copy made the build succeed. So the compile risk of this edit is retired: it is not merely written, it is proven to assemble and link. go_worker_dispatch, go_worker_img2img and go_chat_dispatch shared one copy-pasted read loop whose `if r <= 0` collapsed a TIMEOUT and a clean EOF onto one branch, so a 180s timeout that read nothing returned 0, every caller tested only >0, the candidate loop exhausted, and the daemon reported worker ABSENCE. MEASURED: a socket watch showed the worker ESTABLISHED and held 164s while the caller was told no worker was available. Fix EXTRACTED to one gw_read_full rather than re-typed at three sites, returning 0-3 busy-timeout and 0-4 partial-read alongside the existing 0-1 socket and 0-2 connect, preserving the >0 success contract. The orchestrator now names BUSY-TIMEOUT, PARTIAL-READ, CONNECT-REFUSED or no-candidate-passed-its-probe, and only says all-down when anyup==0 -- the old message was false on its own control flow, since that path is reachable only AFTER go_worker_up returned 1. DONE-RULE: a planted slow-worker fixture must produce BUSY-TIMEOUT and must NOT produce worker-absence.
Authored effort: 2. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G18 — Async render dispatch so no render is ever dropped
Prerequisites: G17 (acceptance unverified)
Contract: go_job_submit
Acceptance: BUILT AND DEPLOYED 2026-08-30 in the same genorch deploy as G17 (DEPLOYED-GREEN listen=VERIFIED, live sha ecee75fb): POST /api/generate_async and /api/img2img_async are live routes. UPDATE 2026-08-30: CHAT rides the async lane too (POST /api/chat_async; GO_JOB_CHAT reuses go_job_submit with the handler unchanged) and the emitted UI polls it -- the measured 17.6 s completion against the 15 s edge window was the whole of 'chat does not work'; Studio already used the batch lane. DONE-RULE MET 2026-08-30 -- nx_gen_dispatch_gate T3 (an answering worker held 6 s behind the submit round-trip) reads 25/25 GREEN against the LIVE daemon a36b5d27, after reading RED 23/25 against 5ef619e9 on exactly the two async teeth: the forked job child inherited the client socket, so the submit reply never reached EOF until the job finished; the child now closes it first. The bite is the live history, both directions. go_job_submit is EXTRACTED from go_handle_batch, which was the daemon's only async route, and go_handle_batch now CALLS it so there is one implementation that cannot drift. The mechanism cost nothing because it already existed: every handler writes its complete HTTP response to cfd, so a forked child opening the job FILE and passing that fd AS cfd makes any synchronous handler async with ZERO changes to the handler. Routes POST /api/generate_async and POST /api/img2img_async added AHEAD of their sync counterparts, because go_starts is a PREFIX match and the specific route is unreachable otherwise -- that ordering is load-bearing and is asserted in the patch. Sync routes unchanged for back-compat and the job reply is additive, so nothing that works today stops working. Compile-verified on the laptop sovereign toolchain before landing: asm 1451202 -> 1454989 B, elf 306,603 -> 307,396 B across the two edits. DONE-RULE: a render whose duration exceeds the dispatch budget must still land a cid, proven by a planted slow-worker fixture.
Authored effort: 2. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G19 — The /gen page is emitted, never handcrafted
Prerequisites: None declared; this does not establish execution eligibility.
Contract: gu_emit
Acceptance: nx_gen_ui_emit emits elara_ui.html from knowledge/gen_ui.conf: two emits byte-identical; every behaviour marker present; a conf missing a key refuses by name with exit 3 and writes nothing; the bytes the daemon serves equal the emitted bytes (nx_gen_ui_gate, 13 teeth). Any UI change is a conf row or an emitter change shipped through the gate -- never an edit to the served file. STATUS 2026-08-30: DONE-RULE MET -- nx_gen_ui_gate GREEN 13/13 on the NAS including served-page-is-the-emitted-page; emitter and gate promoted and registered; the served page is the emitted page.
Authored effort: 1. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
G28 — Broken-parts census of /gen as the browser sees it, on a beat
Prerequisites: None declared; this does not establish execution eligibility.
Contract: gh_album_years
Acceptance: nx_galx_health gen probes the orchestrator on loopback every beat -- /health, every /api/recent image fetched and sniffed as an image, the album day labels -- writes gen.rows and gen.census under knowledge/status/galxhealth and the gen.stamp gauge this page renders; the gate drives the classifiers on planted fixtures; a leg that did not run reads BLIND, never zero
Authored effort: 0.5. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.

Learning and practice paths

No structured learning path is declared for this plan. Existing research, roadmap and worklog remain available above.

Rung dossiers — every record that bears on a rung, in one place

Joined by symbol, never typed. Each dossier is assembled on every publish from the rows this board already carries: the plan's rung row, the matrix row that carries its contract symbol (its status re-measured by the one symbol ruler), the gate the organ declares and the verdict the evidence referee recorded on its last run, the reference marks in the rung text, the design rows that name the rung and the work log filed against it. The capability matrix links each cell to its dossier and each dossier links back to its cell and forward to the rungs it stands on; a dependency naming no rung on this plan is written as UNRESOLVED rather than linked. Absences are named: NO-MATRIX-ROW, NO-GATE-DECLARED, UNRECORDED.

Gate verdicts: UNSTAMPED — no per-gate verdict artifact is published for this domain yet (nx_swcompare_evidence gen writes it beside its stamp); every gate below reads UNRECORDED until it runs.

Gate declarations: knowledge/organ_gate.conf read from this working directory (1 rows) and from the estate root (196 rows); rows in both counted once: 0; declarations read: 197. Every declaration in either copy counts, so a stub beside the publisher can never hide the root.

G1 Trigger-first slot order and ONE composer · cell PRESENT · gate NO-GATE-DECLARED

Contract	`go_load_trigger` · executor Organ · 1 u · stands on nothing
Definition of done	elara_trigger.txt leads the prompt on both photo paths; the probe's T1 asserts trigger-then-camera against the live data; daemon rebuilt, content-diff read, deployed with the pre-change binary banked by hash
Matrix cell	PROMPT: realism trigger prefix as data (elara_trigger.txt) -- the June slot order restored, trigger leads the prompt · `runtime/_hdl_build/nx_companion_identity.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_companion_identity`; declare `nx_companion_identity TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G2 Prompt-grammar port (nx_photo_grammar) · cell WATCHING · gate NO-GATE-DECLARED

Contract	`pg_compose` · executor Organ · 2 u · stands on G1
Definition of done	content.db vocabulary (zimage_reinforcements, realism_microdetails, photo_styles, lighting, expressions, prompt_fragments, quality_vocabulary) seeded as planes; slot-ordered composer replaces the hand tail; probe-gated; June-repro dhash against the original must not regress from the G1 baseline
Matrix cell	PROMPT: full prompt-grammar composer over the mined vocabulary planes (the 1,640-row content.db port) · `runtime/nx_photo_grammar.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_photo_grammar`; declare `nx_photo_grammar TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G3 Live unconditional branch, selected by intent · cell WATCHING · gate NO-GATE-DECLARED

Contract	`gw_guidance_select` · executor Organ · 2 u · stands on G9, G16
Definition of done	SYMBOL CORRECTED 2026-08-30 to match the matrix cell, which had been renamed while this row still said gw_model_select -- the board and the plan were declaring two different contract NAMES for one rung, and both were ABSENT-PROVEN. RE-SCOPED by measurement: the gate is the CFG SETTING not the model (cfg 1.0 with and without a negative produced byte-identical sha256 96de208a; cfg 4.0 produced 667f268b against 7b32b0ff), and the cfg-ladder arm is DEAD -- at the true June factors cfg 1.5 and 2.5 BOTH destroy the render with elf ears and fused fingers, so there is no usable window. The served checkpoint is zimage_turbo_nsfw_2602-Q8_0.gguf, an NSFW FINETUNE. CORRECTED 2026-08-30 BY HASH: a clean base checkpoint DOES exist on disk -- /volume1/ai/models/diffusion/z_image_turbo-Q8_0.gguf measures 6,577,440,704 bytes with sha256 df1c5baa86d1398c979495a6072dbcee79444fdb884a2445582ba0769c44e9a1, byte-identical to the publisher's lfs.oid in leejet/Z-Image-Turbo-GGUF (the stable-diffusion.cpp author's own quant repo), so the earlier no-clean-checkpoint census had not looked in the diffusion store and G9 is no longer this rung's blocker; the next arm is a render experiment on the clean checkpoint (does the negative fire at cfg 1.5 to 2.5 without the finetune's fantasy contamination), which nobody has run. PARTIAL CREDIT ALREADY BANKED: removing a contaminated LoRA (G16) made a clothed-scene instruction render EXACTLY the buttoned white cotton sundress asked for, so some of this wall was LoRA hygiene rather than guidance. DONE-RULE: 20 clothed-scene renders measure clothed under the wardrobe judge, published through the rigor envelope with n, an interval, the seed and the engine-config hash, against a pre-declared negative control that must NOT measure clothed.
Matrix cell	CONTROL: a LIVE unconditional branch -- a negative prompt that actually fires, selected by intent (the ER-8 keystone) · `runtime/_hdl_build/nx_gen_worker.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_gen_worker`; declare `nx_gen_worker TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G4 Batch board · cell WATCHING · gate NO-GATE-DECLARED

Contract	`go_batch_list` · executor Organ · 1 u · stands on nothing
Definition of done	named jobs with a scene list, a row appended per render as it lands, list, status with progress, stop and resume; the UI polls the board; an interrupted batch keeps every finished cid
Matrix cell	UI: batch BOARD -- named jobs, per-image progress as each render lands, list, stop and resume · `runtime/_hdl_build/nx_gen_orchestrator.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_gen_orchestrator`; declare `nx_gen_orchestrator TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G5 External referee run · cell WATCHING · gate NO-GATE-DECLARED

Contract	`gb_bench_run` · executor Local model · 2 u · stands on G2
Definition of done	HPSv2, PickScore and GenEval protocols executed on our renders; our numbers published beside the papers' tables; the run is a gate with a negative control
Matrix cell	JUDGE: external referee run PUBLISHED WITH ITS STATISTICS -- HPSv2, PickScore and GenEval on our renders with n, an interval, the seed and the engine-config hash · `runtime/nx_gen_bench.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_gen_bench`; declare `nx_gen_bench TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G6 ControlNet and pose conditioning · cell WATCHING · gate NO-GATE-DECLARED

Contract	`gw_controlnet` · executor Organ · 2 u · stands on G3
Definition of done	pose and structure maps dispatched through the worker; the 3D lane's skeletons are the deterministic source
Matrix cell	CONTROL: ControlNet and pose or structure conditioning · `runtime/_hdl_build/nx_gen_worker.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_gen_worker`; declare `nx_gen_worker TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G7 Mask inpaint and outpaint · cell PRESENT · gate NO-GATE-DECLARED

Contract	`gw_inpaint_mask` · executor Organ · 1 u · stands on nothing
Definition of done	COMPOSITION HALF LANDED 2026-08-30: gw_inpaint_mask (nx_gen_worker.nx) appends the engine's own mask and inpainting_mask_invert fields to the A1111 img2img body and refuses by named code (EMPTY, NOT-BASE64, INVERT-NOT-BOOLEAN); go_handle_img2img reads optional mask and mask_invert and returns a 400 naming the rule before any dispatch; nx_gen_inpaint_gate GREEN 15/15 via /api/gate_run and BITE-PROVEN (killed=1, not_reached=0), promoted, registered, declared in gen.gates; daemon built (sha 932d6141, contentdiff lost_from_live=0), pre-change binary banked by hash (a36b5d27), deployed via gdeploy@genorch with an audited pre-deploy-gate override (evidence-honesty RED on four unrelated domains), live hash re-read and the listen edge witnessed. DONE-RULE (render half OWED, GPU shared with a game at deploy time): mask rides the A1111 img2img shape the engine already serves; edited region judged by the coherence organ
Matrix cell	CONTROL: mask-based inpaint and outpaint · `runtime/_hdl_build/nx_gen_worker.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_gen_worker`; declare `nx_gen_worker TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G8 Upscale on render · cell WATCHING · gate NO-GATE-DECLARED

Contract	`gw_upscale` · executor Organ · 0.5 u · stands on nothing
Definition of done	ESRGAN in-engine wired as a gen verb; upscaled cid ingested beside the source with a GENREC link
Matrix cell	CONTROL: super-resolution upscale of a render (ESRGAN compiles into the library and is LINKED OUT of the server) · `runtime/_hdl_build/nx_gen_worker.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_gen_worker`; declare `nx_gen_worker TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G9 Model acquisition · cell PRESENT · gate NO-GATE-DECLARED

Contract	`mg_fetch` · executor Organ · 1 u · stands on nothing
Definition of done	HF, civitai and civitaiarchive pulls with an Authorization header, SHA256-verified by nx_vecfetch, registered in a model plane; a mismatch refuses
Matrix cell	MODEL: hash-verified acquisition from HF, civitai and civitaiarchive into the model registry · `runtime/nx_model_get.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_model_get`; declare `nx_model_get TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G10 Identity LoRA training loop · cell WATCHING · gate NO-GATE-DECLARED

Contract	`lt_train_deploy` · executor Local model · 2 u · stands on G9
Definition of done	favorites to dataset to train to deploy; the tuned model must EXCEED the baseline on the same prompts under the judge panel or it is not wired
Matrix cell	MODEL: identity LoRA training loop -- favorites to dataset to train to judge-gated deploy · `runtime/nx_lora_train.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_lora_train`; declare `nx_lora_train TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G11 Workflow graph executor · cell WATCHING · gate NO-GATE-DECLARED

Contract	`gen_graph_exec` · executor Organ · 3 u · stands on nothing
Definition of done	the pathway lib gains an executor whose cells are the existing organs; partial re-execution on unchanged upstream cells
Matrix cell	WORKFLOW: execute a workflow graph over the gen worker (ComfyUI-class pipelines) · `runtime/nx_gen_graph.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_gen_graph`; declare `nx_gen_graph TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G12 Model swap with TTL · cell WATCHING · gate NO-GATE-DECLARED

Contract	`mx_swap_ttl` · executor Organ · 1 u · stands on nothing
Definition of done	LLM and diffusion time-share the 16 GB card under a residency TTL; VRAM per model measured and published
Matrix cell	RESOURCE: on-demand model swap with a TTL -- LLM and diffusion time-sharing one 16 GB card · `runtime/nx_mesh_mux.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_mesh_mux`; declare `nx_mesh_mux TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G13 Prompt weighting and wildcards · cell PRESENT · gate NO-GATE-DECLARED

Contract	`cw_prompt_weight` · executor Organ · 1 u · stands on G2
Definition of done	WEIGHTING HALF LANDED 2026-08-30 on the contract knowledge/gen_prompt_weight.conf declared before the code (permil units, envelope [min,max] as the first two ints, out-of-envelope REFUSED by named code never clamped, slot weights as additive ints 3..5 with 1000 the identity so the composed prompt is byte-identical until data changes): cw_prompt_weight, cw_weight_slot (shared composer; a data slot out of envelope falls back to the baseline slot) and cw_weight_admit (go_handle_generate refuses a user group with permil and byte offset named) in nx_companion_wardrobe.nx; the engine honours the syntax (elder-sdcpp conditioner.hpp per-token weights, mean-preserving). nx_gen_promptweight_gate GREEN 32/32 via /api/gate_run, BITE-PROVEN on the third attempt (the first two refused by build admission, exit 6), promoted, registered, declared in gen.gates; live on the deployed daemon (sha 932d6141): a (skin:13) group answered rule=OUT-OF-ENVELOPE permil=13000 before any dispatch. DONE-RULE (wildcard half OWED): weights and wildcard expansion as data; a wildcard row that changes nothing is a defect to rewrite, never delete
Matrix cell	CONTROL: prompt weighting, wildcards and embeddings · `runtime/_hdl_build/nx_companion_wardrobe.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_companion_wardrobe`; declare `nx_companion_wardrobe TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G14 Hires and refiner pass · cell WATCHING · gate NO-GATE-DECLARED

Contract	`gw_hires_pass` · executor Organ · 0.5 u · stands on G8
Definition of done	two-pass render through the worker; latency published beside the single pass
Matrix cell	CONTROL: hires-fix and refiner two-pass rendering · `runtime/_hdl_build/nx_gen_worker.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_gen_worker`; declare `nx_gen_worker TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G15 Workflow ingest and reproduce · cell WATCHING · gate NO-GATE-DECLARED

Contract	`wi_reproduce` · executor Organ · 1 u · stands on G11
Definition of done	a rival's PNG metadata or workflow reproduced on our engine; fidelity measured by the June-repro ruler (dhash and MAD)
Matrix cell	WORKFLOW: ingest a rival's workflow or PNG metadata and reproduce the render · `runtime/nx_workflow_ingest.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_workflow_ingest`; declare `nx_workflow_ingest TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G16 LoRA and checkpoint contamination screen before adoption · cell PRESENT · gate NO-GATE-DECLARED

Contract	`lc_screen` · executor Organ · 1 u · stands on nothing
Definition of done	every acquired LoRA or checkpoint renders ONE fixed-seed screening image before it may enter elara_loras.txt or the served model path; a contaminated artifact is refused by name with the tell that fired. FOUND 2026-08-30: amateur_photography_zimage_v1 was LIVE in elara_loras.txt and produced elf ears at cfg 1.0 in a single-LoRA ladder, then a catastrophic identity-loss and facial-artifact render in the live pair -- while the platform registry reported ten UNRELATED LoRAs as NOT YET TRAINED and knew nothing about the artifacts actually loaded. The screen is the acquisition gate G9 needs.
Matrix cell	QC: contamination screen on every acquired LoRA and checkpoint BEFORE it may enter the served path · `runtime/_hdl_build/nx_lora_screen_lib.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_lora_screen_lib`; declare `nx_lora_screen_lib TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G17 Dispatch honesty -- a refusal that names the failing conjunct · cell PRESENT · gate NO-GATE-DECLARED

Contract	`gw_read_full` · executor Organ · 2 u · stands on nothing
Definition of done	BUILT AND DEPLOYED 2026-08-30: /api/build (src_path verified, sha ecee75fb, 301306 B) then gdeploy@genorch DEPLOYED-GREEN listen=VERIFIED, pre-change binary banked by hash (banked_genorch_pre_g17_d1e95cb1.elf). The live daemon carries BUSY-TIMEOUT and the conjunct-naming refusals (string-grep flipped matches=0 to present at the live path). DONE-RULE MET 2026-08-30: nx_gen_dispatch_gate GREEN 15/15 against the LIVE daemon (silent worker reads BUSY-TIMEOUT, absent worker reads the health-probe refusal fast, neither reads registry-down; the gate's first run leaked to the real worker through the SSOT's absolute fallback and now writes its own swarm row plus an isolation tooth). Originally: the planted slow-worker fixture -- WRITTEN 2026-08-30 as nx_gen_dispatch_gate (a silent worker that listens and never accepts, and an absent worker, each against the LIVE binary launched on a throwaway port with budget=2 from a scratch cwd; ~80 s, run via nx_job_run; RED by construction against the pre-fix build, so the bite is the live history). ALSO FOUND THE SAME DAY: the CHAT path still carried the false registry-down message -- the generate path was cured and its sibling was not -- fixed with the same clastfail/canyup attribution (five named conjuncts) and riding the queued daemon build. DEPLOY-LANE DEFECT FILED on the way: the first two PROMOTED receipts changed nothing because hc_deploy_one resolves stages beside the LIVE path only -- see the control-plane debt row of 2026-08-30. The identical transformation was applied to the laptop tree and landed on EXACTLY 7632 bytes, byte-count-equal to the NAS artifact, then compiled with the sovereign toolchain: nx_gen_orchestrator_daemon built clean, asm=1451202B, elf 306,603 B, no gcc. CONTROL RUN FIRST AND IT DISCRIMINATED: the pre-fix 6396 B source produced the SAME unrelated error at line 71 with gw_read_full ABSENT from the symbol table, while the fixed source produced it at line 88 with gw_read_full PRESENT -- a 17-line shift exactly equal to the inserted helper. That unrelated error was a LOCAL tree defect, nx_netprobe_lib.nx truncated to 0 bytes on the laptop while the NAS holds it at 2528 B; repairing it from the NAS copy made the build succeed. So the compile risk of this edit is retired: it is not merely written, it is proven to assemble and link. go_worker_dispatch, go_worker_img2img and go_chat_dispatch shared one copy-pasted read loop whose `if r <= 0` collapsed a TIMEOUT and a clean EOF onto one branch, so a 180s timeout that read nothing returned 0, every caller tested only >0, the candidate loop exhausted, and the daemon reported worker ABSENCE. MEASURED: a socket watch showed the worker ESTABLISHED and held 164s while the caller was told no worker was available. Fix EXTRACTED to one gw_read_full rather than re-typed at three sites, returning 0-3 busy-timeout and 0-4 partial-read alongside the existing 0-1 socket and 0-2 connect, preserving the >0 success contract. The orchestrator now names BUSY-TIMEOUT, PARTIAL-READ, CONNECT-REFUSED or no-candidate-passed-its-probe, and only says all-down when anyup==0 -- the old message was false on its own control flow, since that path is reachable only AFTER go_worker_up returned 1. DONE-RULE: a planted slow-worker fixture must produce BUSY-TIMEOUT and must NOT produce worker-absence.
Matrix cell	GEN: a refusal that names the failing conjunct -- absence, connection reset and busy-timeout are distinct outcomes · `runtime/_hdl_build/nx_gen_worker.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_gen_worker`; declare `nx_gen_worker TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-12 requirement 2026-09-12 recovery and new-domain pilot: reconcile Elder AI and Companion historical user workflows against current native runtime, API/MCP bindings and served outcomes. Shared scope is retained in media.plan migration-recovery pilot; analyst/business case in finance.plan and projectmgmt.plan. Verified source roots C:/Users/elder/elder-ai-platform and C:/Users/elder/companion are reference evidence, not automatic code adoption. Old availability claims and stale answer prose must be reconciled with current observations before any gap/coverage score. Preserve working behavior and identify unwired or regressed functions with reproductions. Same native research/inventory/rank/dispatch/verify/emit workflow must later qualify a new-domain target. Beach/swarm retain execution priority.

G18 Async render dispatch so no render is ever dropped · cell PRESENT · gate NO-GATE-DECLARED

Contract	`go_job_submit` · executor Organ · 2 u · stands on G17
Definition of done	BUILT AND DEPLOYED 2026-08-30 in the same genorch deploy as G17 (DEPLOYED-GREEN listen=VERIFIED, live sha ecee75fb): POST /api/generate_async and /api/img2img_async are live routes. UPDATE 2026-08-30: CHAT rides the async lane too (POST /api/chat_async; GO_JOB_CHAT reuses go_job_submit with the handler unchanged) and the emitted UI polls it -- the measured 17.6 s completion against the 15 s edge window was the whole of 'chat does not work'; Studio already used the batch lane. DONE-RULE MET 2026-08-30 -- nx_gen_dispatch_gate T3 (an answering worker held 6 s behind the submit round-trip) reads 25/25 GREEN against the LIVE daemon a36b5d27, after reading RED 23/25 against 5ef619e9 on exactly the two async teeth: the forked job child inherited the client socket, so the submit reply never reached EOF until the job finished; the child now closes it first. The bite is the live history, both directions. go_job_submit is EXTRACTED from go_handle_batch, which was the daemon's only async route, and go_handle_batch now CALLS it so there is one implementation that cannot drift. The mechanism cost nothing because it already existed: every handler writes its complete HTTP response to cfd, so a forked child opening the job FILE and passing that fd AS cfd makes any synchronous handler async with ZERO changes to the handler. Routes POST /api/generate_async and POST /api/img2img_async added AHEAD of their sync counterparts, because go_starts is a PREFIX match and the specific route is unreachable otherwise -- that ordering is load-bearing and is asserted in the patch. Sync routes unchanged for back-compat and the job reply is additive, so nothing that works today stops working. Compile-verified on the laptop sovereign toolchain before landing: asm 1451202 -> 1454989 B, elf 306,603 -> 307,396 B across the two edits. DONE-RULE: a render whose duration exceeds the dispatch budget must still land a cid, proven by a planted slow-worker fixture.
Matrix cell	GEN: async render dispatch -- every render returns a job id and no render is dropped on a slow worker · `runtime/_hdl_build/nx_gen_orchestrator.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_gen_orchestrator`; declare `nx_gen_orchestrator TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

G19 The /gen page is emitted, never handcrafted · cell PRESENT · gate NO-GATE-DECLARED

Contract	`gu_emit` · executor Organ · 1 u · stands on nothing
Definition of done	nx_gen_ui_emit emits elara_ui.html from knowledge/gen_ui.conf: two emits byte-identical; every behaviour marker present; a conf missing a key refuses by name with exit 3 and writes nothing; the bytes the daemon serves equal the emitted bytes (nx_gen_ui_gate, 13 teeth). Any UI change is a conf row or an emitter change shipped through the gate -- never an edit to the served file. STATUS 2026-08-30: DONE-RULE MET -- nx_gen_ui_gate GREEN 13/13 on the NAS including served-page-is-the-emitted-page; emitter and gate promoted and registered; the served page is the emitted page.
Matrix cell	GEN: the /gen companion page is EMITTED from data, never hand-authored -- strings, defaults, endpoints and theme tokens are conf rows; served bytes must equal emitted bytes · `runtime/nx_gen_ui_emit.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_gen_ui_emit`; declare `nx_gen_ui_emit TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-07 land Client lifecycle emitter promoted after native build4838 and nine native configuration cases: valid and escaped labels emit, seven malformed inputs refuse with no output. Actual emitted JavaScript passed15 lifecycle checks in a simulated DOM: hidden pages abort owned reads and pause state polling; restored pages resume one poll owner and retain accepted job identities; in-memory chat uses the existing retention policy. Emitted HTML23266 bytes SHA2f09249851e09de6107d40c147a11eb4f9d7f231e88b1af9c81180993a9c172b. Release navigation points to this existing ledger. Contentdiff5259 remains RED for reviewed intentional replacements of old polling and storage functions. UI publication and authenticated HTTP verification are pending at this record; browser visual/heap checks and backend model TTL RAM/VRAM reclamation remain open. No SOTA or memory-release qualification claimed. 2026-09-08 land Client lifecycle UI delivered through the canonical emitter7084 and authenticated origin verification7093: HTTP200,23266-byte HTML SHA2f09249851e09de6107d40c147a11eb4f9d7f231e88b1af9c81180993a9c172b exactly matches the tested artifact and served file. Visible release-summary and worklog links point to this domain. Native configuration cases9/9 and actual emitted JavaScript lifecycle checks15/15 passed. Client-owned reads and state polling pause while hidden; accepted backend jobs remain owned by the server. Browser visual/heap acceptance, service-worker cancellation integration and actual backend RAM/VRAM unloading remain unverified. This is origin delivery proof, not WAN/CDN or SOTA qualification. 2026-10-08 land album clock fixed at the root: go_album_append now stamps sys_now_realtime_sec(); daemon rebuilt over /api/build (sha a5638247, 549,248 B, contentdiff lost 6 runs all shared-library messages, 44 source paths gained), pre-change binary banked by hash at /volume1/ai/gen/banked_genorch_pre_album_c41a6886.elf, deployed via gdeploy@genorch (the first deploy promoted a no-op because the row stages beside the live path -- copied the stage beside live and deployed again; live hash re-read a5638247). The three historical rows were recovered exactly from the monotonic offsets (uptime 70.5 days exceeds every row, so no reboot intervened): elara_album.tsv rewritten 234 -> 243 B with the original banked as elara_album.tsv.pre-clockfix-20261008

G28 Broken-parts census of /gen as the browser sees it, on a beat · cell PRESENT · gate NO-GATE-DECLARED

Contract	`gh_album_years` · executor Organ · 0.5 u · stands on nothing
Definition of done	nx_galx_health gen probes the orchestrator on loopback every beat -- /health, every /api/recent image fetched and sniffed as an image, the album day labels -- writes gen.rows and gen.census under knowledge/status/galxhealth and the gen.stamp gauge this page renders; the gate drives the classifiers on planted fixtures; a leg that did not run reads BLIND, never zero
Matrix cell	OPS: broken-parts census of /gen on a beat -- health, every recent image fetched and sniffed, album day labels (G28) · `runtime/nx_galx_health_lib.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_galx_health_lib`; declare `nx_galx_health_lib TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-10-08 land nx_galx_health promoted (sha c80e161e, registered, offc twin) and its gate nx_galx_health_gate 52 of 52 GREEN on the NAS, bite-proven (killed=1 not_reached=0); clock rows galxhealth (census) and galxhealthgen (gen) bound at 21600 s; gauge rows on media.gauge and gen.gauge 2026-10-08 measure first /gen reading 2026-10-08 on loopback :18795: health ok, recent_cids=48 img_probed=48 img_ok=48 img_fail=0, album_days=2 album_days_before_floor=2 -- both days filed under 1970 because go_album_append stamped rows with sys_now_us (CLOCK_MONOTONIC, seconds since boot), not the wall clock

dossiers=20 cells_joined=20 no_matrix_row=0 (sum 20) gates_recorded=0 gates_unrecorded=0 no_gate_declared=20 (sum 20 of cells_joined) evgates=UNSTAMPED

Capability comparisons

Compare the field, search individual capabilities and open their source and adoption evidence. Documented presence does not establish comparative quality.

Position map — centrality and distinctiveness

The four-quadrant map the field uses for brand strategy (Dawar and Bagga, HBR June 2015), re-derived from this matrix on every publish. Centrality is the share of the category's feature mass a player covers, each feature weighted by how many hold it; distinctiveness is the average lead over each rival on the rows the player holds; breadth is the depth-weighted share of the whole matrix (the bubble); depth is how deeply the rows held are held; momentum is the day-over-day move off the spine (green rising, red falling, grey until day two); the dashed path runs first day → previous day → today. Dividers are the category means. Axes are fitted to the field of play, so read the tick numerals, not the frame. Rival marks are documented presence, so a rival's position reads the record, never its quality. The picture grades its own readability below; the table beside it is the same data for a screen reader or a second method.

Rows. 54 feature rows mapped of 62 matrix rows; 8 substrate rows (kernels, bounds and instruments the plan's rungrole marks substrate) are the estate's scaffolding, not category features, and stay off the map

Views. 2D cut: centrality, distinctiveness · 3D cube: centrality, distinctiveness, breadth · axes registered: centrality, distinctiveness, breadth, depth, momentum · a board picks its own in knowledge/compare/gen.cdmap (cut|x|y, cube|x|y|z); absent = the HBR defaults

bubble area = breadth · ring = momentum (green rising, red falling, grey until day two) · dashed = category means · axes fitted to the field of play: centrality 500–1000, distinctiveness 100–600 of 0–1000 permil (the full range put every player in one corner)
readability of the 2D cut, self-graded by the layout ruler: label overlaps 0 · labels over marks 0 · off-canvas 0 · unresolved labels 0 · mark overlaps 1 (a fact of the data: two players that close are that close) · data spread 595 permil of the plot · quadrant words unseated 0
text contrast, measured with wcag2-ratio (floors from contrast.conf), light theme: labels 16.24 (floor 4.50) · notes 16.24 (floor 4.50) · quadrant words 5.89 (floor 4.50) · tick numerals 5.89 (floor 4.50) · axis titles 5.59 (floor 4.50) · dark theme: labels 13.78 (floor 4.50) · quadrant words 5.55 (floor 4.50) · axis titles 5.97 (floor 4.50) · dark classes under their floor 0 (one figure serves both themes: a dark shortfall is a token to fix, never a class to hide) · classes refused under their floor 0 (a refused class is not drawn; the scale classes are measured, never hidden) · export: SVG PNG (receipt, rendered by the estate's own rasteriser from this page)

axes fitted: centrality 500–1000 · distinctiveness 100–600 · breadth 250–600 permil · farther bubbles are painted first, nearer ones over them · the floor shadow is each bubble's (x, y) at height 0
readability of the 3D cube, self-graded by the layout ruler: label overlaps 0 · labels over marks 0 · off-canvas 0 · unresolved labels 0 · mark overlaps 8 (a fact of the data: two players that close are that close) · data spread 333 permil of the plot

Nishi ComfyUI SwarmUI Forge InvokeAI Fooocus Midjourney
10 panels over 5 registered axes, every pair once (the lower axis on x, the higher on y) · each panel fitted to its own field of play, first and last tick numerals shown · no labels in a small cut, the legend names the colours; the grade below reports the mark terms only (MM summed over the panels, spread averaged)
readability of the small multiples, self-graded by the layout ruler: label overlaps 0 · labels over marks 0 · off-canvas 0 · unresolved labels 0 · mark overlaps 10 (a fact of the data: two players that close are that close) · data spread 636 permil of the plot

Player	Quadrant	centrality	distinctiveness	breadth	depth	momentum	Rows held	Days on spine	First seen
Nishi	Unconventional	658	492	561	798	+16 since 2026-09-23	38	10	2026-09-15
ComfyUI	Mainstream	866	223	512	864	+3 since 2026-09-23	32	10	2026-09-15
SwarmUI	Aspirational	889	269	555	857	-3 since 2026-09-23	35	10	2026-09-15
Forge	Mainstream	820	168	450	839	-5 since 2026-09-23	29	10	2026-09-15
InvokeAI	Mainstream	834	214	487	877	-18 since 2026-09-23	30	10	2026-09-15
Fooocus	Mainstream	811	159	432	804	-22 since 2026-09-23	29	10	2026-09-15
Midjourney	Peripheral	635	150	308	694	+30 since 2026-09-23	24	10	2026-09-15

Day	Matrix rows reviewed	Players recorded
2026-09-15	61	7
2026-09-16	61	7
2026-09-17	61	7
2026-09-18	61	7
2026-09-19	61	7
2026-09-20	61	7
2026-09-21	61	7
2026-09-22	61	7
2026-09-23	61	7
2026-10-08	62	7

How to read this evidence. A Nishi source mark records whether the named symbol is declared in the selected source file. It does not establish working behavior or accepted delivery. A watching tag names an expected source entry; discovering it updates source coverage only. A dark tag means the file exists but the named symbol was not found: inspect the source and contract before assuming the capability is missing or renaming code. Peer codes are authored assessments; unassessed means no comparison is established, not absence. Qualification, deployment identity, runtime observations and acceptance must carry their own evidence. Adoption records the catalog's observed path through source, build, promotion, registration and invocation; libraries use importer evidence and gates use declared execution surfaces. A full catalog path does not establish quality, freshness or a successful user outcome. Partial and unknown states retain their stated next action. Census stamps: importers asof 1790637268, gate census asof 1790104131 (unix seconds; -1 = census absent).

Capability matrix — measured against source

◉ authored stronger claim; source present● present◐ partial○ absent? unassessed; excluded from relative comparison · click any capability for its evidence

Peer cell coverage: authored assessed 372, unassessed 0. Authored assessments are not verified outcomes.

Capability	Nishi	ComfyUI	SwarmUI	Forge	InvokeAI	Fooocus	Midjourney
GEN
text-to-image on the sovereign GPU worker -- OpenAI-shape dispatch, seed-deterministic Measured: `go_worker_dispatch` exists in `runtime/_hdl_build/nx_gen_worker.nx`, verified at emit. Every local UI renders on the user's own GPU [comfyui] [swarmui] [forge] [invokeai] [fooocus]; Midjourney is cloud-only [midjourney]. Ours dispatches to elder-sdcpp (a stable-diffusion.cpp fork [sdcpp]) serving Z-Image-Turbo, an 8-NFE distilled DiT that fits 16 GB [zimage]; same prompt plus seed = same image. Adoption: `LIB-WIRED importers=2 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), InvokeAI (leads), Fooocus (leads)	●	◉	◉	◉	◉	◉	●
async batch job -- submit returns a job id, the browser polls, any batch size survives the 15 s edge timeout Measured: `go_handle_batch` exists in `runtime/_hdl_build/nx_gen_orchestrator.nx`, verified at emit. ComfyUI queues asynchronously with partial graph re-execution [comfyui]; SwarmUI grids fan a batch across a swarm of GPUs [swarmui]. Ours forks a detached child per job and writes batchjob-id.json; the batch BOARD (named jobs, per-image progress, stop and resume) is the watch row below. Adoption: `LIB-WIRED importers=6 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind ComfyUI (leads), SwarmUI (leads), InvokeAI (leads)	●	◉	◉	●	◉	●	●
async render dispatch -- every render returns a job id and no render is dropped on a slow worker Measured: `go_job_submit` exists in `runtime/_hdl_build/nx_gen_orchestrator.nx`, verified at emit. SHIPPED 2026-08-30: source landed, built over /api/build and DEPLOYED via gdeploy@genorch, DEPLOYED-GREEN listen=VERIFIED, live sha ecee75fb 301306 B -- POST /api/generate_async and /api/img2img_async return a job id immediately and reuse the batch machinery. Upstream stable-diffusion.cpp's own server exposes an async job API -- 202 with a poll URL, 429 when the queue is full, cancel for queued and generating jobs [sdcppasync] -- while our fork is fully synchronous on one mutex, so this daemon-side submit is the analogue until the engine gains a queue. DONE-RULE MET 2026-08-30, BITE-PROVEN ON THE LIVE HISTORY: nx_gen_dispatch_gate T3 plants a worker that ANSWERS only after a 6 s delay and requires the submit to return a job id BEFORE that delay can elapse, a poll to read pending first, and the planted reply to land through /api/batch no earlier than the delay. Against daemon 5ef619e9 it went RED 23/25 on exactly the two async teeth -- the forked job child INHERITED the client socket, so the parent's already-written reply never reached EOF until the job finished and any read-to-close client saw an async submit that lasted the whole render (browsers were spared only by Content-Length; the /gen gateway reads the backend to EOF). One line (the child closes the client socket first), daemon a36b5d27 DEPLOYED-GREEN listen=VERIFIED, and the same gate reads 25/25 GREEN. The emitted UI polls the async lane for chat and regenerate (single-render Studio submits ride the batch lane it always used). The gate runs about four minutes (three chat tries at the conf budget) and is run by nx_job_run, deliberately not listed in gen.gates until the evidence runner's budget is declared. (Note history: this row once read G17 by mistake -- two rows were claiming one rung.) MEASURED 2026-08-29: the single-photo and chat-photo paths block on one socket and are abandoned at a 180 s timeout while the worker is still rendering -- a socket watch showed the connection ESTABLISHED and held for 164 s from the NAS while the caller was told no worker was available. Only the batch route forks a child today, and that machinery generalises for free because it already hands the child a file descriptor as its response socket. Every rival has had a real queue for years and Midjourney is fully async, so this is PARITY and not an exceed; the throughput is unchanged because the engine is one mutex with no queue and no cancel -- what changes is that a render is never lost and a failure is legible. The c3 code is read from the SUPERSEDED Forge mirror and is UNPROVEN like the rest of that column. Adoption: `LIB-WIRED importers=6 nonval=1` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind ComfyUI (leads), SwarmUI (leads), InvokeAI (leads), Midjourney (leads)	●	◉	◉	●	◉	●	◉
the /gen companion page is EMITTED from data, never hand-authored -- strings, defaults, endpoints and theme tokens are conf rows; served bytes must equal emitted bytes Measured: `gu_emit` exists in `runtime/nx_gen_ui_emit.nx`, verified at emit. G19, added 2026-08-30 (operator: emit every UI from the first byte up; /gen must not be handcrafted). The served page had been a hand-authored file since seq264. nx_gen_ui_emit reads knowledge/gen_ui.conf (52 required keys; a missing key REFUSES by name and writes nothing) and emits the page with the NX-DERIVED stamp. Proven on the sovereign toolchain: two emits byte-identical, every behaviour marker present (chat post, async studio batch, error surfacing, three tabs, family-gallery link), and the missing-key neg-control refuses with exit 3 and no file. THE FLIP IS THE RECEIPT, NOT THE PROOF: nx_gen_ui_gate (13 teeth including served-page-is-the-emitted-page) is written and compiles; PROVEN ON THE BOX 2026-08-30: nx_gen_ui_emit and nx_gen_ui_gate built, promoted and registered on the NAS, and nx_gen_ui_gate ran GREEN 13/13 via nx_job_run INCLUDING served-page-is-the-emitted-page -- the daemon serves exactly what the emitter emits from the conf. ALBUM 2026-08-31: the Gallery tab (renamed Photos) is the conversation album by day -- the page fetches GET /api/album (days newest-first from elara_album.tsv under the operator clock offset in gen_day_offset.conf) and draws day headers, while /gallery stays the ONE full gallery; emitter sha ed1ace73 promoted, the page re-emitted on the box and nx_gen_ui_gate re-ran GREEN 14/14 including the new gallery-tab-is-the-conversation-album-by-day tooth. Rival codes are UNREAD: no rival's UI provenance has been measured. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). dossiernot adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked) Pros Implementing source symbol found; runtime acceptance separate; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, MidjourneyCons not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	●	○	○	○	○	○	○
a refusal that names the failing conjunct -- absence, connection reset and busy-timeout are distinct outcomes Measured: `gw_read_full` exists in `runtime/_hdl_build/nx_gen_worker.nx`, verified at emit. Watch G17 -- SYMBOL CORRECTED 2026-08-30 AND THE SOURCE HAS LANDED. The contract was written as _ABSENT_:gw_outcome_name, which is ABSENT-PROVEN (matches=0 over 23,354 files, corpus_complete=1) and could NEVER have flipped, while the symbol that actually shipped went unwatched -- the board was structurally unable to report its own completed work. gw_read_full is now DEFINED at nx_gen_worker.nx:29 and CALLED at :55, :79 and :104. ⚠THE FLIP IS THE RECEIPT, NOT THE PROOF: the source contract is met, the DONE-RULE IS NOT -- DONE-RULE MET 2026-08-30: nx_gen_dispatch_gate GREEN 15/15 against the LIVE daemon (a silent worker that listens and never accepts reads BUSY-TIMEOUT; an absent worker reads the health-probe refusal within seconds -- the isolation tooth that proves no real worker was dispatched, added after the first run leaked to production through the SSOT's absolute fallback; neither reads registry-down); the CHAT path was found still carrying the registry-down lie the same day and was cured with the same five named conjuncts; BUILD+DEPLOY LANDED 2026-08-30 over the API (DEPLOYED-GREEN listen=VERIFIED, live sha ecee75fb 301306 B), so the live daemon's refusal now names the failing conjunct -- the retired false message (no gpu worker available) is the ONE string the promote ruler counted as lost, deliberately. ⚠NAME HAZARD: gw_read_full is a FIVE-WAY duplicate across the estate (nx_mesh_gateway:177, nx_gen_gateway:194, nx_vault_gateway:107, nx_gallery_gateway:178, nx_gen_worker:29) carrying TWO different signatures -- (cfd,req,cap) in the four gateways against (fd,respbuf,respcap) here. The generator scopes its measurement to the DECLARED organ so this cell is unambiguous, but nobody may read "gw_read_full exists" as a single estate-wide fact. MEASURED 2026-08-29 by reading the source: the dispatch read loop collapses a timeout and a clean end-of-file onto one branch, so a 180 s timeout that read nothing returns zero bytes; the caller tests only greater-than-zero, the candidate loop advances, and the daemon emits no gpu worker available with the registry named as the cause. FOUR distinguishable states already exist and are destroyed at the call site -- probe-down, socket or connect failure, connected-and-rendering, and a partial read then timeout, which today returns a positive byte count, reads as SUCCESS and resurfaces later as worker returned no images. The return space was already three-valued and the caller tested one boolean. EXCEED DELIBERATELY NOT CLAIMED: no rival refusal text has been measured, and an exceed asserted against an unmeasured comparison is the defect this board exists to prevent, so the zeros here mean UNREAD and not a measured No. Adoption: `LIB-WIRED importers=2 nonval=1` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	○	○	○
img2img on an existing gallery cid (sketch-to-art, refine) Measured: `go_handle_img2img` exists in `runtime/_hdl_build/nx_gen_orchestrator.nx`, verified at emit. All six rivals refine from an image; ours takes a gallery cid and re-ingests the result under a new cid. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=6 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), InvokeAI (leads), Fooocus (leads), Midjourney (leads)	●	◉	◉	◉	◉	◉	◉
the engine's prompt-embedded control block -- steps, cfg, size and seed ride INSIDE the prompt Measured: `go_append_turbo_args` exists in `runtime/_hdl_build/nx_gen_orchestrator.nx`, verified at emit. Midjourney's parameters also ride in the prompt text [midjourney]; the local UIs expose the same knobs as form fields. Ours measured the knob ARRIVES: 8 steps 12.0 s, 30 steps 25.0 s at 512x512 on the 5080 through the live daemon (gen_steps.conf, 2026-08-04). Adoption: `LIB-WIRED importers=6 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	●	●	●	●	●	●
default step count as DATA, re-read per request (gen_steps.conf carries the measured latency curve) Measured: `go_default_steps` exists in `runtime/_hdl_build/nx_gen_orchestrator.nx`, verified at emit. Rivals persist UI settings; ours re-reads a conf per request with the measurement written beside the number, so the quality-vs-latency knob is an operator edit with no rebuild. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=6 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	●	●	●	●	●	○
PROMPT
realism trigger prefix as data (elara_trigger.txt) -- the June slot order restored, trigger leads the prompt Measured: `go_load_trigger` exists in `runtime/_hdl_build/nx_companion_identity.nx`, verified at emit. LANDED 2026-08-27. The pre-docker gen-img composed every render as lora-tags, trigger words, prompt (svc-config default_loras row of 2026-04-25) and the sovereign path had dropped the trigger; the 2026-08-19 June-repro A/B measured it as the dominant term of the June look (dhash 74 to 48). Fooocus ships style presets with activation fragments [fooocus]; in the other UIs LoRA trigger words are typed by hand (Part). Adoption: `LIB-WIRED importers=3 nonval=2` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	◐	◐	◐	◐	●	○
full prompt-grammar composer over the mined vocabulary planes (the 1,640-row content.db port) Open — watching `runtime/nx_photo_grammar.nx` : `pg_compose`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Watch G2: the old platform's Director's Shot-Sheet grammar (zimage reinforcements, microdetails, lighting, expressions) as sovereign planes plus a slot-ordered composer; Fooocus's preset library is the field's nearest analog [fooocus]. The ranker flagged this rung UNMAPPED on 2026-08-27 -- this row is the loop's own fix. dossierwatching pg_compose Pros none measured yetCons behind ComfyUI, SwarmUI, Forge, InvokeAI (leads), Fooocus, Midjourney; open contract; expected source declaration not established	○	●	●	●	◉	●	●
ONE photo-prompt composer shared by the daemon and its GPU-free probe (trigger, camera, body) Source present; authored stronger claim: `go_compose_photo_prompt` in `runtime/_hdl_build/nx_companion_wardrobe.nx`, verified at emit. EXCEED with the why: the slot order exists in exactly one function and nx_elara_prompt_probe proves it without a GPU (T1 trigger-then-camera, T7 non-vacuity). No rival ships a testable prompt composer; theirs is whatever the user typed. Adoption: `LIB-WIRED importers=3 nonval=1` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	○
data-driven companion identity anchor, front-loaded for the DiT's RoPE (elara_appearance.txt) Source present; authored stronger claim: `go_load_appearance` in `runtime/_hdl_build/nx_companion_identity.nx`, verified at emit. EXCEED with the why: a persistent first-class identity object per companion, editable as data with no rebuild, Z-Image-counterbalanced (age and ethnicity drift measured and pinned). Midjourney's character reference is image-based consistency, not an editable genome (Part) [midjourney]. Adoption: `LIB-WIRED importers=3 nonval=2` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	◐
occlusion-aware wardrobe compiler -- per-zone clothed-or-bare with a positive fabric assertion Measured: `go_build_photo_prompt` exists in `runtime/_hdl_build/nx_companion_wardrobe.nx`, verified at emit. Exists and is gated (T6 and T7), but the MEASURED wall stands: on a cfg-distilled NSFW-prior base with no negative prompt, clothing control is unreliable. Not claimed as an exceed until the negative-capable path (watch row below) lands and 20 clothed renders measure clothed. Adoption: `LIB-WIRED importers=3 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	○	○	○
camera and photographic style variety rotated by seed (elara_shots.txt) Measured: `go_pick_shot` exists in `runtime/_hdl_build/nx_companion_identity.nx`, verified at emit. Fooocus styles are the richest preset library in the field [fooocus]; Midjourney has stylize and style parameters [midjourney]; ComfyUI reaches variety through nodes [comfyui]. Adoption: `LIB-WIRED importers=3 nonval=2` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind Fooocus (leads)	●	●	●	●	●	◉	●
quality tail measured against the 568,944-row beauty corpus, kept as data (elara_quality.txt) Source present; authored stronger claim: `go_load_quality` in `runtime/_hdl_build/nx_companion_identity.nx`, verified at emit. EXCEED with the why: every term in the tail was tested for lift in the post-repair window of the old platform's scorer (a term present in 98 percent of a corpus cannot be measured by presence, so the baseline vocabulary is kept as baseline, not claimed as lift). Fooocus ships hand-curated quality presets [fooocus]. Adoption: `LIB-WIRED importers=3 nonval=2` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	●	○
realism LoRA stack as data, with a CONTAMINATION SCREEN before adoption Measured: `go_load_loras` exists in `runtime/_hdl_build/nx_companion_identity.nx`, verified at emit. Every local UI loads LoRAs [fooocus]; Midjourney has no LoRA surface. Ours reads elara_loras.txt per request. 🔴CORRECTED 2026-08-30 BY MEASUREMENT, AND THE EARLIER READING WAS BACKWARDS: the recovered platform registry lists ten LoRAs at is_available 0 NOT YET TRAINED, which reads as an empty lane -- but those are NOT the artifacts on disk. Ten DIFFERENT real LoRA files sit in models/unified/loras and TWO WERE LIVE IN elara_loras.txt, so the registry did not know what the stack was actually loading. Measured at the true June factors, seed 1480332350, 960x1088 and 960x1280, only the LoRA stack varying: the LIVE pair (RealisticSnapshot 0.7 + amateur_photography 0.45) produced a CATASTROPHIC render -- identity lost to brunette, a black artifact across the face, melted anatomy, garbled scene text, and an apparent age contradicting the appearance file's own declared twenty-one. The single-LoRA ladder had already indicted the same artifact: amateur_photography_zimage_v1 is the ONE arm that produced ELF EARS at cfg 1.0, where the zero-LoRA baseline never does. ultra_realist_beta2 0.6 ALONE measured best of four arms and RealisticSnapshot 0.7 alone measured clean. ✅FIXED IN DATA, NO REBUILD: elara_loras.txt rewritten to the single clean LoRA. ⇒ THE REUSABLE CAPABILITY IS THE SCREEN, NOT THE STACK -- a contaminated artifact announces itself in ONE render with a visible binary tell, so every acquired LoRA or checkpoint gets a one-render screen before adoption, which is exactly what the model-acquisition rung needs and what the registry could not tell us. ⚠n=1 per arm; the fix is justified by the size of the defect, not by a distribution. Adoption: `LIB-WIRED importers=3 nonval=2` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Midjourney; fully adopted on the estate ladderCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), InvokeAI (leads), Fooocus (leads)	●	◉	◉	◉	◉	◉	○
QC
contamination screen on every acquired LoRA and checkpoint BEFORE it may enter the served path Measured: `lc_screen` exists in `runtime/_hdl_build/nx_lora_screen_lib.nx`, verified at emit. STATIC HALF LANDED 2026-08-30 (nx_lora_screen.elf PROMOTED+REGISTERED; nx_lora_screen_gate 12/12 GREEN with three planted-bad fixtures biting and a well-formed control silent): the 8-byte header length and every declared data_offsets extent are checked against the file's bytes, so a truncated, forged or payload-carrying artifact is FLAGGED by name with its numbers, and .ckpt/.pt/.pth/.bin are REFUSED outright (CWE-502, pickled code executes on load). FIRST LIVE RESULT, and it indicts the file adopted as clean that same morning: ultra_realist_beta2.safetensors declares 170,065,920 bytes of tensor data in a file holding 89,377,936 (FLAG DECLARED-EXCEEDS-FILE) while amateur_photography_zimage_v1 and z_real_v1 both screen CLEAN with sum, max-end and region equal to the byte -- the LoRA serving today is a truncated or mis-headed acquisition and must be re-pulled hash-verified through mg_fetch before any further render claim rests on it. The RENDER half of the done-rule (one fixed-seed screening image per candidate, adjudicated before adoption) is still owed; this organ is its admission precondition. Watch G16, added 2026-08-30 because the rung existed in gen.plan with NO watch cell at all and was therefore invisible to both the generator and the ranker. FOUND BY MEASUREMENT, NOT PROPOSED: amateur_photography_zimage_v1 was LIVE in elara_loras.txt and is the ONE arm of a four-arm single-LoRA ladder that produced ELF EARS AT CFG 1.0, where the zero-LoRA baseline never does; in the live two-LoRA pair at 960x1280 it produced a catastrophic render -- identity lost to brunette, a black artifact across the face, melted anatomy, garbled scene text, and an apparent age contradicting the appearance file's own declared twenty-one. Swapping it for ultra_realist_beta2 0.6 restored identity, freckles, clean hands AND made a clothed-scene instruction render exactly the garment requested, so this screen also carries partial credit against G3. ⇒THE CAPABILITY IS THAT CONTAMINATION HAS A CHEAP VISIBLE BINARY TELL: one fixed-seed render per candidate artifact, adjudicated before adoption, which is precisely the admission gate G9 needs and precisely what the platform capability registry could NOT provide -- it listed ten LoRAs as NOT YET TRAINED while ten DIFFERENT real files sat on disk and two were serving. RIVAL CODES ARE DELIBERATELY ZERO MEANING UNREAD, NOT A MEASURED NO: no rival's artifact-admission behaviour has been measured, and an exceed asserted against an unmeasured comparison is the defect this board exists to prevent, so nishiExceed stays 0 until it is. Adoption: `LIB-WIRED importers=2 nonval=1` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	○	○	○
pre-spend anatomy-risk score of the composed prompt (no GPU, factors named) Source present; authored stronger claim: `go_anatomy_risk` in `runtime/_hdl_build/nx_anatomy_risk.nx`, verified at emit. EXCEED with the why: a mechanical refusal-with-remedy BEFORE GPU spend; the daemon appends risk, factors and scene to anatomy_audit.tsv so the rate is trackable. Nothing in the field surfaces a pre-spend verdict. Adoption: `LIB-WIRED importers=4 nonval=4` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	○
coherence repair of incoherent scene words before render (opt-in, flag-only by default) Source present; authored stronger claim: `go_coherence_repair` in `runtime/_hdl_build/nx_render_coherence.nx`, verified at emit. EXCEED with the why: the render-bug register's top class (wet-without-context, 40 percent of bugs) is rewritten to a coherent descriptor; opt-in so the user's exact words are never silently overridden. Adoption: `LIB-WIRED importers=2 nonval=2` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	○
GPU-free composition probe -- eight teeth including a non-vacuity tooth on the opposite wardrobe branch Source present; authored stronger claim: `compose` in `runtime/_hdl_build/nx_elara_prompt_probe.nx`, verified at emit. EXCEED with the why: the composer is proven against the REAL data files in seconds; a prompt regression is caught before a single render. Re-based on the shared composer 2026-08-27 so it can no longer pass on an order the daemon does not use. Adoption: `BUILT-UNPROMOTED` — PARTIAL: compiled, never promoted to the serving root: /api/promote it. not adopted: compiled, never promoted to the serving root: /api/promote it Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, MidjourneyCons not adopted yet: compiled, never promoted to the serving root: /api/promote it	◉	○	○	○	○	○	○
COMPANION
chat and photo-on-request through one daemon -- the Qwen3 text encoder doubles as the companion LLM Source present; authored stronger claim: `go_handle_chat` in `runtime/_hdl_build/nx_gen_orchestrator.nx`, verified at emit. EXCEED with the why: dual use of the DiT's own text encoder [zimage] on one 16 GB card, so talk and render share a process; server-authoritative persona and memory ride the same route. No image UI in the field is a companion. ASYNC 2026-08-30: a completion measured 17.6 s on the shared card against a 15 s edge window, so chat now rides the same job lane as renders (POST /api/chat_async then poll /api/batch/<job>; the emitted UI polls it) and the per-try dispatch budget is DATA (knowledge/gen_chat_budget.conf, 60 s) instead of a 25 s literal. Adoption: `LIB-WIRED importers=6 nonval=1` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	○
a NUMERIC affective vector compiled into the diffusion prompt (affection, dopamine, cortisol and oxytocin to expression tokens) Source present; authored stronger claim: `go_state_update` in `runtime/_hdl_build/nx_companion_state.nx`, verified at emit. EXCEED NARROWED 2026-08-29 to the AFFECT half only: per-turn state persists across sessions and compiles as a continuous numeric vector into expression tokens on the photo prompt. The OUTFIT half is RETRACTED -- shipping companion front-ends already inject wardrobe into an image prompt (the SillyTavern outfit extensions and the SLAYimages class), so wardrobe injection is not ours. What survives attack is the numeric vector, since the nearest shipping prior art collapses affect to one label from a fixed emotion set to select a sprite. SCOPE STATED: this exceed is measured against the COMPANION field, which none of this board's six rival columns represent, and the mirrors that would ground it are OWED -- see the gen-state-exceed-ungrounded debt row. Adoption: `LIB-WIRED importers=1 nonval=1` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	○
character-card compiled persona SSOT (elara_card.json to persona, age fail-closed) Measured: `ccm_json_str` exists in `runtime/_hdl_build/nx_card_compile.nx`, verified at emit. The card is the byte-one persona source; the compiler refuses an under-18 card. Product-adjacent to the gen lane; no rival column has the concept. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked) Pros Implementing source symbol found; runtime acceptance separate; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, MidjourneyCons not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	●	○	○	○	○	○	○
ASSET
GENREC provenance written INTO the PNG (prompt, seed, model, steps, size, host as tEXt) Measured: `png_insert_genrec` exists in `runtime/_hdl_build/nx_png_textw.nx`, verified at emit. ComfyUI embeds the whole workflow in the PNG and reloads it [comfyui]; Forge, InvokeAI and Fooocus embed parameters [invokeai] [fooocus]; Midjourney images carry none. Ours is a sovereign PNG tEXt writer (CRC32) with the GENREC keys the cid canon hashes. Adoption: `LIB-WIRED importers=4 nonval=2` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Midjourney; fully adopted on the estate ladderCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), InvokeAI (leads), Fooocus (leads)	●	◉	◉	◉	◉	◉	○
content-addressed cid = hash of the GENREC factors, so a render is named by what produced it Source present; authored stronger claim: `nx_store_ingest_compute_cid` in `runtime/_hdl_build/nx_store_ingest.nx`, verified at emit. EXCEED with the why: the render is named by its generation factors rather than by a counter and a timestamp as every rival does, and nx_gen_cid_gate referees the canon. HONEST LIMIT measured 2026-08-29 and not yet fixed: the canon carries seed, model, sampler, steps, host, prompt, size and args, but the harvest silently truncates prompt at 2047 bytes and args at 4095, so two prompts differing only past the cap hash IDENTICALLY and the second ingest TRUNCATES the earlier blob. The composed chat prompt is already 1371 bytes and the grammar port adds pose, props, lighting and two CJK blocks, so G2 re-arms exactly the data-loss class this gate exists to prevent. Scheduler and flow shift are absent from the canon because they are never SENT, which is the same defect from the other side. 🔴AND THE REFEREE IS LIVE ORDNANCE: nx_gen_cid_gate is PROMOTED 156,922 B against BUILT 61,985 B -- the AHEAD class, where the served binary carries 94,937 bytes the current source cannot rebuild, so a naive rebuild-and-promote DESTROYS capability. Four separate designs across four boards each proposed exactly that. Bank the live artifact by hash to a durable path before any rebuild, and note nx_contentdiff compares STRING RUNS so a code-only delta is invisible to it -- when size moves the wrong way the SIZE is the evidence and the ruler is the one that is silent. The absolute claim that two prompts at one seed can never collide is WITHDRAWN until the truncation is fixed. Adoption: `LIB-WIRED importers=11 nonval=6` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	○
ingest pipeline -- decode, inject GENREC, cid, blob, store, sidecar (idempotent, both response shapes) Measured: `gp_response` exists in `runtime/_hdl_build/nx_gen_pipeline.nx`, verified at emit. All rivals save renders with metadata to an output folder or board; the family-gallery claim MEASURED FALSE 2026-08-30 (a gen cid 404d at the family gallery img route -- gen blobs live in /volume1/ai/gen and the gallery resolves only its own cid tsv) and FIXED 2026-08-31: go_gallery_feed (conf-gated, knowledge/gen_gallery_feed.conf) forks nx_galx_saveimg per render -- copy to /volume1/ai/images/gen, cid-tsv append, view prepend, LIVE with no gallery restart -- wired after the generate, chat-photo and img2img ingests; a feed failure announces on stderr and never fails the render. Deployed in daemon c41a6886 and WITNESSED END-TO-END 2026-08-31: a live render (gen cid acf4d271) landed in /volume1/ai/images/gen under its CONTENT cid 362c9c91 with a gallery cid-tsv row and a view entry written AT RENDER TIME by the daemon's feed -- proven by an idempotent manual re-save answering view_added=0 over a second identical tsv row. ⚠the first witness probe checked the GENREC-factor cid and read the feed as broken: the save-to-gallery canon names by CONTENT hash, a DIFFERENT cid namespace from the gen factors canon -- probe the subject the consumer actually writes. The unauthenticated :18090 probe 301s every request (auth wall), so the HTTP 200 leg is witnessed from the operator's session. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=4 nonval=2` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	●	●	●	●	●	●
O(1) cid-to-path index for serving (replaces a per-request flat scan) Measured: `gs_cidindex_lookup` exists in `runtime/_hdl_build/nx_galx_cid_index.nx`, verified at emit. Every rival serves its own outputs; the index is what makes ours serve 233k gallery rows without a scan. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=8 nonval=7` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	●	●	●	●	●	●
any media file becomes a first-class gallery image with no restart (the save-to-gallery primitive) Measured: `si_hex` exists in `runtime/nx_galx_saveimg.nx`, verified at emit. InvokeAI boards are the strongest asset organiser in the field [invokeai]. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked) Pros Implementing source symbol found; runtime acceptance separateCons behind InvokeAI (leads); not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	●	●	●	●	◉	●	●
JUDGE
full-population mechanical ranking of gallery renders Source present; authored stronger claim: `do_rank` in `runtime/_hdl_build/nx_gen_rank.nx`, verified at emit. EXCEED with the why: every render in the store is scoreable by one organ, whole population and never a sample. No rival ships a ranker over its own outputs. Adoption: `RUN-BY:clock` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	○
bilateral-symmetry beauty score per cid (Q1000) Source present; authored stronger claim: `go_image_symmetry` in `runtime/_hdl_build/nx_image_symmetry_lib.nx`, verified at emit. EXCEED with the why: a sovereign per-image score with a stated method; the old platform's 568k-row scorer is the calibration corpus. Adoption: `LIB-WIRED importers=3 nonval=3` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	○
mechanical character-image judge (identity axes per render) Source present; authored stronger claim: `cjc_hw` in `runtime/_hdl_build/nx_charjudge.nx`, verified at emit. EXCEED with the why: the same organ that judges the whole estate's character renders judges these; its measuring range is written in its own debt row (eye-window literals) rather than hidden. Adoption: `RUN-BY:fork:nx_charjudge_census` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	○
external referee run PUBLISHED WITH ITS STATISTICS -- HPSv2, PickScore and GenEval on our renders with n, an interval, the seed and the engine-config hash Open — watching `runtime/nx_gen_bench.nx` : `gb_bench_run`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Watch G5. CORRECTED 2026-08-29: the clause that no rival column ships a referee was FALSE, and its refutation sat inside this board's own pinned SwarmUI mirror, which states SwarmUI can automatically install improved-aesthetic-predictor and PickScore [swarmui] -- two preference scorers, auto-installed, so c2 is 2. Running the published protocols [hpsv2] [pickscore] [geneval] is therefore PARITY, not the exceed. The exceed is the STATISTICS leg -- n, a Wilson interval, the seed, the engine-config hash and a trivial baseline beside every number -- which no rival publishes and which this board's own rigor-envelope debt row already requires. dossierwatching gb_bench_run Pros none measured yetCons behind SwarmUI (leads); open contract; expected source declaration not established	○	○	◉	○	○	○	○
WORKER
non-blocking TCP health probe and failover across GPU workers (a powered-off worker never hangs a dispatch) Measured: `np_probe_up` exists in `runtime/nx_netprobe_lib.nx`, verified at emit. SwarmUI's founding feature is a swarm of GPU backends for one user [swarmui]; ComfyUI, Forge, InvokeAI and Fooocus are single-backend processes; Midjourney is a managed fleet. Adoption: `LIB-WIRED importers=3 nonval=3` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of ComfyUI, Forge, InvokeAI, Fooocus; fully adopted on the estate ladderCons behind SwarmUI (leads)	●	○	◉	○	○	○	●
role-to-address SSOT re-read per call (swarm_nodes.conf) Measured: `se_addr` exists in `runtime/nx_swarm_endpoint_lib.nx`, verified at emit. SwarmUI configures backends in its UI [swarmui]. Ours is a conf the whole swarm reads; the gen daemon's own cwd landmine (it cannot see the conf from /volume1/ai/gen and falls back to argv) is a filed defect, stated here rather than hidden. Adoption: `LIB-WIRED importers=8 nonval=7` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of ComfyUI, Forge, InvokeAI, Fooocus, Midjourney; fully adopted on the estate ladderCons behind SwarmUI (leads)	●	○	◉	○	○	○	○
VRAM-budgeted placement scoring across the swarm Measured: `sp_pick` exists in `runtime/nx_swarm_place.nx`, verified at emit. SwarmUI distributes grids across backends [swarmui]; ours scores candidates against a VRAM budget from the hub inventory. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked) Pros Implementing source symbol found; runtime acceptance separate; ahead of ComfyUI, Forge, InvokeAI, FooocusCons behind SwarmUI (leads); not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	●	○	◉	○	○	○	●
RESOURCE
GPU residency guard -- restart-resident, busy, wait and start decided natively, thresholds as data Source present; authored stronger claim: `gg_decide` in `runtime/nx_gpuguard.nx`, verified at emit. EXCEED with the why: the worker holds ZERO VRAM at idle (offload mode) and a native organ decides wedged-vs-loading from the service's OWN footprint, busy dominating every other rule; ComfyUI, Forge and Fooocus ship offload and low-VRAM modes [comfyui] [forge] [fooocus] but nothing decides for you. Cost named honestly: a cold offload render measured 167 s on 2026-08-27 while the card was shared with a game. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked) Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, MidjourneyCons not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	◉	●	○	●	○	●	○
multi-model VRAM-LRU residency policy (the Triton model-management analog) -- a gated simulator, not yet wired to the worker Measured: `mx_run` exists in `runtime/nx_mesh_mux.nx`, verified at emit. Honest: the policy is proven on a simulator with a fixture registry; the worker still serves one diffusion model plus its encoder. SwarmUI manages many models across backends [swarmui]. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked) Pros Implementing source symbol found; runtime acceptance separate; ahead of MidjourneyCons behind SwarmUI (leads); not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	●	●	◉	●	●	●	○
on-demand model swap with a TTL -- LLM and diffusion time-sharing one 16 GB card DARK — `runtime/nx_mesh_mux.nx` EXISTS but does not declare `mx_swap_ttl`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. Watch G12: the llama-swap-class capability the Elara dual-use need names; the local UIs swap checkpoints on demand, none with a residency TTL policy. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them dossierwatching mx_swap_ttl Pros none measured yetCons behind ComfyUI, SwarmUI (leads), Forge, InvokeAI, Fooocus; open contract; expected source declaration not established	○	●	◉	●	●	●	○
CONTROL
a LIVE unconditional branch -- a negative prompt that actually fires, selected by intent (the ER-8 keystone) DARK — `runtime/_hdl_build/nx_gen_worker.nx` EXISTS but does not declare `gw_guidance_select`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. Watch G3, the measured controllability wall -- MECHANISM CORRECTED and FIRST ARM REFUTED, both by control on 2026-08-29. (1) The gate is the CFG SETTING, not the model: at seed 1480332350 with zero LoRAs, cfg 1.0 with and without a negative produced the SAME image (sha256 96de208a both) while cfg 4.0 produced DIFFERENT ones (667f268b against 7b32b0ff), because uncond plus s times cond minus uncond is exactly cond at s equals 1. The turbo model DOES take a negative and it is INERT at cfg 1. (2) The cfg-ladder arm is DEAD: run at the true June factors (14 steps, 832x1088, same seed, negative present, only cfg varying), cfg 1.5 and cfg 2.5 BOTH destroy the render -- elf ears, six fused fingers, the kitchen scene replaced, waxy skin -- and at the cfg where the negative is finally active the hands are WORSE than at cfg 1.0 where it was inert. There is no usable window. (3) The real prior was found in the engine's own process arguments: the served checkpoint is zimage_turbo_nsfw_2602-Q8_0.gguf, an NSFW FINETUNE, so the prior is baked into the model file and CFG above 1 amplifies its fantasy contamination rather than the negative. RETRACTED remedies: cyberrealisticZImage_v40 is a Turbo derivative that cannot clear this rung's own accept test, and Realism_Engine_Klein_V2 measures 1.02 GB on disk against 5.7 to 6.8 GB for real checkpoints here, confirming it is a LoRA. CORRECTED 2026-08-30 BY HASH, NOT BY REREADING: a clean base checkpoint DOES exist on disk -- /volume1/ai/models/diffusion/z_image_turbo-Q8_0.gguf measures 6,577,440,704 bytes with sha256 df1c5baa86d1398c979495a6072dbcee79444fdb884a2445582ba0769c44e9a1, byte-identical to the publisher's lfs.oid in leejet/Z-Image-Turbo-GGUF (the stable-diffusion.cpp author's own quant repo, fetched over the sovereign stack and pinned in knowledge/fetched/rf_8fa9586c.raw), so the four-checkpoint census that declared this rung BLOCKED BY G9 had not looked in the diffusion store. G9 is no longer the blocker; the unrun experiment is a render ladder on the CLEAN checkpoint at the true June factors (does the negative fire at cfg 1.5 to 2.5 without the finetune's fantasy contamination), and gw_guidance_select is the verb that points the worker at it by intent. Arms in cost order now: acquire a clean Z-Image-Turbo checkpoint and select by intent, a de-distillation adapter, attention-space negative guidance, neutral-base plus edit-after, and only LAST a swap to Z-Image base, a TRADE not an upgrade since CFG doubles the forward count so 50 steps is 100 evaluations against Turbo's 8 [zimage]. Every local UI runs negatives on models that were never distilled [fooocus]. dossierwatching gw_guidance_select Pros none measured yetCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), InvokeAI (leads), Fooocus (leads), Midjourney; open contract; expected source declaration not established	○	◉	◉	◉	◉	◉	●
ControlNet and pose or structure conditioning DARK — `runtime/_hdl_build/nx_gen_worker.nx` EXISTS but does not declare `gw_controlnet`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. Watch G6. Forge integrates ControlNet natively [forge]; ComfyUI has ControlNet nodes [comfyui]; Fooocus ships ImagePrompt and structure modes [fooocus]; Midjourney's style and character references are the partial analog [midjourney]. dossierwatching gw_controlnet Pros none measured yetCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), InvokeAI (leads), Fooocus (leads), Midjourney; open contract; expected source declaration not established	○	◉	◉	◉	◉	◉	◐
mask-based inpaint and outpaint Measured: `gw_inpaint_mask` exists in `runtime/_hdl_build/nx_gen_worker.nx`, verified at emit. LANDED 2026-08-30 (composition half): gw_inpaint_mask appends the engine's own mask and inpainting_mask_invert fields to the A1111 img2img body (nx_gen_worker.nx); go_handle_img2img accepts an optional mask (RAW base64 PNG, one channel, 255 = repaint, 0 = keep) and mask_invert (0 or 1) on POST /api/img2img and /api/img2img_async and REFUSES a malformed mask BY RULE NAME before any GPU is touched (NOT-BASE64, INVERT-NOT-BOOLEAN, EMPTY). PROVEN: nx_gen_inpaint_gate GREEN 15/15 through /api/gate_run (on vs off bodies byte-different, both server-parsed field names present, invert 0 vs 1 differ, four named refusals write nothing) and BITE-PROVEN by mutation of nx_gen_worker.nx (valid_mutants=1, not_reached=0, killed=1, restore byte-identical); gate promoted, registered and declared in gen.gates. Daemon built over /api/build (src_path verified, sha 932d6141, 359221 B; nx_contentdiff lost_from_live=0 over 595 runs), the pre-change binary banked by hash to /volume1/ai/gen/banked_genorch_pre_g7g13_a36b5d27.elf, deployed via gdeploy@genorch under an AUDITED pre-deploy-gate override (the blocker was evidence-honesty on four OTHER domains -- webscraping, office, videocodec, autodev -- gen itself read MECH-OK), live hash re-read 932d6141 and the :18795 REFUSED-to-ACCEPT edge witnessed. DEPLOY-LANE LESSON, measured twice today: hc_deploy_one resolves its stage BESIDE THE LIVE PATH, so the first deploy reported PROMOTED and DEPLOYED-GREEN while live, .prev and the nishihost stage were all byte-unchanged; copying the built artifact to /volume1/ai/gen/<name>.sov.elf.new first is what made the second deploy real. THE FLIP IS THE RECEIPT, NOT THE PROOF: the done-rule's RENDER half (a masked img2img through the live daemon, the edited region judged by the coherence organ) is OWED -- the GPU was shared with a game at deploy time and the T1 headroom gate held. Named bound: the mask rides the daemon's 131,071 B request window, pre-existing and not widened here. Watch G7, RE-PRICED 2026-08-29 from engineering to WIRING because the engine half is already done and reachable. Our fork decodes a multipart-uploaded mask into a one-channel image and passes it as the mask_image member of the image-generation parameter struct, and the served sd-server binary carries the matching inpaint literals, so the route answers today at the images-edits endpoint; it predates our fork point rather than needing a rebase. What is absent is the NishiLang verb and its gate. InvokeAI's unified canvas leads [invokeai]; Fooocus inpaints and outpaints [fooocus]; Midjourney varies a region and pans [midjourney]. Adoption: `LIB-WIRED importers=2 nonval=1` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), InvokeAI (leads), Fooocus (leads), Midjourney (leads)	●	◉	◉	◉	◉	◉	◉
super-resolution upscale of a render (ESRGAN compiles into the library and is LINKED OUT of the server) DARK — `runtime/_hdl_build/nx_gen_worker.nx` EXISTS but does not declare `gw_upscale`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. Watch G8, SHARPENED 2026-08-29 with a positive control: the upscaler compiles into the static library, its loader strings are present in the CLI binary and ABSENT from the served sd-server binary, and the server source contains ZERO references to upscale or esrgan, so the linker dropped it. The upscale strings that ARE in the server trace to the shared argument table -- help text, not capability. The C API is already declared. This is a server route plus a link reference, the cheapest of the four control rungs, and it needs no rebase. The gallery serves an on-demand 2x upscale per asset today; the gen lane's own upscale-on-render is the gap. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them dossierwatching gw_upscale Pros none measured yetCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), InvokeAI (leads), Fooocus (leads), Midjourney (leads); open contract; expected source declaration not established	○	◉	◉	◉	◉	◉	◉
prompt weighting, wildcards and embeddings Measured: `cw_prompt_weight` exists in `runtime/_hdl_build/nx_companion_wardrobe.nx`, verified at emit. LANDED 2026-08-30 (weighting half) on the contract knowledge/gen_prompt_weight.conf declared BEFORE the code existed: PERMIL units, a permitted envelope [min,max] as the first two ints (100 2000 = 0.1..2.0), out-of-envelope REFUSED by named code and never clamped (the conf's own reasoning: a typo like skin:13 must not become a render that reads as a model defect), slot weights for trigger, camera and identity as ADDITIVE ints 3..5 with 1000 the identity, so the composed prompt is BYTE-IDENTICAL to the pre-G13 prompt until an operator edits data. WHY IT IS A LIVE CONTROL AND NOT A NO-OP: elder-sdcpp routes the Z-Image user turn through parse_prompt_attention and multiplies each token's hidden state by its weight with a mean-preserving rescale (conditioner.hpp, read from the source 2026-08-30). Wired at two sites: cw_weight_slot in the ONE shared photo composer (a DATA slot out of envelope falls back to the bare slot -- the baseline render, never a mutated one) and cw_weight_admit in go_handle_generate (a USER group out of envelope is refused with its permil and byte offset named). PROVEN: nx_gen_promptweight_gate GREEN 32/32 through /api/gate_run including compose-level on-vs-off through the REAL composer from a scratch cwd, promoted and registered; and LIVE on the deployed daemon (sha 932d6141) -- POST /api/generate with a (skin:13) group answered prompt weight refused rule=OUT-OF-ENVELOPE permil=13000 allowed=100..2000 at_byte=33 before any dispatch, no GPU touched. BITE-PROVEN on the third attempt by mutation of nx_companion_wardrobe.nx (valid_mutants=1, not_reached=0, killed=1, restore byte-identical, twin restored; the first two attempts were refused by build admission, exit 6, nothing compiled, nothing claimed) and declared in gen.gates. OWED: the wildcard-expansion half of the done-rule. Watch G13. Fooocus wildcards and weights [fooocus]; Midjourney multi-prompts with weights [midjourney]. Adoption: `LIB-WIRED importers=3 nonval=1` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), Fooocus (leads)	●	◉	◉	◉	●	◉	●
hires-fix and refiner two-pass rendering DARK — `runtime/_hdl_build/nx_gen_worker.nx` EXISTS but does not declare `gw_hires_pass`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. Watch G14, RECLASSIFIED 2026-08-29 as a pure REBASE and not engineering: it is 0 percent present in our fork -- zero source matches and zero literals in BOTH binaries -- and 100 percent present upstream, where a complete A1111-style hires pass with its own parameter struct and ten upscaler modes is declared in the public header [sdcppsrc]. CONFIRMED BY INDEPENDENT FETCH 2026-08-30: the header carries 19 hires and 20 upscaler symbol occurrences and the server routes carry 15 and 18, so writing this by hand would duplicate a shipped upstream subsystem at subsystem scale. Fooocus refiner passes [fooocus]; Midjourney upscales variants [midjourney]. dossierwatching gw_hires_pass Pros none measured yetCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), InvokeAI (leads), Fooocus (leads), Midjourney; open contract; expected source declaration not established	○	◉	◉	◉	◉	◉	●
WORKFLOW
node-graph substrate -- a cell-graph manifest lib with typed cells and edges, NOT wired to gen Measured: `nx_pathway_add_edge` exists in `runtime/nx_pathway.nx`, verified at emit. Honest: the graph substrate exists (nx_pathway, the ComfyUI-node analogue) but no organ executes a gen graph over it. ComfyUI is the node graph [comfyui]; InvokeAI ships workflows and nodes [invokeai]; SwarmUI exposes the raw Comfy graph tab [swarmui]. Adoption: `LIB-WIRED importers=6 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Forge, Fooocus, Midjourney; fully adopted on the estate ladderCons behind ComfyUI (leads), InvokeAI (leads)	●	◉	●	○	◉	○	○
execute a workflow graph over the gen worker (ComfyUI-class pipelines) Open — watching `runtime/nx_gen_graph.nx` : `gen_graph_exec`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Watch G11: the pathway lib gains an executor whose cells are the existing organs (compose, dispatch, judge, ingest); partial re-execution is the bar ComfyUI sets [comfyui]. dossierwatching gen_graph_exec Pros none measured yetCons behind ComfyUI (leads), SwarmUI, InvokeAI (leads); open contract; expected source declaration not established	○	◉	●	○	◉	○	○
ingest a rival's workflow or PNG metadata and reproduce the render DARK — `runtime/nx_workflow_ingest.nx` EXISTS but does not declare `wi_reproduce`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. Watch G15. ComfyUI reloads a workflow from a PNG [comfyui]; the others read their own parameter metadata. The June-repro ladder (h74 to h48) is the ruler already built for this. dossierwatching wi_reproduce Pros none measured yetCons behind ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus; open contract; expected source declaration not established	○	●	●	●	●	●	○
MODEL
hash-verified acquisition from HF, civitai and civitaiarchive into the model registry Measured: `mg_fetch` exists in `runtime/nx_model_get.nx`, verified at emit. LANDED 2026-08-30 (nx_model_get.elf PROMOTED+REGISTERED; nx_model_get_gate 16/16 GREEN with seven neg-controls bite-proven) and PROVEN LIVE against HF ground truth: a redirect-following, Range-windowed pull (one 8 MiB window in memory at a time, hashed as it lands, resumable from its .part) computed sha256 6de6373d80bd5c34fdae0b1e7a8aa8c7575b73d95ad1d62cd703f33ff6b4493c for z_image/q8_0.png in leejet/Z-Image-Turbo-GGUF, exactly the publisher's lfs.oid; a deliberately wrong expected hash was REFUSED with the partial deleted; a re-run on the verified file answered OK-ALREADY-VERIFIED with no network touched. It COMPOSES nx_https_fetch_range, the streaming nx_sha256 already inside the fetch unit and nx_weights_pull's fail-closed compare rather than duplicating any of them. DONE-RULE STILL PARTIAL, named: the Authorization header leg (civitai gated pulls) and the model-plane row are not wired -- receipts go to knowledge/status/modelget.log today. SwarmUI and InvokeAI ship model managers [swarmui] [invokeai]; Fooocus auto-downloads its presets [fooocus]. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). dossiernot adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked) Pros Implementing source symbol found; runtime acceptance separate; ahead of MidjourneyCons behind SwarmUI (leads), InvokeAI (leads), Fooocus (leads); not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	●	●	◉	●	◉	◉	○
identity LoRA training loop -- favorites to dataset to train to judge-gated deploy Open — watching `runtime/nx_lora_train.nx` : `lt_train_deploy`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Watch G10. Training lives in extensions for ComfyUI and SwarmUI (Part); Midjourney personalisation is the cloud analog [midjourney]. Accept rule pre-declared: the tuned model must EXCEED the baseline on the same prompts under the judge panel or it is not wired. dossierwatching lt_train_deploy Pros none measured yetCons behind ComfyUI, SwarmUI, Midjourney; open contract; expected source declaration not established	○	◐	◐	○	○	○	●
UI
chat-first companion app -- presence and mood card, gallery and studio tabs, served from disk with no rebuild Measured: `go_serve_ui` exists in `runtime/_hdl_build/nx_gen_orchestrator.nx`, verified at emit. SwarmUI's Generate tab, InvokeAI's canvas and Fooocus's one-page UI lead on generation ergonomics [swarmui] [invokeai] [fooocus]; Midjourney's web editor is polished [midjourney]. Ours is edit-and-push HTML. Adoption: `LIB-WIRED importers=6 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind SwarmUI (leads), InvokeAI (leads), Fooocus (leads), Midjourney (leads)	●	●	◉	●	◉	◉	◉
batch BOARD -- named jobs, per-image progress as each render lands, list, stop and resume DARK — `runtime/_hdl_build/nx_gen_orchestrator.nx` EXISTS but does not declare `go_batch_list`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. Watch G4: the old platform's batch start, status, queue, stop, drain and resume surface, rebuilt on the async job files; a row appends per render so an interrupted batch keeps what it did. ComfyUI's queue view and SwarmUI's grid generator are the bars [comfyui] [swarmui]. dossierwatching go_batch_list Pros none measured yetCons behind ComfyUI (leads), SwarmUI (leads), Forge, InvokeAI (leads), Fooocus, Midjourney; open contract; expected source declaration not established	○	◉	◉	●	◉	●	●
OPS
MCP and REST tool surface for every gen verb (generate, batch, img2img, chat, recent, anatomy) Measured: `g_render_body` exists in `runtime/_hdl_build/nx_gen.nx`, verified at emit. ComfyUI and InvokeAI expose full APIs [comfyui] [invokeai]; ours is the nx_gen tool over /mcp, so a seat renders from a call and reads the cid back. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_forkfresh_gate -- a mention, not a run. not adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_forkfresh_gate -- a mention, not a run Pros Implementing source symbol found; runtime acceptance separate; ahead of MidjourneyCons behind ComfyUI (leads), InvokeAI (leads); not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_forkfresh_gate -- a mention, not a run	●	◉	●	●	◉	●	○
self-healing worker and daemon keepers (five-minute beats) with a banked-by-hash rollback binary Source present; authored stronger claim: `gg_probe` in `runtime/nx_gpuguard.nx`, verified at emit. EXCEED with the why: the laptop engine and the NAS daemons relaunch themselves; the pre-change binary is banked and hash-verified before any deploy (2026-08-27 bank fd82c028). Midjourney is managed for you; the local UIs restart by hand. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked) Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, MidjourneyCons not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	◉	○	○	○	○	○	●
SOVEREIGNTY
the gen control plane is sovereign ELF -- zero Python, zero Docker, zero node runtime Source present; authored stronger claim: `main` in `runtime/_hdl_build/nx_gen_orchestrator_daemon.nx`, verified at emit. EXCEED with the why: gateway, orchestrator, pipeline, judges and tools compile with nx_cc to static ELFs; the engine beneath them is a ggml C++ fork [sdcpp], the one foreign component, named. Every rival is a Python stack [comfyui] [forge] [invokeai] [fooocus] or a C# host over one [swarmui]. Adoption: `BUILT-UNPROMOTED` — PARTIAL: compiled, never promoted to the serving root: /api/promote it. not adopted: compiled, never promoted to the serving root: /api/promote it Pros Source symbol found; stronger capability claim is authored; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, MidjourneyCons not adopted yet: compiled, never promoted to the serving root: /api/promote it	◉	○	○	○	○	○	○
CONTROL
structural conditioning from a pose, depth or lineart map DARK — `runtime/_hdl_build/nx_gen_worker.nx` EXISTS but does not declare `gc_struct_condition`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. G20 gap watching gc_struct_condition, and the single largest reason this lane feels behind the local-UI field. Without it an image can be told what to be ABOUT but never what SHAPE to contain. Our own 3D lane already emits the exact pose and depth maps such a conditioner consumes, so the input side is solved and only the worker entry is missing rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them watching gc_struct_condition Pros none measured yetCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), InvokeAI (leads), Fooocus, Midjourney; open contract; expected source declaration not established	○	◉	◉	◉	◉	◐	◐
per-region prompting inside one image DARK — `runtime/_hdl_build/nx_gen_worker.nx` EXISTS but does not declare `gc_region_prompt`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. G21 gap watching gc_region_prompt. This is what makes two characters in one frame controllable rather than sampled, and it is the same axis the manga research bar exposes when it places characters in-panel by masked cross attention rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them watching gc_region_prompt Pros none measured yetCons behind ComfyUI (leads), SwarmUI, Forge (leads), InvokeAI; open contract; expected source declaration not established	○	◉	●	◉	●	○	○
identity adapter so a character survives across images DARK — `runtime/_hdl_build/nx_gen_worker.nx` EXISTS but does not declare `gc_ident_adapter`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. G22 gap watching gc_ident_adapter. The deterministic path here is genome plus seed and it is a strict EQUALITY rather than a similarity score, but it only holds for beings the 3D lane renders. An adapter is what carries identity for a character that came from a style instead of from a genome rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them watching gc_ident_adapter Pros none measured yetCons behind ComfyUI (leads), SwarmUI, Forge (leads), InvokeAI, Fooocus, Midjourney (leads); open contract; expected source declaration not established	○	◉	●	◉	●	●	◉
inpaint and img2img entry on the worker DARK — `runtime/_hdl_build/nx_gen_worker.nx` EXISTS but does not declare `gc_inpaint`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. G23 gap watching gc_inpaint. Repairing one hand in one frame without re-rolling the whole image is the difference between a toy and a tool, and the sketch-to-art rung enters through the same door rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them watching gc_inpaint Pros none measured yetCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), InvokeAI (leads), Fooocus (leads), Midjourney; open contract; expected source declaration not established	○	◉	◉	◉	◉	◉	●
MODEL
a clean base checkpoint beside the finetune, selected by intent DARK — `runtime/_hdl_build/nx_gen_worker.nx` EXISTS but does not declare `gw_model_intent`: something shipped at this path under another name, and this mark would read open forever. Repoint the contract to the real entry point (the plan rung AND this row) or rename the function; the flip follows on the next beat, and the `comparewatch-` plane row reads DARK until then. G24 gap watching gw_model_intent. The served checkpoint is an NSFW finetune, so its prior is baked into the model FILE and contaminates SFW work with fantasy artefacts that no prompt can remove. A clean base is ALREADY ON DISK and hash-verified byte-identical to the publisher, so this is a selection path rather than an acquisition rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them watching gw_model_intent Pros none measured yetCons behind ComfyUI (leads), SwarmUI (leads), Forge (leads), InvokeAI (leads), Fooocus; open contract; expected source declaration not established	○	◉	◉	◉	◉	●	○
RATING
content rating carried as DATA on every generated asset Open — watching `runtime/nx_gen_rating.nx` : `gr_rate_row`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. G25 gap watching gr_rate_row. Every local UI leaves rating to the operator's own filesystem and Midjourney forbids adult content outright. Binding a rating to the ASSET at emit is what lets one gallery serve a family viewer and an adult viewer without a second pipeline, and it is the precondition for the SFW and NSFW split this estate needs rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them watching gr_rate_row Pros none measured yetCons behind ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus; open contract; expected source declaration not established	○	●	●	●	●	●	○
adult invariant fail-closed at construction, wired onto the gen path Open — watching `runtime/nx_gen_rating.nx` : `gr_adult_invariant`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. G27 gap watching gr_adult_invariant. The invariant EXISTS and is mutation-proven in the character lane -- ws_set refuses with WS_E_MINOR below WS_ADULT_AGE 18 -- and what is absent is its wiring onto this path. Every column in this field handles the question with a policy filter or a moderation queue if at all, and a filter is a mechanism that can be talked around whereas a refusal at construction cannot watching gr_adult_invariant Pros none measured yetCons open contract; expected source declaration not established	○	○	○	○	○	○	○
OPS
broken-parts census of /gen on a beat -- health, every recent image fetched and sniffed, album day labels (G28) Measured: `gh_album_years` exists in `runtime/nx_galx_health_lib.nx`, verified at emit. LANDED 2026-10-08: nx_galx_health gen (clock row galxhealthgen, 21600 s) writes knowledge/status/galxhealth/gen.census, gen.rows and the gen.stamp gauge this page renders. First reading: health ok, 48 of 48 recent images decode, 2 of 2 album days under 1970 -- the album clock defect, fixed the same day (gen.plan G19 log row). Rival codes UNREAD, deliberately: no rival's self-census has been measured. Adoption: `LIB-UNCENSUSED` — PARTIAL: a library absent from the import-graph census (younger than the last eco-graph build, or unreachable from buildroot/runtime). dossiernot adopted: a library absent from the import-graph census (younger than the last eco-graph build, or unreachable from buildroot/runtime) Pros Implementing source symbol found; runtime acceptance separate; ahead of ComfyUI, SwarmUI, Forge, InvokeAI, Fooocus, MidjourneyCons not adopted yet: a library absent from the import-graph census (younger than the last eco-graph build, or unreachable from buildroot/runtime)	●	○	○	○	○	○	○

Rival-claim provenance. Every Yes, Best or Part mark in a rival column is a claim about someone else's product. A cited row carries a reference mark that resolves to a pinned mirror (the refs gate measures that); an uncited row is an observation read and is badged in its evidence until a reference lands. The badge is the receipt, never the proof: a mark proves a mirror exists, not that the mirror supports the code. The rows are counted below.

Rival claims, counted
Class	Count	Note
rows with rival marks	44	rows carrying a Yes, Best or Part in a rival column
cited	31	a reference mark resolves to a pinned mirror
uncited	13	an observation read, badged until a reference lands
rows, summed	44	RECONCILED: the parts sum to the total 44

Delivery and evidence

Inspect risks, technical debt, rendered observations, experiments and references. Read scope and limitations alongside every result.

Debt register

Id	Sev	What it is	Unblock
gen-rigor-envelopenot in this scope	6	Every realism and referee number this board publishes (G5 HPSv2, PickScore, GenEval; the June-repro dhash) is a value with no n, no interval, no seed and no engine-config hash beside it, so two renders on different engines read as comparable when they are not (measured 2026-08-27: the same GENREC on the current engine breaks anatomy at dhash 95 to 98 while the June engine did not).	Publish through the shared rigor envelope gv_envelope that /compare/autodev rung AD1 names -- n, a Wilson interval, seed, engine-config hash -- or read REFUSED; G5 depends on it, and the operator made proven evidence mandatory on both boards.
gen-daemon-cwd-confnot in this scope	5	The gen daemon runs with cwd /volume1/ai/gen, so knowledge/swarm_nodes.conf (the role-to-address SSOT) is not visible and the worker address silently falls back to argv; deploy_verify.sh carried the laptop's OLD address (.193) until 2026-08-27 while gen_keeper.sh carried the current one (.192) -- two launchers, two answers.	Make the daemon read the conf by an estate-relative absolute path (ep_artifact_path) and retire the argv fallback to a loud refusal; one launcher.
offload-cold-rendernot in this scope	4	Offload mode (zero VRAM at idle) reloads 10.5 GB per job: a cold render measured 167 s and a warm one 221 s on 2026-08-27 with the card shared with a game, against 15 s warm on 2026-08-14 with the card free. The orchestrator's dispatch timeout is 180 s, so a cold render on a busy card can succeed on the GPU and be dropped by the daemon.	Measure duration-vs-cid over N cold and N warm renders (full population, never three), then make the photo dispatch async like the batch path rather than abandoning offload.

Release history and work log

The worked plan, on the board. Every leg appends what it measured, landed, retracted, learned and left queued, per rung, so a crash leaves the next seat a ledger here rather than a transcript to mine. Kinds: measure, land, retract, lesson, queue. Newest last.

When	Rung	Kind	Entry
2026-09-07	G19	land	Client lifecycle emitter promoted after native build4838 and nine native configuration cases: valid and escaped labels emit, seven malformed inputs refuse with no output. Actual emitted JavaScript passed15 lifecycle checks in a simulated DOM: hidden pages abort owned reads and pause state polling; restored pages resume one poll owner and retain accepted job identities; in-memory chat uses the existing retention policy. Emitted HTML23266 bytes SHA2f09249851e09de6107d40c147a11eb4f9d7f231e88b1af9c81180993a9c172b. Release navigation points to this existing ledger. Contentdiff5259 remains RED for reviewed intentional replacements of old polling and storage functions. UI publication and authenticated HTTP verification are pending at this record; browser visual/heap checks and backend model TTL RAM/VRAM reclamation remain open. No SOTA or memory-release qualification claimed.
2026-09-08	G19	land	Client lifecycle UI delivered through the canonical emitter7084 and authenticated origin verification7093: HTTP200,23266-byte HTML SHA2f09249851e09de6107d40c147a11eb4f9d7f231e88b1af9c81180993a9c172b exactly matches the tested artifact and served file. Visible release-summary and worklog links point to this domain. Native configuration cases9/9 and actual emitted JavaScript lifecycle checks15/15 passed. Client-owned reads and state polling pause while hidden; accepted backend jobs remain owned by the server. Browser visual/heap acceptance, service-worker cancellation integration and actual backend RAM/VRAM unloading remain unverified. This is origin delivery proof, not WAN/CDN or SOTA qualification.
2026-09-12	G17	requirement	2026-09-12 recovery and new-domain pilot: reconcile Elder AI and Companion historical user workflows against current native runtime, API/MCP bindings and served outcomes. Shared scope is retained in media.plan migration-recovery pilot; analyst/business case in finance.plan and projectmgmt.plan. Verified source roots C:/Users/elder/elder-ai-platform and C:/Users/elder/companion are reference evidence, not automatic code adoption. Old availability claims and stale answer prose must be reconciled with current observations before any gap/coverage score. Preserve working behavior and identify unwired or regressed functions with reproductions. Same native research/inventory/rank/dispatch/verify/emit workflow must later qualify a new-domain target. Beach/swarm retain execution priority.
2026-10-08	G28	land	nx_galx_health promoted (sha c80e161e, registered, offc twin) and its gate nx_galx_health_gate 52 of 52 GREEN on the NAS, bite-proven (killed=1 not_reached=0); clock rows galxhealth (census) and galxhealthgen (gen) bound at 21600 s; gauge rows on media.gauge and gen.gauge
2026-10-08	G28	measure	first /gen reading 2026-10-08 on loopback :18795: health ok, recent_cids=48 img_probed=48 img_ok=48 img_fail=0, album_days=2 album_days_before_floor=2 -- both days filed under 1970 because go_album_append stamped rows with sys_now_us (CLOCK_MONOTONIC, seconds since boot), not the wall clock
2026-10-08	G19	land	album clock fixed at the root: go_album_append now stamps sys_now_realtime_sec(); daemon rebuilt over /api/build (sha a5638247, 549,248 B, contentdiff lost 6 runs all shared-library messages, 44 source paths gained), pre-change binary banked by hash at /volume1/ai/gen/banked_genorch_pre_album_c41a6886.elf, deployed via gdeploy@genorch (the first deploy promoted a no-op because the row stages beside the live path -- copied the stage beside live and deployed again; live hash re-read a5638247). The three historical rows were recovered exactly from the monotonic offsets (uptime 70.5 days exceeds every row, so no reboot intervened): elara_album.tsv rewritten 234 -> 243 B with the original banked as elara_album.tsv.pre-clockfix-20261008

On these two registers. Rows are declared in the domain's plan file and carry the debt id, which is the join key back to the sovereign debt plane — that plane, not this page, is the authority on state. Reconciling them automatically (the regen reading the plane and refreshing these rows) is a named, owed rung; until it lands, treat an id here as a pointer to look up, not a status to trust.

Evidence for every claim. This is not an opinion matrix: each capability we mark present is grounded in real source and proven by a live-forked gate. See the autonomously-generated evidence page → (per-axis source lines + executed self-test verdicts) or the machine-readable evidence.json.

Honest verdict. The coverage above records source-symbol presence. Stronger comparisons are authored claims, not measured runtime superiority or accepted outcomes. A watch contract names an expected source symbol; finding it changes source discovery only. Qualification, deployment and runtime acceptance require their own evidence.

Person · product · place — not yet measured for this domain

Every compare carries this layer. Declare knowledge/compare/gen.ppp (rows surface|nishi or c1..c4|label|url|connect naming OUR live surface and each rival's front door), run nx_ppp_probe domain gen, and this section fills itself on the next beat: the same ruler on both sides — privacy and CX (third-party hosts, tracker classes, cookies, security headers), design and longevity (design hygiene, computed WCAG contrast, render-blocking resources, unsized media, script weight, theme and motion queries), findability (landmarks, skip link, on-site search, breadcrumb, headings, internal links).

Mined from the field — review and competitive intelligence proposed to this board

Rows written by nx_intelmine_propose from nx_reviewmine_lib signals (Steam reviews under the exceeds / meets / mixed / does-not-meet rubric, the installed-title census) and routed here by capability_map.conf. DEFECT names a rival failing that a rung here answers; DEMANDED a capability buyers ask for and do not get; WE-DO-BETTER an exceed the reviews corroborate; THEY-DO-WELL a bar this board must meet; SHIPS a rival capability to match. A proposal is a lead, never a rung — it closes only when a rung with a gate lands, and nothing in this band changes a measured cell.

Signal	Kind	Term	Title	Proposed rung	Evidence
THEY-DO-WELL	feature	story	Portal 2 620	Story: 620 -- narrative generation graded by the panel	praised=11 complained=0 mixed=0 demanded=0

proposals 1|malformed rows 0 (counted, never rendered: a row that is not prop| with 10 fields)|read-capped 0

The field — discovered, not chosen

Rows written by nx_field_discover from gen.seeds: the industry's own lists (Wikipedia wikitext, GitHub topics, awesome lists) read mechanically, every candidate counted across seeds. The matrix columns above are a SEAT'S pick; this band is the population they were picked from, and the stats line measures one against the other. A rival here is a lead, never a verdict — it earns a column when its capabilities are read and pinned.

Discovery receipt. One row per counter the run wrote, in the receipt's own order; the rival rows follow in their own table.

Field discovery, counted
Class	Count	Note
seeds	6
scanned	5
fetched	5
reused	0
failed	1
named	6
candidates	6
mentions	6
capped	0

Rank	Rival	Seeds	Mentions	First seed	Kind	Link
1	ComfyUI	1	1	col0-comfyui	named	https://raw.githubusercontent.com/comfyanonymous/ComfyUI/master/README.md
2	SwarmUI	1	1	col1-swarmui	named	https://raw.githubusercontent.com/mcmonkeyprojects/SwarmUI/master/README.md
3	Forge	1	1	col2-forge	named	https://raw.githubusercontent.com/lllyasviel/stable-diffusion-webui-forge/main/README.md
4	InvokeAI	1	1	col3-invokeai	named	https://raw.githubusercontent.com/invoke-ai/InvokeAI/main/README.md
5	Fooocus	1	1	col4-fooocus	named	https://raw.githubusercontent.com/lllyasviel/Fooocus/main/readme.md
6	Midjourney	1	1	col5-midjourney	named	https://docs.midjourney.com/hc/en-us/articles/32023408776205-Parameter-List

Gauges — a heartbeat, never a bare number

Each row names a gauge stamp written by its measuring beat. This page re-judges the stamp on every publish with the same ruler the beat uses (nx_gauge_lib): FRESH shows the value; STALE withholds it (older than two beats); BLIND withholds it (an axis abstained); ABSENT means nothing has measured. A stale gauge never reads as zero.

gauge	state	reading	note
Broken parts on /gen as the browser sees them (G28): /health, every /api/recent image fetched and sniffed as an image, album day labels before 2020 (a clock defect, not a photo)	FRESH	FRESH gauge=0 axes=3/3 age_s=11486	written by nx_galx_health gen on the galxhealthgen clock row beside knowledge/status/galxhealth/gen.census and gen.rows. axes = health, recent-images, album: a leg that did not run reads BLIND, never zero. First measured 2026-10-08: health ok, 48 of 48 recent images decode, 2 of 2 album days filed under 1970 because the daemon stamped rows with CLOCK_MONOTONIC (fixed in source the same day).

Exclusive and shared capabilities

One capability is graded on one board; every other board references it. The fleet duplicate census (nx_rivalcensus dupes, on the fleetdupes clock row) reads every board's matrix and names each capability graded as a row on two or more boards, with a home derived from evidence: rung when exactly one board's plan contracts the symbol (that board is starred), gates when no plan contracts it and exactly one board's executed gate map runs the organ's gate (that board is plussed), ambiguous when several boards carry the deciding evidence (a decision for their seats), and none when no evidence decides it, with the reason beside it: unmapped (no board of the duplicate declares a gates map; declare the maps) or unproven (a map exists and runs none of the organ's gates; gate the organ on its proving board, or declare it a shared library by policy). A home elsewhere means this board should carry an @xrow reference to the home board's row rather than a graded row of its own; a home here means the other boards should. The table is this board's slice of one estate-wide census: the shape is shared, the rows are the board's.

This board grades 62 capabilities: 51 exclusive to it and 11 shared with other boards. Shared rows on this page: 11 = home here 2 + home elsewhere 1 + ambiguous 1 + undecided 7 + refused 0 (partition 11); undecided = unmapped 0 + unproven 7 + unevidenced 0; homonyms 2; since the previous census: new 0, resolved 0. Census FRESH age_s=6953.

Capability	Organ	Boards grading it	Home	Rule	Evidence	Standing on this board
main	`-`	browser:runtime/_hdl_build/nx_uiq_perf.nx, graphviz:runtime/_hdl_build/nx_eco_graph_forest.nx, graphviz:runtime/_hdl_build/nx_eco_graph_png.nx, graphviz:runtime/_hdl_build/nx_eco_graph_forest.nx, graphviz:runtime/_hdl_build/nx_eco_graph_wiki.nx, familyatlas:runtime/_hdl_build/nx_eco_graph_png.nx, familyatlas:runtime/_hdl_build/nx_eco_graph_card.nx, ecosysdesign:runtime/_hdl_build/nx_eco_graph_arch.nx, typography:runtime/nx_font_ttf_emit.nx, koikatsu:runtime/_hdl_build/nx_retarget_emit.nx, pmdash:runtime/nx_dr_ocm_cli.nx, failmodes:runtime/nx_compare_refs_gate.nx, ecosystem:runtime/nx_compare_rank_fleet.nx, gen:runtime/_hdl_build/nx_gen_orchestrator_daemon.nx, capregistry:runtime/nx_catalog.nx, capregistry:runtime/_hdl_build/nx_coe_intoto.nx, capregistry:runtime/nx_toolreg_reconcile.nx, capregistry:runtime/nx_capsweep.nx, datavis:runtime/_hdl_build/nx_chartlay_gate.nx, speech:runtime/nx_voice_eval.nx	—	homonym	-	homonym
cjc_hw	`nx_charjudge.nx`	mangagen, charsim+, graphics, koikatsu*, gen	koikatsu	rung	plan	home elsewhere adopt: `@xrow koikatsu\|cjc_hw\|home by rung (fleet duplicate census)`
si_hex	`nx_galx_saveimg.nx`	mangagen, gen	—	none	unproven	undecided
do_rank	`nx_gen_rank.nx`	mangagen, sharehub, gen	—	none	unproven	undecided
go_image_symmetry	`nx_image_symmetry_lib.nx`	mangagen, gen, aesthetictwin	—	none	unproven	undecided
lt_train_deploy	`-`	mangagen:runtime/nx_galx_lora_train.nx, gen:runtime/nx_lora_train.nx	—	homonym	-	homonym
go_compose_photo_prompt	`nx_companion_wardrobe.nx`	gen, companionchat	—	none	unproven	undecided
go_load_appearance	`nx_companion_identity.nx`	gen, synthroom	—	none	unproven	undecided
lc_screen	`nx_lora_screen_lib.nx`	gen*, mediaingest	gen	rung	plan	home here
go_handle_chat	`nx_gen_orchestrator.nx`	gen, companionchat	—	none	unproven	undecided
ccm_json_str	`nx_card_compile.nx`	gen, companionchat	—	none	unproven	undecided
mx_swap_ttl	`nx_mesh_mux.nx`	gen, companionchat	—	ambiguous	plan	ambiguous
mg_fetch	`nx_model_get.nx`	gen*, dataio	gen	rung	plan	home here

Gaps from the record — what the estate does that no board carries

The record census (nx_goalmap record) reads the invoked-tool population and every plan queue row and files each organ or directive that NO matrix, plan or gates row names. A row here is a callout the boards missed: adjudicate it onto a board or declare it infrastructure. Census state STALE (age 1396058 s), sources read 4 of 7 declared — a BLIND census is a FLOOR: unread sources can only add rows.

kind	name	board	source	evidence

rows shown 0|this board's directives 0|estate-wide un-boarded organs 452 (listed in full on /compare/ecosystem)|census rows 639|malformed 0 (counted, never rendered)

References

Beyond a link list. Every reference below resolves twice — the publisher's copy and, where banked, the estate's own non-rottable library mirror with a content pin — and carries its evidence class plus the exact claim on this page it grounds. Keyed marks like [key] in the matrix notes jump here. A dash means honestly absent, never assumed.

[comfyui] ComfyUI README (comfyanonymous/ComfyUI, master): the node-graph engine -- asynchronous queueing, partial graph re-execution, smart VRAM and RAM management, model offloading, workflow embedded in and reloaded from PNGs. publisher · read in our library knowledge/fetched/cmp_gen_comfyui.md · pin hd792c3aa57675e2677bb9fb149e651fef460a4ba4fa8f8ea1063538b951efa09 · accessed 2026-08-27 · vendor-docGrounds: GEN: async batch job
[swarmui] SwarmUI README (mcmonkeyprojects/SwarmUI, master): a Generate tab over a Comfy workflow tab, a swarm of GPU backends for one user, grid generation, model management. publisher · read in our library knowledge/fetched/cmp_gen_swarmui.md · pin h93851e61925c4d2a98c28b50ee83ab55e3e6fc730ad3ed49fdd94a512cfe393d · accessed 2026-08-27 · vendor-docGrounds: WORKER: non-blocking TCP health probe and failover across GPU workers
[forge] Stable Diffusion WebUI Forge README (lllyasviel, main): the A1111-lineage UI with native ControlNet integration, LoRA loading, offload modes and Flux support. publisher · read in our library knowledge/fetched/cmp_gen_forge.md · pin he99a0cdc5711816a3da4edb1d70f41dcf3396a6e1fd524f9dd236ce6b1a76cc0 · accessed 2026-08-27 · vendor-docGrounds: CONTROL: ControlNet and pose or structure conditioning
[invokeai] InvokeAI README (invoke-ai/InvokeAI, main): the unified canvas with in- and out-painting, workflows and nodes, boards and model management, an API. publisher · read in our library knowledge/fetched/cmp_gen_invokeai.md · pin h48293a6dbe76e2980f0532b64ee8bd2e236d48ff00c8601bc35c9e0bdf80646a · accessed 2026-08-27 · vendor-docGrounds: CONTROL: mask-based inpaint and outpaint
[fooocus] Fooocus readme (lllyasviel/Fooocus, main): the one-page UI with styles and presets, wildcards and weights, inpaint and outpaint, upscale, refiner passes, low-VRAM offload and auto-downloaded models. publisher · read in our library knowledge/fetched/cmp_gen_fooocus.md · pin h89e1efb0ef1245a6b2abc4db41d5b76e3179656f875db99675900dc477e067bf · accessed 2026-08-27 · vendor-docGrounds: PROMPT: camera and photographic style variety rotated by seed
[sdcpp] stable-diffusion.cpp README (leejet, master): the ggml C++ inference engine our elder-sdcpp fork is built on -- the one non-sovereign component in the gen lane. publisher · read in our library knowledge/fetched/cmp_gen_sdcpp.md · pin h258a8da1b214542f52937934bc97c06100b0b64bfb32f6516a3c04353b4b47f2 · accessed 2026-08-27 · source-readGrounds: SOVEREIGNTY: the gen control plane is sovereign ELF
[zimage] Z-Image-Turbo model card (Tongyi-MAI, Hugging Face): the distilled single-stream DiT -- 8 NFEs, sub-second on an H800, fits 16 GB consumer VRAM, photorealism and instruction adherence, Qwen3 text encoder. publisher · read in our library knowledge/fetched/cmp_gen_zimage.md · pin h87af5e1d7933dbfe091c4048a20805541130690ee813b8039b7f396b47e64338 · accessed 2026-08-27 · vendor-docGrounds: GEN: text-to-image on the sovereign GPU worker
[sdcppsrc] stable-diffusion.cpp public header at master, fetched and read 2026-08-30: the hires-fix upscaler enum and parameter struct are declared here and embedded in both the image and video generation parameter structs, and the upscaler context creation and upscale entry points are declared. Neither construct existed at our fork point, which is the grounds for pricing the hires pass as a REBASE and the upscaler as a LINK REFERENCE rather than as engineering. WITNESSED INDEPENDENTLY: 18628 bytes, sha256 5c7c463ba2acec617d4e8989c701f22518276f5c2f052bcca249dbf7f56d323c, symbol counts upscaler 20 hires 19 cancel 6 mask_image 1. NAS MIRROR OWED -- the body is banked laptop-side; fetch through nx_research_fetch to knowledge/fetched and fill the mirror and pin fields before any gate relies on this row. publisher · accessed 2026-08-30 · source-readGrounds: CONTROL: hires-fix and refiner two-pass rendering
[sdcppasync] stable-diffusion.cpp server routes at master, fetched and read 2026-08-30: the image and video generation routes return 202 with a poll URL and a job status, return 429 when the queue is full, expose cancel for queued and generating jobs, and publish a capabilities endpoint. That async job API is entirely absent from our fork, which is fully synchronous on a single mutex with no queue and no cancel -- the structural context for the dispatch-drop defect measured 2026-08-29. WITNESSED INDEPENDENTLY: 23668 bytes, sha256 39511afff2d7397d4015ac86f364869abfb828047a62287bd309b42fe1d7929b, symbol counts poll_url 2 202 2 cancel 9 upscaler 18 hires 15. NAS MIRROR OWED -- the body is banked laptop-side; fetch through nx_research_fetch to knowledge/fetched and fill the mirror and pin fields before any gate relies on this row. publisher · accessed 2026-08-30 · source-readGrounds: GEN: async render dispatch
[hpsv2] Wu et al. Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis. arXiv 2306.09341. publisher · read in our library knowledge/fetched/cmp_gen_hpsv2.html · pin heb603c9e47e29126cce93eb981315b77f8fed06eef319d17d81e6a41c71320e7 · accessed 2026-08-27 · published-paperGrounds: JUDGE: external human-preference referee
[pickscore] Kirstain et al. Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation (PickScore). arXiv 2305.01569. publisher · read in our library knowledge/fetched/cmp_gen_pickscore.html · pin h80a8945a13ac9404be47cfbc9d7d01dbaae828cc09e1f3b54f781e25a575f87b · accessed 2026-08-27 · published-paperGrounds: JUDGE: external human-preference referee
[geneval] Ghosh et al. GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignment. arXiv 2310.11513. publisher · read in our library knowledge/fetched/cmp_gen_geneval.html · pin h71cb2deb745573e5acc3ad500ee3ae685e0bc4d836ccfdb0ee583caaca246571 · accessed 2026-08-27 · published-paperGrounds: JUDGE: external human-preference referee
[midjourney] Midjourney documentation, Prompt Basics (the parameter-list article resolves here): prompt-embedded parameters, stylize and style, character and style references, region variation, personalisation. publisher · read in our library knowledge/fetched/cmp_gen_midjourney.html · pin h2609fae5308a4f632014e438ffbc134414194ba8a45ae2d9a552b2a5d3d6423a · accessed 2026-08-27 · vendor-docGrounds: GEN: the engine's prompt-embedded control block
[arxiv_t2i_202610] arXiv API listing, text-to-image generation, 30 newest submissions, read 2026-10-08 (entries dated 2026-10-01 to 2026-10-07): UltraText Bench (2610.09823), Iris-3B pixel-space diffusion (2610.09450), phase-wise velocity distillation (2610.08070), preference-and-rubric post-training of frontier text-to-image models (2610.02967), recursive self-improvement in unified multimodal models (2610.03002), DEPICT alignment scoring (2610.03617), NegT2IBench (2610.03084), AREX few-step flow sampling (2610.03483), safe generation and concept erasure (2610.05908, 2610.05000, 2610.05401). publisher · read in our library knowledge/fetched/cmp_gen_arxiv_t2i_202610.xml · pin hdd2584c0d238c2d499a9c51baf316da164d977ff1f41629889f4e5730eb2a7b7 · accessed 2026-10-08 · mirrored-indexGrounds: The current month's listing for SB-GEN-T2I: what the field is publishing in October 2026 (few-step distillation, pixel-space training, preference and rubric post-training, unified multimodal models, text-rendering and negation benchmarks, safety).
[hf_t2i_202610] Hugging Face model hub API, text-to-image pipeline, top 25 by downloads, read 2026-10-08: SDXL base 4.37M, Qwen-Image-2.1 GGUF quants 1.93M and 0.81M (model released 2026-09-20), Qwen-Image-2.1 viggle-turbo and Pruna few-step distills (2026-09-22/23), Z-Image-Turbo and its GGUF 0.62M and 0.90M (Tongyi, 2025-11, Apache-2.0), FLUX.1-dev 0.70M and schnell 0.66M, Qwen-Image-2512 GGUF, Qwen-Image-Lightning. publisher · read in our library knowledge/fetched/cmp_gen_hf_t2i_downloads_202610.json · pin hb5368b9807eae2d5b305fa3f39dc3b8dd72a1912344b24dbbd1de0216bfc44cc · accessed 2026-10-08 · mirrored-indexGrounds: The open-weights usage witness for SB-GEN-T2I: the models the local field actually runs this month; a download table is adoption evidence, never a quality claim.

generated by nx_swcompare_matrix (sovereign NishiLang organ) from knowledge/compare/gen.matrix · source checks show implementation presence; runtime and user-outcome evidence are reported separately · JavaScript supports page controls