Nishi Family › Compare › Media Ingestion, Capture and Cataloguing

Nishi Compare · capabilities, evidence and delivery

Media Ingestion, Capture and Cataloguing

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

Nishi vault capture and analysis-on-ingest vs Harvestr (archive plus live fleet) and ctbrec (cam recorder) and yt-dlp (extractor engine) and Video DownloadHelper (in-browser capture) and Stash and Hydrus (self-hosted catalogues) -- August 2026

Overview

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

Where we are. Measured 2026-08-30 against the matrix, not recalled. The sovereign stack owns acquisition primitives (own TLS with a Chrome hello, HLS with AES-128, streaming-to-disk, a full BitTorrent client, zero-network JS harvest, Wayback cross-temporal recovery) and a real catalogue (70-detector byte sniffer, streaming CID, classify-at-the-door vault records -- 160,415 of them -- collections at ingest, thumbnails, keyframe index, usability verification, takedown exclusion, a PII-walled fact extractor). The page-level gaps are the ones the operator hits: the polite pacer exists and is imported by the crawl lane and nx_4chan but by NONE of the capture verbs (nx_vault_capture, nx_mvault_fetch, nx_hls_get, nx_media_grab; corpus_complete=1), there is no exit rotation and no source pre-flight, and the bot-wall classifier that exists (nx_antibot, the webscraping R4 contract) is consulted only by the crawler step, never by a capture verb; reach is 4chan plus one data-row album adapter against Harvestr's 32 patterns and 39 scrapers and yt-dlp's 1800+ sites, with no cookie import, no DASH, no segmented download and no live-record fleet, so the 53,567-recording cam corpus still comes from the retired Python recorder; the thumbnail pipeline is PNG-only and the keyframe index covered 1.7 percent when last measured; and the training half has the stores (capture, embedding, manifest export) without the producers (tagger, encoder, provenance-carrying dataset export). A hand-written Waterfox extension exists laptop-only and pushes to the torrent daemon, not to vault capture. MILESTONE STATUS 2026-08-31, stated against the pre-declared done-rules and NOT against symbol presence: all five M0 contract symbols (vc_pace_hook, sh_preflight, ig_disk_floor, th_decode_any, sa_count) are present and their watch rows are flipped, but M0 is NOT closed, because two of its five rungs fall short of the done-rule written above them. R0 requires FOUR capture verbs to call the one hook and THREE do (nx_vault_capture, nx_mvault_fetch, nx_hls_get); nx_media_grab is deferred because the import tipped its closure past the nx_cc module limit. R3 requires a fixture set of one file per format plus a zero-decode-error re-run over the 07-22 census recordings, and neither was done; its promote also went to a root no thumbnail consumer reads, so the decoder is not live on the gallery path. R1, R2 and R8 are met. The bar is the rung done-rule proven by its gate, never the watch flip, so this milestone reads OPEN until those two are closed.

Where we need to go. Stop being rate-limited by construction (pacing in the capture path, pre-flight, owned-egress rotation, bot-wall verdicts), reach what Harvestr and yt-dlp reach (fan-out, cookies, DASH, segmented download, a live fleet) without leasing ffmpeg or yt-dlp, catalogue at Stash and Hydrus depth (video phash, full thumbnail coverage, sprites), and make every ingested item training-ready with its provenance and licence attached -- served natively in the Nishi browser and OS and, for Windows, Waterfox and every third-party browser, through an emitted extension that talks to the estate over the authenticated edge so no native companion is ever needed.

The unit. 1 u = one measured session-leg. Calibration from landed rungs: the mangagen panel compositor went from existing substrate to shipped and live-verified in ONE leg (2026-08-13); the citations rung went from 3 to 55 domains in one leg across seven seats (2026-08-18); a greenfield engine with a bite-proven gate has measured 2 to 4 legs. Estimates recalibrate as rungs land and PR7 actuals write back.

Cost to stop getting rate-limited: 3 u. Through M0: the pacer already exists, so the first rung is one hook in the capture chokepoint plus a pre-flight and a disk floor; the thumbnail decoder rung rides with it because it is the same class of composing an incumbent that already exists.

Cost to reach what Harvestr reaches: 15.5 u. Through M2: cookies, segmented download, DASH, fan-out, the live fleet, the emitted extension and the two ethics-bounded reach rungs shared with webscraping. The live fleet is the long pole at 3 u.

Cost to catalogue and train from the vault: 27.5 u. Everything below: video phash, sprites, the tagger and encoder (local-model rungs), the provenance-carrying dataset export and the ingest board.

Operator investment 2026-09-10: media intake and search: REQUESTED / UNFINISHED. Compose existing R16/R17/R19/R20 with source-CID provenance: configurable image/video buckets, semantic retrieval, shop-media garment versus body-reference separation, uncertainty-labelled digital-twin candidates and library links. Reversible keep/archive/remove proposals require review; preserve original source and additive history. No automatic source deletion or training authorization. Supporting backlog; Beach GE57 quality remains primary and GE61 measures delegation benefit.

Water tutorial to configurable capability 2026-09-10: REQUESTED / VIDEO CONTENT UNVERIFIED. Operator source https://www.youtube.com/watch?v=IdUDj5XJI7s is a lead, not observed technique evidence. Verify actual video/transcript through permitted existing media tools, retaining timecodes and source rights. Desired reusable library capability: configurable water that can serve photoreal or anime-styled worlds. Proposed 1/2/3 progression: identify and reproduce individual useful tutorial techniques with inputs and evidence; expose composable parameter/time/material/export operations with repeatable checks; publish a reviewed one-call recipe with user controls and Compare investment evidence. Intermediate useful techniques count when independently qualified; final appearance alone is insufficient. Link existing water and Infinigen work under gameengine/graphics, not a separate video project. No claim video watched, code licensed or water release delivered.

Latest recorded release

2026-10-08 · R22

the gallery backend deployed 2026-10-08 (galxserve, sha a9d8d1d7, first health-checked deploy of this daemon through /api/deploy -- its deploy_domain row galxserve mediaingest was missing and the GLOBAL fallback blocked on eight unrelated RED domains until the row was added) carries the thread view's download-all-to-library link (/api/4chan/panel); the fetch and the commit were driven as two sanctioned jobs because no loopback client exists for the authenticated POST /api/4chan/ingest -- the shape the adapter ABC must give every source.

Release history and work log

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

52 of 66 capability source symbols found|15 of them authored stronger-claims|14 open|coverage 787/1000|adoption 23 full / 29 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 = 53 vs present 53 · named + unnamed + withheld = 14 vs open 14 · both reconcile

liar-kill conj=GPQN · all four conjuncts held

graded document: BUILDROOT tree, 39776 bytes · gates map: PRIMARY · stamped 3d 7h ago · source ../knowledge/status/evstamp_mediaingest.verdict

Gap class	What it is, and the work it names
FLIP-READY	A watch contract symbol is now present in the organ it named: the cell can flip on a pure data edit with no new code. These are the cheapest closures on this board and they belong before any new work.

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 mediaingest 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: STALE_INCONSISTENT. Recorded priorities require current acceptance evidence and resource checks before execution.

Recorded ranking withheld

Retained ranking is not bound to the current plan and matrix: STALE_INCONSISTENT. Historical data remains retained. Refresh the ranking after resolving the input differences. Latest execution outcome is unverified.

Declared roadmap — contract, acceptance, executor, effort

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Pacing wired into the capture chokepoint (R0)	vc_pace_hook	nx_vault_capture, nx_mvault_fetch, nx_hls_get and nx_media_grab call ONE hook composing pace_before and pace_after around every fetch; a planted 429 with Retry-After on a fixture host is honoured and the next request waits; the gate proves the wait arithmetically and a neg-control proves an un-hooked fetch does not wait; no second pacing ruler is written	Organ	0.5 u
Source health pre-flight (R1) after R0	sh_preflight	Before queueing work against a host, a bounded probe classifies it UP, SLOW, NULL-ROUTED or WALLED and the verdict is a row the queue reads; a planted dead host is skipped with its reason named; results persist so a flapping CDN is not re-probed on every item	Organ	0.5 u
Disk floor at ingest (R2)	ig_disk_floor	Every capture verb reads the disk axis nx_syscalls already carries and REFUSES below a floor declared in conf, announcing the free bytes and the floor; the gate plants a floor above the current free space and proves the refusal fires	Organ	0.5 u
Thumbnail decoder coverage (R3)	th_decode_any	nx_galx_thumb composes the in-tree JPEG, GIF and WebP decoders and handles palette and 16-bit PNG; a fixture set of one file per format thumbnails end to end; the galx_pipeline.log decode-error class reads zero on a re-run over the recordings the 07-22 census flagged	Organ	1 u
Cookie import (R4)	ci_netscape	A Netscape cookies.txt and the operator's Firefox and Waterfox profile cookies import into the fetch jar with per-site domain filtering; a login-walled fixture page that returns 403 anonymous returns 200 with the imported jar; secrets never printed	Organ	1 u
Segmented multi-connection download (R5) after R0	dl_segments	nx_https_fetch_range composed into a segmented writer with N parallel ranges, resume from a partial file and a whole-file CID check at the end; measured throughput against the single-stream fetch on the same URL is published beside the rung	Organ	1 u
DASH download (R6) after R5	dg_fetch	An .mpd manifest parsed to representations, the best video plus audio fetched as segments and muxed to one playable file through the existing sovereign container path; a public DASH fixture round-trips and plays in the gallery	Organ	1.5 u
Identity fan-out (R7) after R0,R1	if_probe	One username probed across N site patterns held as DATA rows (not code), in parallel under the pacer, every found profile emitted as a row with its site and the media count; a fixture of three sites with one planted hit and two planted misses partitions correctly and the partition sums	Organ	1.5 u
Adapter census (R8)	sa_count	An organ COUNTS the site adapters (data rows plus per-source organs), proves each against its fixture and publishes the number on this page; a census that cannot prove an adapter counts it as UNPROVEN, never as reach	Organ	0.5 u
Live recorder fleet (R9) after R0	lr_watch	A supervisor polls a data-row list of performers across sites, hands a PUBLIC stream to the sovereign HLS chain the instant status flips, records continuously with per-site switches that stop or resume in-flight captures, survives an egress change, and writes poster plus NXVI at record time; measured against the retired Python recorder's corpus shape and proven on one live public stream before any cam site	Organ	3 u
Emitted browser extension (R10)	ge_emit	An organ emits the Waterfox and Firefox MV2 extension and the Chrome and Edge MV3 twin from data (permissions, edge host, media patterns), the extension sniffs media in the tab and pushes one click to nx_vault_capture over the authenticated edge, and the emitted artifacts are served from nishifamily with their hashes published; the laptop-only nishi-grab is retired the day the emitted one loads	Organ	1.5 u
Edge push without a companion (R11) after R10	ge_edge_push	The extension needs no native host on Windows, Waterfox or any third-party browser because the estate does the capture; a fixture run from a Windows browser with no WSL and no installed binary lands an item in the vault; documented beside the native-messaging contract it deliberately avoids	Organ	0.5 u
Egress rotation on owned hosts (R12) after R0	pp_rotate	Rotation across sovereign-owned egress endpoints only (never residential botnets); the pace gate stays in the path and the policy is rows; shared contract with webscraping R3	Organ	1 u
Bot-wall verdict in the capture path (R13) after R0,R1	vc_wall_verdict	Every capture verb runs nx_antibot abt_classify (the webscraping R4 contract, already landed and imported only by nx_web_crawl_step) over a non-2xx or challenge body and records the verdict as an outcome row feeding the source-health row; a planted Cloudflare specimen reads VENDOR_WALL and a clean 200 reads CLEAN (neg-control); never solved	Organ	1.5 u
Video perceptual hash (R14)	vph_scene	A frame-sampled fingerprint per video (first-IDR decoder plus the dHash kernel) stored under the CID, a Hamming nearest over the corpus, and the vault consults it at ingest so a re-encoded mirror collapses; a fixture pair (source and re-encode) matches and a fixture pair of different videos does not	Organ	1.5 u
Trickplay sprites at ingest (R15) after R3	gt_sprite_sheet	A sprite sheet plus index derived per video in the analysis-on-ingest pass; an on-disk byte probe proves the sheet exists for every indexed video; shared contract with media R6	Organ	1.5 u
Auto-tagger and captioner (R16)	ct_tag_item	A sovereign tagger writes measurements, tags and captions into the capture store for every new item (the integer judges first, a local VLM where one is admitted), with the judge panel and a pre-declared agreement bar against a hand-labelled set before it is wired to the ingest beat	Local model	3 u
Semantic index (R17) after R16	msm_search	A sovereign embedding index over the library with recall at 10 PRE-DECLARED to exceed BM25 on a labeled set or the lane stays unwired; shared contract with media R10	Local model	3 u
Dataset export shards (R18) after R16	ds_export_shards	Image and video-clip shards exported from vault queries with a manifest whose rows carry CID, source, licence tier and consent flag; a row without provenance is refused; the manifest partition must sum to the query count; the LoRA loop consumes it directly	Organ	1.5 u
Ingest board (R19) after R9	ib_page	One served page: queue, live fleet, per-site switches, history, failed-with-reason and dry-run, all read from the planes the organs already write; page_verify GREEN and the live judge over it	Organ	1.5 u

Milestones

Milestone	Rungs	Cumulative
M0 · Stop getting rate-limited	R0,R1,R2,R3,R8	3 u
M1 · Reach what Harvestr reaches	R4,R5,R6,R7,R22	10 u
M2 · Record, extend and surface	R9,R10,R11,R12,R13	17.5 u
M3 · Catalogue at Stash and Hydrus depth	R14,R15	20.5 u
M4 · Train from the vault	R16,R17,R18,R19	29.5 u

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Operator flag at ingest: one human tag row per asset on the accept plane, consent class attached (R20) after R19,R16	ct_operator_flag	OPERATOR 2026-09-02: the operator will flag faces and bodies at ingestion as part of the genetic research, and those flags are the population that aesthetictwin AT20 fits its beauty ranges from. Data, not a model: one knowledge/store/accept- row per flagged asset (id, asset cid, tagger identity, tag vocabulary key, consent or licence class, ts) written from the ingest board (R19) or the emitted extension (R10) in one click. R16's auto-tagger PROPOSES and the operator DISPOSES; the row records which. Done when a flagged asset shows its tag row on the ingest board, a re-ingest of the same cid does not duplicate the row (idempotent by cid+tagger), and nx_realperson fails closed on a tagged named person without a consent class. LAST MILE by operator order: placed in M5 after training so the ranker never puts it ahead of the engine gap.	Organ	2 u

Milestones

Milestone	Rungs	Cumulative
M5 · Operator flags become fitted ranges	R20	2 u

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Provenance class gates USE, not just presence (R21)	pc_may_fit_axis	R18 already refuses a manifest row without provenance -- but nx_refcorpus's rc_put_row bounds the provenance and licence fields by LENGTH ONLY, so `unverified` and `cc0` were accepted identically and the word carried no consequence. R21 makes the class MEAN something: knowledge/asset_provenance.conf declares each tag with a per-class k-anonymity floor and a USE (shipped or reference-only), nx_provclass_lib is the ONE classifier both writer and reader compose, and pc_may_fit_axis answers the question the build-path rule structurally cannot -- keeping an asset out of the build path stops us shipping its BYTES, not its IDENTITY propagating through a generator fitted to it, and that output ships. Done-rule: nx_provclass_gate GREEN with the live journal population classifying (no existing row refused), the discriminating pair firing (same tag REFUSES at n=8 and ADMITS at n=32, so the raised floor is proven binding rather than decorative), and three named neg-controls. Feeds R18 and aesthetictwin AT18/AT20: a reference-only class may produce AGGREGATE STATISTICS ONLY and may never fit a beauty range	Organ	0.5 u

Milestones

Milestone	Rungs	Cumulative
M6 · Provenance becomes a wall	R21	2.5 u

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Thread, blog and feed adapters beyond 4chan: Tumblr and Reddit first (R22) after R0,R4,R8,R13	sa_adapt_thread	OPERATOR 2026-10-08: download every media type from a thread like a 4chan /s/ thread (done the same day: 151 of 151 items, jpg + png + webm, committed to the vault records), then the same one-call ingest for Tumblr blogs, Reddit users and subreddits and the other sources the retired media manager reached, better designed. One adapter ABC (nx_source_adapter) with list and fetch verbs over DATA rows per site (URL grammar, JSON or HTML shape, media selectors, pacing class), composing the pacer (R0), the cookie jar for login walls (R4), the bot-wall verdict (R13) and the vault commit; a fixture per site (one planted thread carrying every media kind) round-trips; the adapter census (R8) counts each proven adapter; the gallery panel offers the same download-all for every adapter and a loopback client drives the authenticated ingest route. Done when the Tumblr and Reddit fixtures ingest end to end and the census publishes the reach	Organ	2 u

Ladder verdict. NOT DECLARED. This board names no dated best-in-class or frontier target and no rung roles (sotatarget and rungrole rows on its plan); the ranker labels it NO-LADDER until it does, and until then its rungs climb toward a target nobody has written down. Bars read NO-BAR against the current month 2026-10.

Bar freshness, counted
Class	Count	Note
bars	0	sotabar rows on the plan
fresh	0	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	0	RECONCILED: the parts sum to the total 0
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 23. Rank input binding: STALE_INCONSISTENT. Dependency order is authored. Implementation, acceptance evidence, authority and resource readiness are unverified. No action is recommended or dispatched here.

Plan SHA-256 6e62a489584a31a3b04a49c2d08debe45779d6f6a73ee06fe49d335e34f26de0. Target rows 0; role rows 0. Existing risks and release worklog retain their own scope; no node completion is inferred.

Runtime observation coverage: 0 of 23 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.

R0 — Pacing wired into the capture chokepoint
Prerequisites: None declared; this does not establish execution eligibility.
Contract: vc_pace_hook
Acceptance: nx_vault_capture, nx_mvault_fetch, nx_hls_get and nx_media_grab call ONE hook composing pace_before and pace_after around every fetch; a planted 429 with Retry-After on a fixture host is honoured and the next request waits; the gate proves the wait arithmetically and a neg-control proves an un-hooked fetch does not wait; no second pacing ruler is written
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.
R1 — Source health pre-flight
Prerequisites: R0 (acceptance unverified)
Contract: sh_preflight
Acceptance: Before queueing work against a host, a bounded probe classifies it UP, SLOW, NULL-ROUTED or WALLED and the verdict is a row the queue reads; a planted dead host is skipped with its reason named; results persist so a flapping CDN is not re-probed on every item
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.
R2 — Disk floor at ingest
Prerequisites: None declared; this does not establish execution eligibility.
Contract: ig_disk_floor
Acceptance: Every capture verb reads the disk axis nx_syscalls already carries and REFUSES below a floor declared in conf, announcing the free bytes and the floor; the gate plants a floor above the current free space and proves the refusal fires
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.
R3 — Thumbnail decoder coverage
Prerequisites: None declared; this does not establish execution eligibility.
Contract: th_decode_any
Acceptance: nx_galx_thumb composes the in-tree JPEG, GIF and WebP decoders and handles palette and 16-bit PNG; a fixture set of one file per format thumbnails end to end; the galx_pipeline.log decode-error class reads zero on a re-run over the recordings the 07-22 census flagged
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.
R4 — Cookie import
Prerequisites: None declared; this does not establish execution eligibility.
Contract: ci_netscape
Acceptance: A Netscape cookies.txt and the operator's Firefox and Waterfox profile cookies import into the fetch jar with per-site domain filtering; a login-walled fixture page that returns 403 anonymous returns 200 with the imported jar; secrets never printed
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.
R5 — Segmented multi-connection download
Prerequisites: R0 (acceptance unverified)
Contract: dl_segments
Acceptance: nx_https_fetch_range composed into a segmented writer with N parallel ranges, resume from a partial file and a whole-file CID check at the end; measured throughput against the single-stream fetch on the same URL is published beside the rung
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.
R6 — DASH download
Prerequisites: R5 (acceptance unverified)
Contract: dg_fetch
Acceptance: An .mpd manifest parsed to representations, the best video plus audio fetched as segments and muxed to one playable file through the existing sovereign container path; a public DASH fixture round-trips and plays in the gallery
Authored effort: 1.5. 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.
R7 — Identity fan-out
Prerequisites: R0 (acceptance unverified), R1 (acceptance unverified)
Contract: if_probe
Acceptance: One username probed across N site patterns held as DATA rows (not code), in parallel under the pacer, every found profile emitted as a row with its site and the media count; a fixture of three sites with one planted hit and two planted misses partitions correctly and the partition sums
Authored effort: 1.5. 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.
R8 — Adapter census
Prerequisites: None declared; this does not establish execution eligibility.
Contract: sa_count
Acceptance: An organ COUNTS the site adapters (data rows plus per-source organs), proves each against its fixture and publishes the number on this page; a census that cannot prove an adapter counts it as UNPROVEN, never as reach
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.
R9 — Live recorder fleet
Prerequisites: R0 (acceptance unverified)
Contract: lr_watch
Acceptance: A supervisor polls a data-row list of performers across sites, hands a PUBLIC stream to the sovereign HLS chain the instant status flips, records continuously with per-site switches that stop or resume in-flight captures, survives an egress change, and writes poster plus NXVI at record time; measured against the retired Python recorder's corpus shape and proven on one live public stream before any cam site
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.
R10 — Emitted browser extension
Prerequisites: None declared; this does not establish execution eligibility.
Contract: ge_emit
Acceptance: An organ emits the Waterfox and Firefox MV2 extension and the Chrome and Edge MV3 twin from data (permissions, edge host, media patterns), the extension sniffs media in the tab and pushes one click to nx_vault_capture over the authenticated edge, and the emitted artifacts are served from nishifamily with their hashes published; the laptop-only nishi-grab is retired the day the emitted one loads
Authored effort: 1.5. 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.
R11 — Edge push without a companion
Prerequisites: R10 (acceptance unverified)
Contract: ge_edge_push
Acceptance: The extension needs no native host on Windows, Waterfox or any third-party browser because the estate does the capture; a fixture run from a Windows browser with no WSL and no installed binary lands an item in the vault; documented beside the native-messaging contract it deliberately avoids
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.
R12 — Egress rotation on owned hosts
Prerequisites: R0 (acceptance unverified)
Contract: pp_rotate
Acceptance: Rotation across sovereign-owned egress endpoints only (never residential botnets); the pace gate stays in the path and the policy is rows; shared contract with webscraping R3
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.
R13 — Bot-wall verdict in the capture path
Prerequisites: R0 (acceptance unverified), R1 (acceptance unverified)
Contract: vc_wall_verdict
Acceptance: Every capture verb runs nx_antibot abt_classify (the webscraping R4 contract, already landed and imported only by nx_web_crawl_step) over a non-2xx or challenge body and records the verdict as an outcome row feeding the source-health row; a planted Cloudflare specimen reads VENDOR_WALL and a clean 200 reads CLEAN (neg-control); never solved
Authored effort: 1.5. 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.
R14 — Video perceptual hash
Prerequisites: None declared; this does not establish execution eligibility.
Contract: vph_scene
Acceptance: A frame-sampled fingerprint per video (first-IDR decoder plus the dHash kernel) stored under the CID, a Hamming nearest over the corpus, and the vault consults it at ingest so a re-encoded mirror collapses; a fixture pair (source and re-encode) matches and a fixture pair of different videos does not
Authored effort: 1.5. 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.
R15 — Trickplay sprites at ingest
Prerequisites: R3 (acceptance unverified)
Contract: gt_sprite_sheet
Acceptance: A sprite sheet plus index derived per video in the analysis-on-ingest pass; an on-disk byte probe proves the sheet exists for every indexed video; shared contract with media R6
Authored effort: 1.5. 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.
R16 — Auto-tagger and captioner
Prerequisites: None declared; this does not establish execution eligibility.
Contract: ct_tag_item
Acceptance: A sovereign tagger writes measurements, tags and captions into the capture store for every new item (the integer judges first, a local VLM where one is admitted), with the judge panel and a pre-declared agreement bar against a hand-labelled set before it is wired to the ingest beat
Authored effort: 3. 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.
R17 — Semantic index
Prerequisites: R16 (acceptance unverified)
Contract: msm_search
Acceptance: A sovereign embedding index over the library with recall at 10 PRE-DECLARED to exceed BM25 on a labeled set or the lane stays unwired; shared contract with media R10
Authored effort: 3. 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.
R18 — Dataset export shards
Prerequisites: R16 (acceptance unverified)
Contract: ds_export_shards
Acceptance: Image and video-clip shards exported from vault queries with a manifest whose rows carry CID, source, licence tier and consent flag; a row without provenance is refused; the manifest partition must sum to the query count; the LoRA loop consumes it directly
Authored effort: 1.5. 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.
R19 — Ingest board
Prerequisites: R9 (acceptance unverified)
Contract: ib_page
Acceptance: One served page: queue, live fleet, per-site switches, history, failed-with-reason and dry-run, all read from the planes the organs already write; page_verify GREEN and the live judge over it
Authored effort: 1.5. 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.
R20 — Operator flag at ingest: one human tag row per asset on the accept plane, consent class attached
Prerequisites: R19 (acceptance unverified), R16 (acceptance unverified)
Contract: ct_operator_flag
Acceptance: OPERATOR 2026-09-02: the operator will flag faces and bodies at ingestion as part of the genetic research, and those flags are the population that aesthetictwin AT20 fits its beauty ranges from. Data, not a model: one knowledge/store/accept- row per flagged asset (id, asset cid, tagger identity, tag vocabulary key, consent or licence class, ts) written from the ingest board (R19) or the emitted extension (R10) in one click. R16's auto-tagger PROPOSES and the operator DISPOSES; the row records which. Done when a flagged asset shows its tag row on the ingest board, a re-ingest of the same cid does not duplicate the row (idempotent by cid+tagger), and nx_realperson fails closed on a tagged named person without a consent class. LAST MILE by operator order: placed in M5 after training so the ranker never puts it ahead of the engine gap.
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.
R21 — Provenance class gates USE, not just presence
Prerequisites: None declared; this does not establish execution eligibility.
Contract: pc_may_fit_axis
Acceptance: R18 already refuses a manifest row without provenance -- but nx_refcorpus's rc_put_row bounds the provenance and licence fields by LENGTH ONLY, so `unverified` and `cc0` were accepted identically and the word carried no consequence. R21 makes the class MEAN something: knowledge/asset_provenance.conf declares each tag with a per-class k-anonymity floor and a USE (shipped or reference-only), nx_provclass_lib is the ONE classifier both writer and reader compose, and pc_may_fit_axis answers the question the build-path rule structurally cannot -- keeping an asset out of the build path stops us shipping its BYTES, not its IDENTITY propagating through a generator fitted to it, and that output ships. Done-rule: nx_provclass_gate GREEN with the live journal population classifying (no existing row refused), the discriminating pair firing (same tag REFUSES at n=8 and ADMITS at n=32, so the raised floor is proven binding rather than decorative), and three named neg-controls. Feeds R18 and aesthetictwin AT18/AT20: a reference-only class may produce AGGREGATE STATISTICS ONLY and may never fit a beauty range
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.
R22 — Thread, blog and feed adapters beyond 4chan: Tumblr and Reddit first
Prerequisites: R0 (acceptance unverified), R4 (acceptance unverified), R8 (acceptance unverified), R13 (acceptance unverified)
Contract: sa_adapt_thread
Acceptance: OPERATOR 2026-10-08: download every media type from a thread like a 4chan /s/ thread (done the same day: 151 of 151 items, jpg + png + webm, committed to the vault records), then the same one-call ingest for Tumblr blogs, Reddit users and subreddits and the other sources the retired media manager reached, better designed. One adapter ABC (nx_source_adapter) with list and fetch verbs over DATA rows per site (URL grammar, JSON or HTML shape, media selectors, pacing class), composing the pacer (R0), the cookie jar for login walls (R4), the bot-wall verdict (R13) and the vault commit; a fixture per site (one planted thread carrying every media kind) round-trips; the adapter census (R8) counts each proven adapter; the gallery panel offers the same download-all for every adapter and a loopback client drives the authenticated ingest route. Done when the Tumblr and Reddit fixtures ingest end to end and the census publishes the reach
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.

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 mediaingest 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.

R0 Pacing wired into the capture chokepoint · cell PRESENT · gate NO-GATE-DECLARED

Contract	`vc_pace_hook` · executor Organ · 0.5 u · stands on nothing
Definition of done	nx_vault_capture, nx_mvault_fetch, nx_hls_get and nx_media_grab call ONE hook composing pace_before and pace_after around every fetch; a planted 429 with Retry-After on a fixture host is honoured and the next request waits; the gate proves the wait arithmetically and a neg-control proves an un-hooked fetch does not wait; no second pacing ruler is written
Matrix cell	ACQUIRE: pacing wired into the CAPTURE chokepoint -- vault capture, album fetch, HLS segments, grab · `runtime/nx_vault_capture.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_vault_capture`; declare `nx_vault_capture TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R1 Source health pre-flight · cell PRESENT · gate NO-GATE-DECLARED

Contract	`sh_preflight` · executor Organ · 0.5 u · stands on R0
Definition of done	Before queueing work against a host, a bounded probe classifies it UP, SLOW, NULL-ROUTED or WALLED and the verdict is a row the queue reads; a planted dead host is skipped with its reason named; results persist so a flapping CDN is not re-probed on every item
Matrix cell	ACQUIRE: source health pre-flight -- fast-fail a null-routed CDN before queueing work against it · `runtime/nx_source_health.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_source_health`; declare `nx_source_health TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R2 Disk floor at ingest · cell PRESENT · gate NO-GATE-DECLARED

Contract	`ig_disk_floor` · executor Organ · 0.5 u · stands on nothing
Definition of done	Every capture verb reads the disk axis nx_syscalls already carries and REFUSES below a floor declared in conf, announcing the free bytes and the floor; the gate plants a floor above the current free space and proves the refusal fires
Matrix cell	CATALOGUE: disk budget guard at ingest -- refuse below a named floor, the refusal announces · `runtime/nx_ingest_admit.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_ingest_admit`; declare `nx_ingest_admit TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R3 Thumbnail decoder coverage · cell PRESENT · gate NO-GATE-DECLARED

Contract	`th_decode_any` · executor Organ · 1 u · stands on nothing
Definition of done	nx_galx_thumb composes the in-tree JPEG, GIF and WebP decoders and handles palette and 16-bit PNG; a fixture set of one file per format thumbnails end to end; the galx_pipeline.log decode-error class reads zero on a re-run over the recordings the 07-22 census flagged
Matrix cell	CATALOGUE: thumbnail decoder coverage -- JPEG, WebP, GIF, palette and 16-bit PNG · `runtime/nx_galx_thumb.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_galx_thumb`; declare `nx_galx_thumb TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R4 Cookie import · cell WATCHING · gate NO-GATE-DECLARED

Contract	`ci_netscape` · executor Organ · 1 u · stands on nothing
Definition of done	A Netscape cookies.txt and the operator's Firefox and Waterfox profile cookies import into the fetch jar with per-site domain filtering; a login-walled fixture page that returns 403 anonymous returns 200 with the imported jar; secrets never printed
Matrix cell	ACQUIRE: import cookies from the operator's browser or a cookies.txt · `runtime/nx_cookie_import.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_cookie_import`; declare `nx_cookie_import TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R5 Segmented multi-connection download · cell WATCHING · gate NO-GATE-DECLARED

Contract	`dl_segments` · executor Organ · 1 u · stands on R0
Definition of done	nx_https_fetch_range composed into a segmented writer with N parallel ranges, resume from a partial file and a whole-file CID check at the end; measured throughput against the single-stream fetch on the same URL is published beside the rung
Matrix cell	ACQUIRE: multi-connection segmented download with resume (aria2c-class) · `runtime/nx_dl_segmented.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_dl_segmented`; declare `nx_dl_segmented TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R6 DASH download · cell WATCHING · gate NO-GATE-DECLARED

Contract	`dg_fetch` · executor Organ · 1.5 u · stands on R5
Definition of done	An .mpd manifest parsed to representations, the best video plus audio fetched as segments and muxed to one playable file through the existing sovereign container path; a public DASH fixture round-trips and plays in the gallery
Matrix cell	ACQUIRE: DASH (.mpd) download · `runtime/nx_dash_get.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_dash_get`; declare `nx_dash_get TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R7 Identity fan-out · cell WATCHING · gate NO-GATE-DECLARED

Contract	`if_probe` · executor Organ · 1.5 u · stands on R0, R1
Definition of done	One username probed across N site patterns held as DATA rows (not code), in parallel under the pacer, every found profile emitted as a row with its site and the media count; a fixture of three sites with one planted hit and two planted misses partitions correctly and the partition sums
Matrix cell	ACQUIRE: identity fan-out -- one username probed across N sites in parallel, every profile found · `runtime/nx_ident_fanout.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_ident_fanout`; declare `nx_ident_fanout TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R8 Adapter census · cell PRESENT · gate NO-GATE-DECLARED

Contract	`sa_count` · executor Organ · 0.5 u · stands on nothing
Definition of done	An organ COUNTS the site adapters (data rows plus per-source organs), proves each against its fixture and publishes the number on this page; a census that cannot prove an adapter counts it as UNPROVEN, never as reach
Matrix cell	ACQUIRE: extractor breadth counted, not claimed -- adapter rows proven by fixture · `runtime/nx_site_adapters.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_site_adapters`; declare `nx_site_adapters TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R9 Live recorder fleet · cell WATCHING · gate NO-GATE-DECLARED

Contract	`lr_watch` · executor Organ · 3 u · stands on R0
Definition of done	A supervisor polls a data-row list of performers across sites, hands a PUBLIC stream to the sovereign HLS chain the instant status flips, records continuously with per-site switches that stop or resume in-flight captures, survives an egress change, and writes poster plus NXVI at record time; measured against the retired Python recorder's corpus shape and proven on one live public stream before any cam site
Matrix cell	ACQUIRE: live-stream auto-record fleet -- poll status, record on PUBLIC, per-site switches, continuity across an exit change · `runtime/nx_live_recorder.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_live_recorder`; declare `nx_live_recorder TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R10 Emitted browser extension · cell LANDED · gate NO-GATE-DECLARED

Contract	`ge_emit` · executor Organ · 1.5 u · stands on nothing
Definition of done	An organ emits the Waterfox and Firefox MV2 extension and the Chrome and Edge MV3 twin from data (permissions, edge host, media patterns), the extension sniffs media in the tab and pushes one click to nx_vault_capture over the authenticated edge, and the emitted artifacts are served from nishifamily with their hashes published; the laptop-only nishi-grab is retired the day the emitted one loads
Matrix cell	SURFACE: third-party browser extension EMITTED from the estate -- Waterfox and Firefox (MV2) plus Chrome and Edge (MV3), one-click push to the sovereign stack · `runtime/nx_grab_ext_emit.nx` · LANDED decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_grab_ext_emit`; declare `nx_grab_ext_emit TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R11 Edge push without a companion · cell WATCHING · gate NO-GATE-DECLARED

Contract	`ge_edge_push` · executor Organ · 0.5 u · stands on R10
Definition of done	The extension needs no native host on Windows, Waterfox or any third-party browser because the estate does the capture; a fixture run from a Windows browser with no WSL and no installed binary lands an item in the vault; documented beside the native-messaging contract it deliberately avoids
Matrix cell	SURFACE: Windows and third-party browsers served WITHOUT a native companion -- the extension talks to the estate over the authenticated edge · `runtime/nx_grab_ext_emit.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_grab_ext_emit`; declare `nx_grab_ext_emit TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R12 Egress rotation on owned hosts · cell WATCHING · gate NO-GATE-DECLARED

Contract	`pp_rotate` · executor Organ · 1 u · stands on R0
Definition of done	Rotation across sovereign-owned egress endpoints only (never residential botnets); the pace gate stays in the path and the policy is rows; shared contract with webscraping R3
Matrix cell	ACQUIRE: rate-limit storm response -- exit rotation on OWNED egress, per-site pinning, ride-through of in-flight recordings · `runtime/nx_proxy_pool.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_proxy_pool`; declare `nx_proxy_pool TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R13 Bot-wall verdict in the capture path · cell WATCHING · gate NO-GATE-DECLARED

Contract	`vc_wall_verdict` · executor Organ · 1.5 u · stands on R0, R1
Definition of done	Every capture verb runs nx_antibot abt_classify (the webscraping R4 contract, already landed and imported only by nx_web_crawl_step) over a non-2xx or challenge body and records the verdict as an outcome row feeding the source-health row; a planted Cloudflare specimen reads VENDOR_WALL and a clean 200 reads CLEAN (neg-control); never solved
Matrix cell	ACQUIRE: bot-wall verdict recorded in the CAPTURE path -- Cloudflare, DDoS-Guard, Turnstile named on every failed fetch, never solved · `runtime/nx_vault_capture.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_vault_capture`; declare `nx_vault_capture TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R14 Video perceptual hash · cell WATCHING · gate NO-GATE-DECLARED

Contract	`vph_scene` · executor Organ · 1.5 u · stands on nothing
Definition of done	A frame-sampled fingerprint per video (first-IDR decoder plus the dHash kernel) stored under the CID, a Hamming nearest over the corpus, and the vault consults it at ingest so a re-encoded mirror collapses; a fixture pair (source and re-encode) matches and a fixture pair of different videos does not
Matrix cell	CATALOGUE: video perceptual hash -- frame-sampled fingerprint so re-encoded mirrors collapse · `runtime/nx_vphash.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_vphash`; declare `nx_vphash TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R15 Trickplay sprites at ingest · cell WATCHING · gate NO-GATE-DECLARED

Contract	`gt_sprite_sheet` · executor Organ · 1.5 u · stands on R3
Definition of done	A sprite sheet plus index derived per video in the analysis-on-ingest pass; an on-disk byte probe proves the sheet exists for every indexed video; shared contract with media R6
Matrix cell	CATALOGUE: trickplay scrub-sprite sheet derived at ingest · `runtime/nx_galx_trickplay.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_galx_trickplay`; declare `nx_galx_trickplay TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R16 Auto-tagger and captioner · cell WATCHING · gate NO-GATE-DECLARED

Contract	`ct_tag_item` · executor Local model · 3 u · stands on nothing
Definition of done	A sovereign tagger writes measurements, tags and captions into the capture store for every new item (the integer judges first, a local VLM where one is admitted), with the judge panel and a pre-declared agreement bar against a hand-labelled set before it is wired to the ingest beat
Matrix cell	TRAIN: auto-tagger and captioner producing that capture (VLM plus the integer judges) · `runtime/nx_caption_tagger.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_caption_tagger`; declare `nx_caption_tagger TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R17 Semantic index · cell WATCHING · gate NO-GATE-DECLARED

Contract	`msm_search` · executor Local model · 3 u · stands on R16
Definition of done	A sovereign embedding index over the library with recall at 10 PRE-DECLARED to exceed BM25 on a labeled set or the lane stays unwired; shared contract with media R10
Matrix cell	TRAIN: sovereign encoder producing embeddings and semantic search over the library · `runtime/nx_media_semantic.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_media_semantic`; declare `nx_media_semantic TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R18 Dataset export shards · cell WATCHING · gate NO-GATE-DECLARED

Contract	`ds_export_shards` · executor Organ · 1.5 u · stands on R16
Definition of done	Image and video-clip shards exported from vault queries with a manifest whose rows carry CID, source, licence tier and consent flag; a row without provenance is refused; the manifest partition must sum to the query count; the LoRA loop consumes it directly
Matrix cell	TRAIN: dataset export with provenance and licence per item -- image and video-clip shards, a manifest that reconciles · `runtime/nx_dataset_export.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_dataset_export`; declare `nx_dataset_export TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

R19 Ingest board · cell WATCHING · gate NO-GATE-DECLARED

Contract	`ib_page` · executor Organ · 1.5 u · stands on R9
Definition of done	One served page: queue, live fleet, per-site switches, history, failed-with-reason and dry-run, all read from the planes the organs already write; page_verify GREEN and the live judge over it
Matrix cell	SURFACE: operator dashboard -- queue, live fleet, per-site switches, history, failed-with-reason, dry-run · `runtime/nx_ingest_board.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_ingest_board`; declare `nx_ingest_board TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-10 queue 2026-09-10 operator request retained, not implemented or visually verified: local/remote/rented workers return source provenance, inference uncertainty, quality/test receipts and reviewable selection decisions; NAS coordinates and retains durable results rather than assuming heavy compute placement. Link containers CT1 and gameengine GE61.

R20 Operator flag at ingest: one human tag row per asset on the accept plane, consent class attached · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`ct_operator_flag` · executor Organ · 2 u · stands on R19, R16
Definition of done	OPERATOR 2026-09-02: the operator will flag faces and bodies at ingestion as part of the genetic research, and those flags are the population that aesthetictwin AT20 fits its beauty ranges from. Data, not a model: one knowledge/store/accept- row per flagged asset (id, asset cid, tagger identity, tag vocabulary key, consent or licence class, ts) written from the ingest board (R19) or the emitted extension (R10) in one click. R16's auto-tagger PROPOSES and the operator DISPOSES; the row records which. Done when a flagged asset shows its tag row on the ingest board, a re-ingest of the same cid does not duplicate the row (idempotent by cid+tagger), and nx_realperson fails closed on a tagged named person without a consent class. LAST MILE by operator order: placed in M5 after training so the ranker never puts it ahead of the engine gap.
Matrix cell	NO-MATRIX-ROW — no matrix row carries `ct_operator_flag`; the ranker reads this rung as unmapped and no cell can flip for it. Add the row and the join completes on the next publish.
Gate	the gate join needs a matrix row: no organ is named for this rung
Work log	2026-09-02 queue nothing built by operator order (last mile). The consumer rung is aesthetictwin AT20; the fit machinery is aesthetictwin AT18; the tag vocabulary is a data file to be admitted with the first flag, never typed in an organ.

R21 Provenance class gates USE, not just presence · cell PRESENT · gate NO-GATE-DECLARED

Contract	`pc_may_fit_axis` · executor Organ · 0.5 u · stands on nothing
Definition of done	R18 already refuses a manifest row without provenance -- but nx_refcorpus's rc_put_row bounds the provenance and licence fields by LENGTH ONLY, so `unverified` and `cc0` were accepted identically and the word carried no consequence. R21 makes the class MEAN something: knowledge/asset_provenance.conf declares each tag with a per-class k-anonymity floor and a USE (shipped or reference-only), nx_provclass_lib is the ONE classifier both writer and reader compose, and pc_may_fit_axis answers the question the build-path rule structurally cannot -- keeping an asset out of the build path stops us shipping its BYTES, not its IDENTITY propagating through a generator fitted to it, and that output ships. Done-rule: nx_provclass_gate GREEN with the live journal population classifying (no existing row refused), the discriminating pair firing (same tag REFUSES at n=8 and ADMITS at n=32, so the raised floor is proven binding rather than decorative), and three named neg-controls. Feeds R18 and aesthetictwin AT18/AT20: a reference-only class may produce AGGREGATE STATISTICS ONLY and may never fit a beauty range
Matrix cell	PROVENANCE: the provenance class gates USE, not just presence · `runtime/nx_provclass_lib.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_provclass_lib`; declare `nx_provclass_lib TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-03 land conf + lib + gate landed on the NAS 2026-09-03 (asset_provenance.conf 3,244 B, 12 classes; nx_provclass_lib 9,251 B sha 7a43dacd; nx_provclass_gate 7,139 B sha b8b09288). 15/16 on the laptop -- the ONE failure is the anti-vacuity tooth journal-population-was-actually-read, because the journal is NAS-side; it caught every-live-journal-licence-tag-classifies scoring a VACUOUS actual=0 expected=0 PASS on first run, on its own author. NAS run owed once build admission clears. 2026-09-03 measure the live refcorpus journal carries 47 rows and FIVE undeclared licence vocabularies (ORIGINAL, sovereign, royalty-free, CC-BY-mixed, REFERENCE_MEASURED). All five are declared in the conf, so the new wall refuses no existing row -- measured directly, not asserted, and the gate will assert it on the NAS. 2026-09-03 lesson OPERATOR DECISION 2026-09-03: SFMLab / Open3DLab / SmutBase content is admitted as class `unverified` -- reference-only, k-floor 32. One platform, three front doors, user-uploaded with reactive-DMCA enforcement, so a licence is not restrictive but ABSENT. At n>=32 a distribution fact is a fact about a population and carries no copyright; at n=8 it can still describe one work. The sites sit behind Anubis proof-of-work and NO solver was built: webscraping R4 already declared that boundary before this seat arrived.

R22 Thread, blog and feed adapters beyond 4chan: Tumblr and Reddit first · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`sa_adapt_thread` · executor Organ · 2 u · stands on R0, R4, R8, R13
Definition of done	OPERATOR 2026-10-08: download every media type from a thread like a 4chan /s/ thread (done the same day: 151 of 151 items, jpg + png + webm, committed to the vault records), then the same one-call ingest for Tumblr blogs, Reddit users and subreddits and the other sources the retired media manager reached, better designed. One adapter ABC (nx_source_adapter) with list and fetch verbs over DATA rows per site (URL grammar, JSON or HTML shape, media selectors, pacing class), composing the pacer (R0), the cookie jar for login walls (R4), the bot-wall verdict (R13) and the vault commit; a fixture per site (one planted thread carrying every media kind) round-trips; the adapter census (R8) counts each proven adapter; the gallery panel offers the same download-all for every adapter and a loopback client drives the authenticated ingest route. Done when the Tumblr and Reddit fixtures ingest end to end and the census publishes the reach
Matrix cell	NO-MATRIX-ROW — no matrix row carries `sa_adapt_thread`; the ranker reads this rung as unmapped and no cell can flip for it. Add the row and the join completes on the next publish.
Gate	the gate join needs a matrix row: no organ is named for this rung
Work log	2026-10-08 measure 2026-10-08 thread /s/22460514: nx_4chan thread listed the media, nx_4chan fetch downloaded 151 of 151 (143 jpg, 7 png, 1 webm, 0 failed) into media/4chan/s_22460514, and nx_mvault_walk real 4chan commit recorded 151 unique items (dup 0, commits 1, fatal 0) into the mvault record store (knowledge/store/mvault-, NXR1 records, the /vault gateway's surface). The gallery grid's view index did not change (226,595 records before and after): vault records are a separate surface from the gallery's images and recordings, so a one-click download lands in /vault, not in the /gallery grid -- a surface decision to make explicit on this board, not a defect. 2026-10-08 land the gallery backend deployed 2026-10-08 (galxserve, sha a9d8d1d7, first health-checked deploy of this daemon through /api/deploy -- its deploy_domain row galxserve mediaingest was missing and the GLOBAL fallback blocked on eight unrelated RED domains until the row was added) carries the thread view's download-all-to-library link (/api/4chan/panel); the fetch and the commit were driven as two sanctioned jobs because no loopback client exists for the authenticated POST /api/4chan/ingest -- the shape the adapter ABC must give every source.

dossiers=23 cells_joined=21 no_matrix_row=2 (sum 23) gates_recorded=0 gates_unrecorded=0 no_gate_declared=21 (sum 21 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. 66 feature rows mapped of 66 matrix rows; 0 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/mediaingest.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 400–900, distinctiveness 200–500 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 0 (a fact of the data: two players that close are that close) · data spread 741 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 400–900 · distinctiveness 200–500 · breadth 200–700 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 1 · unresolved labels 0 · mark overlaps 6 (a fact of the data: two players that close are that close) · data spread 219 permil of the plot

Nishi Harvestr ctbrec yt-dlp Video DownloadHelper Stash Hydrus Network
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 12 (a fact of the data: two players that close are that close) · data spread 552 permil of the plot

Player	Quadrant	centrality	distinctiveness	breadth	depth	momentum	Rows held	Days on spine	First seen
Nishi	Aspirational	841	455	601	762	0 since 2026-09-23	52	10	2026-09-15
Harvestr	Aspirational	669	392	429	833	0 since 2026-09-23	34	10	2026-09-15
ctbrec	Peripheral	502	229	247	742	0 since 2026-09-23	22	10	2026-09-15
yt-dlp	Mainstream	673	324	393	812	0 since 2026-09-23	32	10	2026-09-15
Video DownloadHelper	Unconventional	476	396	308	884	0 since 2026-09-23	23	10	2026-09-15
Stash	Unconventional	523	454	368	869	0 since 2026-09-23	28	10	2026-09-15
Hydrus Network	Aspirational	772	437	540	849	0 since 2026-09-23	42	10	2026-09-15

Day	Matrix rows reviewed	Players recorded
2026-09-15	66	7
2026-09-16	66	7
2026-09-17	66	7
2026-09-18	66	7
2026-09-19	66	7
2026-09-20	66	7
2026-09-21	66	7
2026-09-22	66	7
2026-09-23	66	7
2026-10-08	66	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 396, unassessed 0. Authored assessments are not verified outcomes.

Capability	Nishi	Harvestr	ctbrec	yt-dlp	Video DownloadHelper	Stash	Hydrus Network
ACQUIRE
sovereign HTTPS fetch -- TLS 1.3 and 1.2 reach, redirects, gzip, Chrome-JA3 hello for CDN fingerprint walls Measured: `nx_https_fetch_follow_chrome` exists in `runtime/nx_https_fetch_follow.nx`, verified at emit. Harvestr leans on curl_cffi Chrome impersonation plus cloudscraper [harvestr]; yt-dlp --impersonate rides the same curl_cffi [ytdlp]; ours is our own TLS with a Chrome-shaped hello (measured 2026-08-12: a 4chan bulk failure was RATE, not fingerprint -- the hello already passes). Adoption: `LIB-WIRED importers=241 nonval=217` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind Harvestr (leads), yt-dlp (leads)	●	◉	●	◉	●	●	●
cascade fetch that can never return silent-empty -- challenge pages rejected, per-strategy attempts log always written Source present; authored stronger claim: `fa_is_real` in `runtime/nx_fetch_any.nx`, verified at emit. EXCEED with the why: the attempts log is DATA the caller banks on failure so absence becomes a record, never silence; Harvestr logs to its dashboard console [harvestr] and yt-dlp prints warnings [ytdlp] -- neither hands the failed strategy ladder back as an artifact. 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 Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus NetworkCons not adopted yet: compiled, never promoted to the serving root: /api/promote it	◉	●	○	●	○	○	●
per-host adaptive pacing -- Retry-After honoured, exponential backoff, decay on success, table persisted across processes Measured: `pace_after` exists in `runtime/_hdl_build/nx_crawl_pace.nx`, verified at emit. Hydrus per-domain bandwidth rules are the published bar [hydrus-dl]; RFC 6585 429 plus Retry-After is the signal [rfc6585]; Harvestr backs off with jitter and honours Retry-After inside its scraper base class and adds a tiered VPN watchdog above it [harvestr]; yt-dlp --sleep-requests and --retry-sleep [ytdlp]. MEASURED 2026-08-30 (corpus_complete=1): importers are nx_4chan, nx_web_crawl_step, nx_url_index, nx_bulk_index, nx_domain_map -- the crawl lane and one board adapter. Adoption: `PROMOTED-UNREGISTERED` — PARTIAL: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: a real binary nobody can call over MCP: /api/tools/register it; 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 ctbrec, Video DownloadHelper, StashCons behind Hydrus Network (leads); not adopted yet: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	●	●	○	●	○	○	◉
pacing wired into the CAPTURE chokepoint -- vault capture, album fetch, HLS segments, grab Source present; authored stronger claim: `vc_pace_hook` in `runtime/nx_vault_capture.nx`, verified at emit. LANDED 2026-08-30 (R0): nx_vault_capture (vc_pace_hook), nx_mvault_fetch and nx_hls_get (burst form) now wrap every fetch in nx_paced_fetch (pf_before/pf_after), which delegates to nx_crawl_pace -- one ruler, no second copy; PROMOTED LIVE, gate nx_paced_fetch_gate 14/14 GREEN. nx_media_grab is the 4th verb and its pacing is DEFERRED (adding the import tipped its deep-crawl closure past the nx_cc module limit; tracked, not a silent gap). EXCEED with the why: fixing this uncovered and fixed a latent FAIL-OPEN in the pacer itself -- the 4096-slot table had no eviction, so once full (measured: nishihost's table at 4096/4096 real hosts) every NEW host got no pacing; R15 adds LRU-by-next_allowed eviction on the write path, proven by the gate reporting occupied=4096 and the write still landing. Harvestr honours Retry-After per site [harvestr]; RFC 6585 429 + Retry-After is the signal [rfc6585]; no rival's per-host limiter is shown to survive its own table filling. 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 Source symbol found; stronger capability claim is authored; ahead of ctbrec, Video DownloadHelper, StashCons 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)	◉	◉	●	◉	○	○	◉
rate-limit storm response -- exit rotation on OWNED egress, per-site pinning, ride-through of in-flight recordings Open — watching `runtime/nx_proxy_pool.nx` : `pp_rotate`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Harvestr: a tiered watchdog restarts the affected bots first and rotates the Mullvad exit only if limits keep climbing, pins one site to a residential proxy, and keeps writing the same file across the rotation gap [harvestr]; yt-dlp is --proxy only [ytdlp]. Shared contract with webscraping R3 -- owned egress only, never residential botnets; the sovereign Tor client (nx_tor_relay) is a candidate egress. dossierwatching pp_rotate Pros none measured yetCons behind Harvestr (leads), yt-dlp; open contract; expected source declaration not established	○	◉	○	●	○	○	○
bot-wall verdict recorded in the CAPTURE path -- Cloudflare, DDoS-Guard, Turnstile named on every failed fetch, never solved DARK — `runtime/nx_vault_capture.nx` EXISTS but does not declare `vc_wall_verdict`: 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. Harvestr: curl_cffi plus cloudscraper plus a stealth Chromium tier that auto-passes invisible Turnstile plus the Accept text/css trick for DDoS-Guard [harvestr]. MEASURED 2026-08-30 (corpus_complete=1): the classifier EXISTS (nx_antibot abt_classify, the webscraping R4 contract, landed) and is imported by nx_web_crawl_step only -- the capture verbs never ask it, so a wall reads as a bare failed fetch. Contract: every capture verb classifies a non-2xx or challenge body and records the verdict as an outcome row; ethics by construction: classify and report, no CAPTCHA farm, robots and terms respected. dossierwatching vc_wall_verdict Pros none measured yetCons behind Harvestr (leads), ctbrec, yt-dlp; open contract; expected source declaration not established	○	◉	●	●	○	○	○
source health pre-flight -- fast-fail a null-routed CDN before queueing work against it Measured: `sh_preflight` exists in `runtime/nx_source_health.nx`, verified at emit. LANDED 2026-08-30 (R1): sh_preflight (nx_source_health) classifies a host UP, SLOW, DEAD or WALLED and PERSISTS the verdict with a 30-min TTL so a flapping CDN is not re-probed per item; sh_should_skip fast-fails DEAD and WALLED before work is queued. Gate nx_source_health_gate 12/12 GREEN -- all five verdicts planted-number, a walled-vs-up neg-control, and a classify then persist then lookup then TTL-expiry then overwrite round-trip, no network. The bounded remote probe nx_source_health_probe composes the Chrome-hello fetch and nx_antibot: LIVE example.com read UP, an unresolvable .invalid host read DEAD skip=1, and a per-vantage-closed 443 read DEAD -- reachability is a property of the vantage. WALLED is emitted only on a hard-block status at or above 400 so a challenged 2xx is never over-blocked; per-capture wall detection is R13. Harvestr routes the April-2026 Coomer null-route to mirrors [harvestr]; we now fast-fail it, and mirror auto-routing is the exceed path. Adoption: `LIB-GATE-ONLY importers=3` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. dossiernot adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of ctbrec, yt-dlp, Video DownloadHelper, StashCons behind Harvestr (leads); not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	◉	○	○	○	○	●
identity fan-out -- one username probed across N sites in parallel, every profile found Open — watching `runtime/nx_ident_fanout.nx` : `if_probe`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Harvestr's core verb: 32 site URL patterns as data, 8-way parallel probing, about 15 s for 50 sites, first hit wins [harvestr]. Ours gathers one page for one entity (nx_media_gather) and has no fan-out. dossierwatching if_probe Pros none measured yetCons behind Harvestr (leads); open contract; expected source declaration not established	○	◉	○	○	○	○	○
site adapters as DATA rows, not code -- a new album site is an INSERT Measured: `mvf_fetch_album_pfx` exists in `runtime/nx_mvault_fetch.nx`, verified at emit. Hydrus downloader parsers and URL classes are shareable data objects [hydrus-dl]; Harvestr splits it: URL patterns in sites.json plus 39 scraper classes in Python [harvestr]; yt-dlp extractors are code, 1800+ of them [ytdlp]; Stash scrapers are YAML plus script packages [stash]. Ours: album pages described by where the links live, two-hop by design. Adoption: `LIB-WIRED importers=3 nonval=2` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper; fully adopted on the estate ladderCons behind Hydrus Network (leads)	●	◐	○	○	○	●	◉
board and gallery adapters shipped -- 4chan browse, preview, download with retry and backoff Measured: `fc_fetch_one` exists in `runtime/nx_4chan.nx`, verified at emit. gallery-dl is the image-gallery bar for reddit, tumblr, pixiv and boards [gallerydl]; the retired Python media manager had reddit, tumblr, twitter and telegram scrapers -- only 4chan is ported (its source header names it the first per-source ingester off Python). 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 ctbrec, StashCons behind Harvestr (leads), yt-dlp (leads), Hydrus Network (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)	●	◉	○	◉	●	○	◉
extractor breadth counted, not claimed -- adapter rows proven by fixture Measured: `sa_count` exists in `runtime/nx_site_adapters.nx`, verified at emit. LANDED 2026-08-30 (R8): nx_site_adapters counts BOTH adapter kinds -- runtime album-adapter rows in the mvfetch- plane (a new site is an INSERT) and per-source ingester organs declared in source_adapters.conf (nx_4chan today) -- every number DERIVED (reg_index_len + conf line count), never hardcoded; gate nx_site_adapters_gate proves the derivation against an independent oracle with a comment-and-blank-count-zero neg-control. The honest count is SMALL and the page says so: this turns extractor breadth from a claim into a number a reach campaign (R7 fan-out + more adapter rows) drives up. yt-dlp 1800+ sites [ytdlp]; Harvestr 32 patterns plus 39 scrapers [harvestr]; VDH is site-agnostic by watching the tab's traffic [vdhcoapp]. 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 separateCons behind Harvestr (leads), yt-dlp (leads), Video DownloadHelper (leads), Hydrus Network (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)	●	◉	●	◉	◉	●	◉
in-page media discovery -- lazy attrs, source tags, HLS and DASH manifests, JSON blobs, CSS url() Measured: `nx_media_harvest` exists in `runtime/nx_media_harvest.nx`, verified at emit. VDH observes the tab's real requests; yt-dlp's generic extractor parses page HTML [ytdlp]; ours classifies every hit into image, gif, video, stream or audio. Adoption: `LIB-WIRED importers=6 nonval=5` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of ctbrec, Stash; fully adopted on the estate ladderCons behind yt-dlp (leads), Video DownloadHelper (leads)	●	●	○	◉	◉	○	●
deep resource-graph crawl -- CSS and JS followed, JS-built URLs folded WITHOUT executing page JS Measured: `nx_media_deepcrawl` exists in `runtime/nx_media_deepcrawl.nx`, verified at emit. Budget-bounded (maxfetch, maxdepth), robust on minified ad bundles; the same crawler the in-browser m key uses. 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=4` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of ctbrec, Stash, Hydrus Network; fully adopted on the estate ladderCons behind Video DownloadHelper (leads)	●	●	○	●	◉	○	○
JS-execution harvest with ZERO network -- page JS run on the sovereign engine, constructed fetch URLs captured pending Source present; authored stronger claim: `mjx_harvest_js` in `runtime/_hdl_build/nx_media_jsexec.nx`, verified at emit. EXCEED with the why: fetches PEND, so running a page cannot hit a rate limit or be fingerprinted by request pattern; Harvestr and VDH capture from a REAL browser session with real requests [harvestr] [vdhcoapp]. Honest limit in source: the interpreter has no fuel budget yet, so a hostile while(true) blocks LIVE wiring -- declared, not hidden. Adoption: `LIB-WIRED importers=9 nonval=2` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of Harvestr, ctbrec, yt-dlp, Stash, Hydrus Network; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	◐	○	●	◉	○	○
site-resilient media-URL classifier -- extension, path pattern, CDN host and token signals, not per-site JSON paths Measured: `xt_classify` exists in `runtime/nx_media_signal.nx`, verified at emit. Per-site extractors break when a site restructures; this keys off universal signals and ranks by confidence, so one broken signal still resolves through the others. 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=27 nonval=12` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Stash; fully adopted on the estate ladderCons behind yt-dlp (leads), Video DownloadHelper (leads)	●	●	●	◉	◉	○	●
self-healing extraction table -- a learned pattern flows through without a code edit Source present; authored stronger claim: `xh_learn` in `runtime/_hdl_build/nx_extract_heal.nx`, verified at emit. EXCEED with the why: on a miss a critic-confirmed media URL is LOGGED and LEARNED (a host or path pattern appended to the table and saved), so the recogniser adapts from its own failures; Hydrus lets a person edit parsers by hand [hydrus-dl]; no rival adapts its recogniser automatically. Adoption: `LIB-WIRED importers=7 nonval=2` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus Network; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	●
discovery to download plan for JS-hardened video -- m3u8 to hls_get, mp4 to video_get, DRM named as the wall Measured: `vs_sniff_page` exists in `runtime/nx_video_sniff.nx`, verified at emit. Honest capability line in source: obfuscation beaten by executing the bundle, AES-128 HLS beaten, Widevine and PlayReady and FairPlay EME never circumvented. 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=23 nonval=12` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Stash, Hydrus Network; fully adopted on the estate ladderCons behind Harvestr (leads), yt-dlp (leads), Video DownloadHelper (leads)	●	◉	●	◉	◉	○	○
HLS download -- master to variant pick, AES-128 segment decrypt, TS stitch, fork-per-segment memory bound Measured: `hg_parse_key` exists in `runtime/nx_hls_get.nx`, verified at emit. RFC 8216 [rfc8216]; Harvestr and ctbrec hand HLS to ffmpeg [harvestr]; yt-dlp hlsnative or ffmpeg [ytdlp]; fMP4 EXT-X-MAP init segments are declared in source as the next rung. Adoption: `PROMOTED-UNREGISTERED` — PARTIAL: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: a real binary nobody can call over MCP: /api/tools/register it; 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 Stash, Hydrus NetworkCons behind Harvestr (leads), ctbrec (leads), yt-dlp (leads), Video DownloadHelper (leads); not adopted yet: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	●	◉	◉	◉	◉	○	○
DASH (.mpd) download Open — watching `runtime/nx_dash_get.nx` : `dg_fetch`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Harvestr via ffmpeg [harvestr]; yt-dlp native DASH with -N parallel fragments [ytdlp]; ours already RANKS a DASH manifest (XT_DASH in the signal classifier) and nothing fetches it. dossierwatching dg_fetch Pros none measured yetCons behind Harvestr (leads), yt-dlp (leads), Video DownloadHelper (leads); open contract; expected source declaration not established	○	◉	○	◉	◉	○	○
multi-connection segmented download with resume (aria2c-class) Open — watching `runtime/nx_dl_segmented.nx` : `dl_segments`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Harvestr aria2c 16 connections [harvestr]; yt-dlp -N and an external aria2c [ytdlp]; the primitive nx_https_fetch_range exists and nothing composes it into a segmented writer. dossierwatching dl_segments Pros none measured yetCons behind Harvestr (leads), yt-dlp (leads), Video DownloadHelper; open contract; expected source declaration not established	○	◉	○	◉	●	○	○
streaming download straight to disk -- TLS record to file fd, multi-GB never buffers in RAM, Referer and Cookie honoured Measured: `nx_video_fetch` exists in `runtime/nx_video_get.nx`, verified at emit. The engine both the Waterfox extension and the native browser feed (source header); CDN redirects followed, a real Firefox UA sent. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `SOURCE-ONLY` — PARTIAL: source exists, never compiled: /api/build it. not adopted: source exists, never compiled: /api/build it Pros Implementing source symbol found; runtime acceptance separate; ahead of StashCons not adopted yet: source exists, never compiled: /api/build it	●	●	●	●	●	○	●
live-stream auto-record fleet -- poll status, record on PUBLIC, per-site switches, continuity across an exit change Open — watching `runtime/nx_live_recorder.nx` : `lr_watch`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. ctbrec and Harvestr's vendored StreaMonitor cover 18 cam platforms with a 5-30 s poll loop and hand the stream to ffmpeg the instant status flips [harvestr]; yt-dlp --live-from-start is one stream at a time [ytdlp]. Ours: the 53,567-recording cam corpus (nx_media_inventory, generated 2026-07-17) came from the retired Python recorder; the sovereign HLS chain is proven on a live public stream (nx_hls_live_probe) and no fleet recorder exists. dossierwatching lr_watch Pros none measured yetCons behind Harvestr (leads), ctbrec (leads), yt-dlp; open contract; expected source declaration not established	○	◉	◉	●	○	○	○
chaturbate room status and edge-URL resolve -- bulk-first polling, single-use edge token honoured Measured: `cb_decide` exists in `runtime/nx_cam_chaturbate.nx`, verified at emit. The SITE half of R9, landed 2026-09-01 and gate-proven 33 of 33 (nx_cam_chaturbate_gate, planted .invalid fixtures, no network). Mechanism MEASURED not recalled: status and URL come from a BULK affiliates onlinerooms feed plus a POST to get_edge_hls_url_ajax, never from scraping the room page. The retired Python detector in the operator's ctbrec logs reported PRIVATE and Streams Found 0 across 13 patterns and recommended Selenium; the HTML that same run saved shows it was bounced to the LANDING page (a next equals slash-question-next parameter plus roomlist markers, and the string it claims to have matched occurs zero times as a room status), so all 13 misses had ONE cause and it was not JavaScript. A second witness for the same age-gate already sat in nx_http_client's own cookie-builder comment. Three states the incumbent paid for in production are teeth here: a PUBLIC room with an EMPTY url is RESTRICTED not PUBLIC (starting a capture on an empty input is what produced its retry loop), a 403 challenge is RATELIMIT not a fast-retryable error (fast retries lengthen a Cloudflare block), and an unrecognised status lands in UNKNOWN never in a known bucket. Bulk-first is the rate-limit answer and not an optimisation: N watched performers cost ONE request, and paying a per-room ajax across a large fleet is what flags an exit IP into a site-wide 403. The resolved URL carries a SINGLE-USE connection-bound edge token, so cb_capture_contract returns the one-consumer handoff rule as a VALUE a gate can assert from outside rather than as a comment no caller can see. Exceed is 0 on purpose: the behaviours were learned from Harvestr's bot, and what is ours is that they are gate-proven, that the two endpoints share ONE status vocabulary, and that we answer exit-IP pressure with the bulk feed and the R0 pacer instead of the residential proxy pool the incumbent uses. Harvestr and ctbrec cells are read from the operator's installed instances (streamonitor sites chaturbate.py, and rooms.json auto_record) which is the strongest evidence on this board; the yt-dlp cell is a capability read not re-verified against a mirror this leg [harvestr]. Adoption: `LIB-GATE-ONLY importers=2` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of Video DownloadHelper, Stash, Hydrus NetworkCons behind Harvestr (leads), ctbrec (leads); not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	◉	◉	●	○	○	○
sovereign BitTorrent acquisition -- magnet to verified pieces, DHT, PEX, MSE, streaming picker, autoseed Source present; authored stronger claim: `tg_autoseed` in `runtime/nx_torrent_get.nx`, verified at emit. EXCEED with the why: none of the six ingests torrents at all; ours is a full client in the sovereign language (rarest-first, endgame, ip-filter, pool width derived from the box) and the gallery pipeline auto-fires on each drop. Adoption: `PROMOTED-UNREGISTERED` — PARTIAL: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: a real binary nobody can call over MCP: /api/tools/register it; 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 Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus NetworkCons not adopted yet: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	◉	○	○	○	○	○	○
cross-temporal recovery -- a missing image recovered from a DIFFERENT Wayback capture than the page that referenced it Source present; authored stronger claim: `fetch_retry` in `runtime/nx_media_xrecover.nx`, verified at emit. EXCEED with the why: CDX-query every capture of the asset URL, pick a status-200 capture, fetch its raw bytes and validate real image magic; no rival recovers dead media from an archive. Adoption: `SOURCE-ONLY` — PARTIAL: source exists, never compiled: /api/build it. not adopted: source exists, never compiled: /api/build it Pros Source symbol found; stronger capability claim is authored; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus NetworkCons not adopted yet: source exists, never compiled: /api/build it	◉	○	○	○	○	○	○
cookie and session auth -- cookie jar with age-gate acceptance carried across redirects Measured: `ff_collect_cookies` exists in `runtime/nx_https_fetch_follow.nx`, verified at emit. yt-dlp --cookies-from-browser and Netscape cookies.txt [ytdlp]; Harvestr per-site cookies.txt plus X auth cookies [harvestr]; Hydrus session cookies per domain [hydrus-dl]; ours collects Set-Cookie within one fetch chain and imports nothing. Adoption: `LIB-WIRED importers=241 nonval=217` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind Harvestr (leads), yt-dlp (leads), Video DownloadHelper (leads), Hydrus Network (leads)	●	◉	●	◉	◉	●	◉
import cookies from the operator's browser or a cookies.txt Open — watching `runtime/nx_cookie_import.nx` : `ci_netscape`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. The login-walled half of the corpus (premium X, fan sites, age-gated tubes) is unreachable without it [ytdlp] [harvestr]. dossierwatching ci_netscape Pros none measured yetCons behind Harvestr (leads), ctbrec, yt-dlp (leads), Video DownloadHelper (leads), Hydrus Network (leads); open contract; expected source declaration not established	○	◉	●	◉	◉	○	◉
CATALOGUE
byte-sniffed type classification -- four trust tiers, 70 detectors incl TS, STL, EPUB, WARC, never extension-trusted Measured: `nx_sniff_full` exists in `runtime/nx_media_sniff.nx`, verified at emit. Hydrus identifies files by bytes too [hydrus-dl]; Harvestr trusts the downloader's extension [harvestr]. MEASURED 2026-07-30: the pre-rewrite sniffer knew six magics and filed 53,567 cam recordings as unknown -- the current one is the root fix. Adoption: `LIB-WIRED importers=9 nonval=8` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec; fully adopted on the estate ladderCons behind Hydrus Network (leads)	●	○	○	●	●	●	◉
content identity -- streaming CID over the whole file at O(1) memory, multi-GB safe Measured: `cid_of_file_chunk` exists in `runtime/nx_canon_cid.nx`, verified at emit. Hydrus dedups on the full SHA-256 [hydrus-dupes]; Harvestr's dedupe is size plus head and tail SHA1, a 99 percent heuristic by its own README [harvestr]; yt-dlp --download-archive dedups by site id, not bytes [ytdlp]. Adoption: `LIB-WIRED importers=116 nonval=86` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper; fully adopted on the estate ladderCons behind Hydrus Network (leads)	●	◐	○	◐	○	●	◉
perceptual near-duplicate for images -- dHash and aHash, Hamming nearest Measured: `nx_phash_dhash` exists in `runtime/nx_phash.nx`, verified at emit. Hydrus similar-file search and duplicate filter run on perceptual hashes [hydrus-dupes]; Stash generates a phash per scene; ours indexes (nx_phash_index) and the vault does not consult it at ingest. Adoption: `LIB-WIRED importers=21 nonval=13` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper; fully adopted on the estate ladderCons behind Stash (leads), Hydrus Network (leads)	●	○	○	○	○	◉	◉
video perceptual hash -- frame-sampled fingerprint so re-encoded mirrors collapse Open — watching `runtime/nx_vphash.nx` : `vph_scene`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Harvestr's cross-mirror dedup is byte-level only [harvestr]; ours has the image hasher and the first-IDR frame decoder (nx_cam_poster) and no bridge between them. dossierwatching vph_scene Pros none measured yetCons behind Stash (leads), Hydrus Network; open contract; expected source declaration not established	○	○	○	○	○	◉	●
classify-at-the-door -- class (gen or real) x type facets, source badge, provenance stamped at ingest Measured: `mv_type_bytes` exists in `runtime/nx_mvault_ingest.nx`, verified at emit. Stash and Hydrus catalogue with tags, performers, studios and import URLs [stash] [hydrus-dl]; Harvestr keeps history.json per performer folder [harvestr]; ours keys TYPE on the bytes, never on the source. Adoption: `LIB-WIRED importers=12 nonval=7` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Video DownloadHelper; fully adopted on the estate ladderCons behind Stash (leads), Hydrus Network (leads)	●	●	●	●	○	◉	◉
one capture chokepoint -- media, page, or whole site into the vault as CID items with collections declared FIRST Source present; authored stronger claim: `vc_ingest_file` in `runtime/nx_vault_capture.nx`, verified at emit. EXCEED with the why: one verb, three scopes, and every collection member MUST resolve to a record (gate tooth with a negative control); VDH saves a page's media to a folder [vdhcoapp]; Hydrus gallery and watcher downloaders import files, not pages or sites [hydrus-dl]. 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 Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus NetworkCons 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)	◉	○	○	○	●	○	●
bulk reindex of an existing corpus -- recursive walk, streaming CID, path-resume plane, batched commits, record-only Measured: `mw_run` exists in `runtime/nx_mvault_walk.nx`, verified at emit. Stash scan and Hydrus import folders are the bar [stash] [hydrus-dl]. MEASURED 2026-08-30: the vault holds 160,415 CID records with the bytes left in place. 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 Harvestr, ctbrec, yt-dlp, Video DownloadHelperCons behind Stash (leads), Hydrus Network (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)	●	○	○	○	○	◉	◉
collections and provenance captured AT ingest -- which album, which site, which page Measured: `mvc_batch_add_pfx` exists in `runtime/nx_mvault_coll.nx`, verified at emit. The 07-31 lesson in source: ingesting first and grouping later cost a 35,615-record backfill; provenance at ingest is free. 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=5` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Video DownloadHelper; fully adopted on the estate ladderCons behind Stash (leads), Hydrus Network (leads)	●	●	●	●	○	◉	◉
image thumbnails -- sovereign decoder plus integer box downscale, CID-addressed cache, fork-per-image memory bound Measured: `th_one` exists in `runtime/nx_galx_thumb.nx`, verified at emit. yt-dlp writes the SITE's thumbnail, it derives none [ytdlp]; Stash and Hydrus derive on scan [stash] [hydrus-dl]; ctbrec keeps a thumbnails dir (observed on the installed instance). 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 Harvestr, Video DownloadHelperCons behind Stash (leads), Hydrus Network (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)	●	○	●	●	○	◉	◉
thumbnail decoder coverage -- JPEG, WebP, GIF, palette and 16-bit PNG Measured: `th_decode_any` exists in `runtime/nx_galx_thumb.nx`, verified at emit. PARTIALLY MET 2026-08-31 (R3 -- the contract SYMBOL is present and the decode composition is real, but this rung is NOT closed against its own pre-declared done-rule, which requires a fixture set of ONE FILE PER FORMAT thumbnailing end to end AND the galx_pipeline.log decode-error class reading ZERO on a re-run over the recordings the 07-22 census flagged; neither was done, and the gate exercised TWO formats, PNG and P6 PPM, not one per format): th_decode_any composes nx_img_bytes_to_rgb -- THE single gate-proven unified decoder (JPEG baseline+progressive, GIF, WebP-lossless, BMP, TIFF, TGA, PCX, ICO, PNM, and PNG all colour types incl palette + sub-byte + 16-bit) -- so nx_galx_thumb is no longer PNG-only. The organ was BUILT-UNPROMOTED and was promoted 2026-08-31 to the NISHIHOST SERVING ROOT (286431 B), which NO thumbnail consumer reads, so this decoder capability is NOT yet live on the gallery path. Measured the same day by exhaustive scan (corpus_complete=1, 23407 files): all three live consumers -- nx_gallery_serve GS_THUMB_ELF via gs_thumb_run, nx_media_server MS_THUMB_ELF, and nx_media_ingest_pipeline mp_resolve -- fork /volume1/ai/galx/nx_galx_thumb.elf, a different 100856 B binary the promote never touched, and the two CLI contracts are incompatible (the promoted build dispatches on a subcommand while gs_thumb_run passes argv[1] as a source path), so copying the promoted artifact across would break every cold-cache thumbnail rather than fix it. AND THE REFEREE POINTS AT THE OTHER ROOT: nx_galx_thumb_gate declares its subject as the BARE path nx_galx_thumb.elf, which resolves against the serving root, so its GREEN 4/4 is evidence about the newly promoted 286431 B artifact and says NOTHING about the 100856 B binary the consumers fork -- the gate and the consumers measure two different binaries, which is exactly why the referee can read GREEN while gallery thumbnails stay broken. The remedy is to build the ONESHOT target (buildroot/runtime/_hdl_build/nx_galx_thumb_oneshot.nx) and install THAT to the galx root; filed as a media-vault debt row superseding 1784922045, which had already recorded this same split on 2026-07-24. Live-proven and gate nx_galx_thumb_gate 4/4 GREEN: a real PNG thumbnails with the path preserved, a P6 PPM thumbnails end to end (a non-PNG format, the R3 capability), and garbage is refused with no wrong image; per-format decode is proven inside nx_img_bytes_to_rgb which this composes and never reimplements -- one decode ruler, no duplicate. yt-dlp writes the site thumbnail and derives none [ytdlp]; Stash and Hydrus derive on scan [stash] [hydrus-dl], both via ffmpeg or libvips while this is sovereign from the compiler up. 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 Harvestr, yt-dlp, Video DownloadHelperCons behind Stash (leads), Hydrus Network (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)	●	○	●	○	○	◉	◉
video poster from the first IDR -- TS and MP4 containers, H.264 CAVLC, CABAC and HEVC named not silent Measured: `build_poster_rgb` exists in `runtime/_hdl_build/nx_cam_poster.nx`, verified at emit. No ffmpeg; the batch mode is idempotent (skip-existing). rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `PROMOTED-UNREGISTERED` — PARTIAL: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: a real binary nobody can call over MCP: /api/tools/register it; 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 Harvestr, Video DownloadHelperCons behind Stash (leads), Hydrus Network (leads); not adopted yet: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	●	○	●	●	○	◉	◉
trickplay scrub-sprite sheet derived at ingest Open — watching `runtime/nx_galx_trickplay.nx` : `gt_sprite_sheet`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Stash generates sprites at generate time; shared contract with media R6 -- the ingest half of the same organ. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them dossierwatching gt_sprite_sheet Pros none measured yetCons behind Stash (leads); open contract; expected source declaration not established	○	○	○	○	○	◉	○
duration and keyframe index at ingest (NXVI) -- H.264 and HEVC NAL-aware, exact seek Measured: `dump_idx` exists in `runtime/nx_ts_index.nx`, verified at emit. MEASURED 2026-07-22: 1,260 indexes for 76,004 catalogued recordings = 1.7 percent, built lazily on first play; the backfill organ (nx_galx_idx_backfill) is BUILT-UNPROMOTED on nishihost while the galx root carries a copy. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `RUN-BY:fork:nx_hostctl` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec, Video DownloadHelper; fully adopted on the estate ladderCons behind Stash (leads)	●	○	○	●	○	◉	●
scene, dead-air, motion and rhythm markers at ingest Source present; authored stronger claim: `ts_classify` in `runtime/nx_ts_marks.nx`, verified at emit. EXCEED with the why: demux-signal analysis emits NXMK markers no ingest tool derives; Stash markers are hand-placed. 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=5 nonval=3` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus Network; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	●	○
analysis-on-ingest pipeline -- thumbs plus NXVI plus posters in one idempotent resumable pass, every missing tool ANNOUNCED Measured: `mp_resolve` exists in `runtime/nx_media_ingest_pipeline.nx`, verified at emit. Stash generate and Hydrus import-time derivation are the bar [stash] [hydrus-dl]. nx_catalog reads this BUILT-UNPROMOTED on nishihost while /volume1/ai/galx carries nx_media_ingest_pipeline.elf (19 elfs there, measured 2026-08-30) -- two artifact roots, name which. 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 Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelperCons behind Stash (leads), Hydrus Network (leads); not adopted yet: compiled, never promoted to the serving root: /api/promote it	●	○	○	○	○	◉	◉
media fact extraction with a structural PII wall -- closed integer vocabulary, EXIF surfaced as a presence flag only Source present; authored stronger claim: `mf_probe` in `runtime/nx_mediafacts_lib.nx`, verified at emit. EXCEED with the why: no byte of the input flows to the output, so GPS, serials and embedded thumbnails are unreadable through this lane; yt-dlp --embed-metadata and ffprobe copy metadata through [ytdlp]. Adoption: `LIB-WIRED importers=5 nonval=4` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus Network; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	●	●	●	●
container probe -- ISO-BMFF box tree walked and malformed sizes flagged Measured: `walk` exists in `runtime/nx_mp4_probe.nx`, verified at emit. ffprobe-class depth (codec, fps, bitrate, tracks) is theirs [stash]; ours validates structure and stops at mdat. Adoption: `PROMOTED-UNREGISTERED` — PARTIAL: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked) Pros Implementing source symbol found; runtime acceptance separateCons behind yt-dlp (leads), Video DownloadHelper (leads), Stash (leads), Hydrus Network (leads); not adopted yet: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	●	●	●	◉	◉	◉	◉
usability verification -- every ingested page or frame actually DECODES, serving is not working Source present; authored stronger claim: `mu_check_book` in `runtime/_hdl_build/nx_media_usable.nx`, verified at emit. EXCEED with the why: OK, DEGRADED or BROKEN per book with the reason per page; Harvestr repairs videos after the fact (video_repair.py) [harvestr]; no cataloguer proves decodability before calling a thing ingested. Adoption: `SOURCE-ONLY` — PARTIAL: source exists, never compiled: /api/build it. not adopted: source exists, never compiled: /api/build it Pros Source symbol found; stronger capability claim is authored; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus NetworkCons not adopted yet: source exists, never compiled: /api/build it	◉	●	○	○	○	○	○
live-vs-recording classifier over frame fingerprints -- slates, loops and ads flagged Source present; authored stronger claim: `nx_rec_classify` in `runtime/nx_recording.nx`, verified at emit. EXCEED with the why: a sustained hold or a periodic recurrence of perceptual hashes marks a frozen or looped segment for trimming; no recorder or cataloguer classifies this. Adoption: `LIB-GATE-ONLY importers=1` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Source symbol found; stronger capability claim is authored; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus NetworkCons not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	◉	○	○	○	○	○	○
takedown and hash-match exclusion at ingest -- dHash blocklist with per-entry Hamming threshold Source present; authored stronger claim: `excl_blocked` in `runtime/nx_media_exclusion.nx`, verified at emit. EXCEED with the why: content removed for being illegal or non-consensual stays removed including its re-encodes; the blocklist is data and starts empty. Adoption: `LIB-WIRED importers=3 nonval=1` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus Network; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	○
ingest audit ladder per artifact -- lane detected from bytes, grades as data, unknown says so Source present; authored stronger claim: `ia_ladder_from` in `runtime/nx_ingest_audit.nx`, verified at emit. Shared with koikatsu: a capability report that only names what it looked for cannot tell you what you have. 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 Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus NetworkCons 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)	◉	○	○	○	○	○	○
media inventory census -- recordings, movies, books, id-stable stale set, NSFW fail-closed Measured: `iv_build` exists in `runtime/_hdl_build/nx_media_inventory_lib.nx`, verified at emit. MEASURED (nx_media_inventory, generated 2026-07-17): 53,567 recordings, 3,357 performers, 2,391 movie files, 2,822 books; counts are whatever the getdents walk measures. 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=2 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of ctbrec, yt-dlp, Video DownloadHelper; fully adopted on the estate ladderCons behind Stash (leads), Hydrus Network (leads)	●	●	○	○	○	◉	◉
consent-gated catalogue durability -- rights ledger snapshot through the append-only seg-store Measured: `nx_dms_catalog_save` exists in `runtime/nx_dms_catalog_store.nx`, verified at emit. Consent travels with the record, never a side table. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-GATE-ONLY importers=1` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelperCons not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	○	○	○	○	●	●
disk budget guard at ingest -- refuse below a named floor, the refusal announces Measured: `ig_disk_floor` exists in `runtime/nx_ingest_admit.nx`, verified at emit. LANDED (R2): ig_disk_floor (nx_ingest_admit) reads the disk axis via sys_fs_avail_bytes and returns IG_ADMIT, IG_REFUSE_FULL or IG_UNMEASURABLE against a named floor from knowledge/status/ingest_floor.conf (default 5 GiB); nx_vault_capture calls it at the top of vc_ingest_file and on a full disk announces the shortfall, unlinks the tmp and returns VC_DISK_REFUSE before writing a partial. Gate nx_ingest_admit_gate 8/8 GREEN -- pure ig_decide proofs across all three states, a full-string verdict neg-control, and a live wrapper confirming a valid verdict against the real vault root. Harvestr pairs min_disk_gb with a disk manager [harvestr]; the disk axis landed in nx_syscalls on 2026-08-27 and the capture path now reads it. Adoption: `LIB-WIRED importers=2 nonval=1` — fully adopted (top of its ladder). dossier Pros Implementing source symbol found; runtime acceptance separate; ahead of yt-dlp, Video DownloadHelper, Stash; fully adopted on the estate ladderCons behind Harvestr (leads)	●	◉	●	○	○	○	●
TRAIN
per-item capture store -- measurements, tags and captions keyed under the CID with a field index Measured: `mv_cap_put` exists in `runtime/nx_mvault_caption.nx`, verified at emit. Stash tags and Hydrus tags plus the PTR are the bar [stash] [hydrus-dl]; ours is the write-API and the tagger below is absent. Adoption: `LIB-GATE-ONLY importers=1` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelperCons behind Stash (leads), Hydrus Network (leads); not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	○	○	○	○	◉	◉
auto-tagger and captioner producing that capture (VLM plus the integer judges) Open — watching `runtime/nx_caption_tagger.nx` : `ct_tag_item`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Stash scrapers fill metadata from sites with no vision [stash]; Hydrus tags arrive from the PTR [hydrus-dl]; nobody captions for training; ours needs the producer. dossierwatching ct_tag_item Pros none measured yetCons behind Stash, Hydrus Network; open contract; expected source declaration not established	○	○	○	○	○	●	●
embedding store -- face, CLIP-class and pose vectors keyed CID plus model, additive Measured: `mv_embed_put` exists in `runtime/nx_mvault_embed.nx`, verified at emit. The store exists; the producer is the media R10 contract below. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-GATE-ONLY importers=1` — PARTIAL: imported only by validation organs (gates, tests, benches): wire it into a shipping program. not adopted: imported only by validation organs (gates, tests, benches): wire it into a shipping program Pros Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Hydrus NetworkCons not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	○	○	○	○	●	○
sovereign encoder producing embeddings and semantic search over the library Open — watching `runtime/nx_media_semantic.nx` : `msm_search`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Shared contract with media R10; recall at 10 is PRE-DECLARED to beat BM25 on a labeled set or the lane stays unwired. dossierwatching msm_search Pros none measured yetCons open contract; expected source declaration not established	○	○	○	○	○	○	○
training manifest export -- favourites to image TAB caption rows from the PNG prompt text Measured: `le_prompt` exists in `runtime/nx_galx_lora_export.nx`, verified at emit. Hydrus exports sidecar txt and json tags per file [hydrus-dl]; LIVE (invoked) on the gallery favourites. Adoption: `RUN-BY:clock` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	○	○	●
caption format emitters -- joycaption, prompt text, TTS, raw JSON Measured: `nx_caption_emit_joycaption` exists in `runtime/nx_caption_emit.nx`, verified at emit. Dataset consumers read the same beats in the shape they want. 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=2 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	○	○	●
dataset export with provenance and licence per item -- image and video-clip shards, a manifest that reconciles Open — watching `runtime/nx_dataset_export.nx` : `ds_export_shards`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. The operator's third goal: ingest powers training. Contract: every shard row carries CID, source, licence tier and consent flag; a row without provenance is REFUSED, and the manifest partition must sum. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them dossierwatching ds_export_shards Pros none measured yetCons behind Hydrus Network; open contract; expected source declaration not established	○	○	○	○	○	○	●
acquired-model integrity screen before a LoRA or checkpoint may be used Source present; authored stronger claim: `lc_screen` in `runtime/_hdl_build/nx_lora_screen_lib.nx`, verified at emit. EXCEED with the why: safetensors header parsed, tensor offsets scanned, metadata excerpted, an oversize header FLAGGED not truncated; no ingest tool screens models. Adoption: `LIB-WIRED importers=2 nonval=1` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus Network; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○	○	○
SURFACE
browser-native -- the m key deep-harvests the page and the grab verb files it into an isolated grabs gallery Measured: `gr_emit_gallery` exists in `runtime/nx_media_grab.nx`, verified at emit. VDH is the in-browser bar [vdhcoapp]; ours keeps web grabs SEPARATE from the training-favourite corpus by construction. 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 Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec, yt-dlp, Stash, Hydrus NetworkCons behind Video DownloadHelper (leads); not adopted yet: compiled, never promoted to the serving root: /api/promote it	●	○	○	○	◉	○	○
sniff endpoint any client can call -- GET /sniff?url= returns the best route Measured: `send_resp` exists in `runtime/_hdl_build/nx_media_sniff_daemon.nx`, verified at emit. Loopback-bound by default, fronted by the opaque gateway to go public; the CoApp role served instead of installed [vdhcoapp]. Adoption: `SOURCE-ONLY` — PARTIAL: source exists, never compiled: /api/build it. not adopted: source exists, never compiled: /api/build it Pros Implementing source symbol found; runtime acceptance separate; ahead of Harvestr, ctbrec, yt-dlp, StashCons behind Video DownloadHelper (leads); not adopted yet: source exists, never compiled: /api/build it	●	○	○	○	◉	○	●
third-party browser extension EMITTED from the estate -- Waterfox and Firefox (MV2) plus Chrome and Edge (MV3), one-click push to the sovereign stack Measured: `ge_emit` exists in `runtime/nx_grab_ext_emit.nx`, verified at emit. MEASURED 2026-08-30: a hand-written Gecko MV2 extension exists LAPTOP-ONLY (nxc2/sites/nishi-grab, 2026-07-23: webRequest media sniffer, badge, push to the torrent daemon over the authenticated edge) -- never published into the estate, pushes to torrent not to vault capture, no MV3 twin. VDH v10 needs no companion at all [vdhcoapp]; a native host would have to speak the per-browser native-messaging contract [mdn-nativemsg]. Contract: the extension is emitted from data by an organ, pushes to nx_vault_capture, and is served from nishifamily. PROGRESS 2026-09-01, and the cell stays _ABSENT_ ON PURPOSE: the EMIT half is built and proven -- nx_grab_ext_emit (26,600 B) emits BOTH twins from knowledge/grab_ext.conf, doing the MV2-to-MV3 structural split itself (background.scripts to service_worker, browser_action to action, and all-urls moved out of permissions into host_permissions) so the twins cannot diverge in policy, with nx_grab_ext_emit_gate GREEN 32 of 32 (PROMOTED + REGISTERED, run via nx_job_run) whose load-bearing teeth check the permissions array CONTENTS in BOTH directions, because a tooth that only asserted host_permissions exists would pass an emitter that wrote all-urls into both arrays, which is the actual MV3 packaging error. The symbol ge_emit now exists as the real entry point. THE RUNG IS STILL NOT DONE, and the reason is a dependency this row and the plan never named: R10 says the extension pushes to nx_vault_capture over the authenticated edge, and THAT ENDPOINT DOES NOT EXIST. Measured 2026-09-01 over 23,421 organ sources, coverage_complete=1 corpus_complete=1: the string /api/vault appears in NO organ, while /torrent/api/add and /torrent/api/video DO exist as real handlers in nx_torrent_daemon.nx. nx_vault_capture is a CLI organ with no HTTP surface at all (entry points vc_capture_media, vc_capture_page, vc_capture_site), so a serving surface must authenticate the session and invoke those. Retargeting the extension at the torrent route would reproduce the exact defect this row exists to remove while reading green, so it was NOT done. That endpoint fetches arbitrary user-supplied URLs and is therefore SSRF-capable, which by this estate's own measured incident (an unauthenticated fetcher bound to INADDR_ANY answering from another host) makes standing up a PUBLIC route an operator decision rather than a defect fix. DECLARED LIMIT of the emitted artifact: the background body keeps its per-tab media list in memory, which is correct for the MV2 persistent page and lossy on MV3 where the service worker is torn down when idle -- the capture push is unaffected, the fix is storage.session mirroring plus rehydration on wake, and it is deliberately deferred to its own increment with its own teeth rather than smuggled in beside the structural split. Adoption: `REGISTERED-UNAUTHORISED` — PARTIAL: registered, no cap ever minted; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). dossiernot adopted: registered, no cap ever minted; 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 Harvestr, ctbrec, yt-dlp, StashCons behind Video DownloadHelper (leads); not adopted yet: registered, no cap ever minted; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	●	○	○	○	◉	○	●
Windows and third-party browsers served WITHOUT a native companion -- the extension talks to the estate over the authenticated edge DARK — `runtime/nx_grab_ext_emit.nx` EXISTS but does not declare `ge_edge_push`: 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. Every rival installs natively on Windows (Python, Java, exe, CoApp) [harvestr] [vdhcoapp]; the sovereign toolchain emits x86-64 ELF and runs under WSL, so the honest Windows answer today is the edge model -- the NAS does the work, Windows needs only the browser -- and VDH itself moved off its companion [vdhcoapp]. dossierwatching ge_edge_push Pros none measured yetCons behind Harvestr (leads), ctbrec (leads), yt-dlp (leads), Video DownloadHelper (leads), Stash (leads), Hydrus Network (leads); open contract; expected source declaration not established	○	◉	◉	◉	◉	◉	◉
operator dashboard -- queue, live fleet, per-site switches, history, failed-with-reason, dry-run Open — watching `runtime/nx_ingest_board.nx` : `ib_page`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. Harvestr's Flask dashboard and ctbrec's web UI [harvestr]; ours has /gallery and the vault gateway and no ingest board. dossierwatching ib_page Pros none measured yetCons behind Harvestr (leads), ctbrec (leads), Video DownloadHelper, Stash (leads), Hydrus Network (leads); open contract; expected source declaration not established	○	◉	◉	○	●	◉	◉
resumable and incremental -- a re-run skips what is already held Measured: `mw_pathkey` exists in `runtime/nx_mvault_walk.nx`, verified at emit. Path-resume plane keyed on path plus size (declared: a same-size rewrite is missed); yt-dlp --download-archive [ytdlp]; Harvestr history.json with a rolling window per site [harvestr]. 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 Video DownloadHelperCons behind Harvestr (leads), yt-dlp (leads), Stash (leads), Hydrus Network (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)	●	◉	●	◉	○	◉	◉
whole ingest stack in one sovereign language -- own TLS, HLS, AES, decoders, torrent; no ffmpeg, yt-dlp, aria2c or Chromium Source present; authored stronger claim: `hg_remux` in `runtime/nx_hls_get.nx`, verified at emit. EXCEED with the why: Harvestr is yt-dlp plus aria2c plus ffmpeg plus curl_cffi plus a Playwright fork [harvestr]; Stash requires ffmpeg [stash]; VDH shipped ffmpeg inside its CoApp [vdhcoapp]; one auditable stack from the compiler up. Adoption: `PROMOTED-UNREGISTERED` — PARTIAL: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: a real binary nobody can call over MCP: /api/tools/register it; 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 Harvestr, ctbrec, yt-dlp, Video DownloadHelper, Stash, Hydrus NetworkCons not adopted yet: a real binary nobody can call over MCP: /api/tools/register it; no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)	◉	○	○	○	○	○	○

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	58	rows carrying a Yes, Best or Part in a rival column
cited	44	a reference mark resolves to a pinned mirror
uncited	14	an observation read, badged until a reference lands
rows, summed	58	RECONCILED: the parts sum to the total 58

Delivery and evidence

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

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
2026-09-02	R20	queue	nothing built by operator order (last mile). The consumer rung is aesthetictwin AT20; the fit machinery is aesthetictwin AT18; the tag vocabulary is a data file to be admitted with the first flag, never typed in an organ.
2026-09-03	R21	land	conf + lib + gate landed on the NAS 2026-09-03 (asset_provenance.conf 3,244 B, 12 classes; nx_provclass_lib 9,251 B sha 7a43dacd; nx_provclass_gate 7,139 B sha b8b09288). 15/16 on the laptop -- the ONE failure is the anti-vacuity tooth journal-population-was-actually-read, because the journal is NAS-side; it caught every-live-journal-licence-tag-classifies scoring a VACUOUS actual=0 expected=0 PASS on first run, on its own author. NAS run owed once build admission clears.
2026-09-03	R21	measure	the live refcorpus journal carries 47 rows and FIVE undeclared licence vocabularies (ORIGINAL, sovereign, royalty-free, CC-BY-mixed, REFERENCE_MEASURED). All five are declared in the conf, so the new wall refuses no existing row -- measured directly, not asserted, and the gate will assert it on the NAS.
2026-09-03	R21	lesson	OPERATOR DECISION 2026-09-03: SFMLab / Open3DLab / SmutBase content is admitted as class `unverified` -- reference-only, k-floor 32. One platform, three front doors, user-uploaded with reactive-DMCA enforcement, so a licence is not restrictive but ABSENT. At n>=32 a distribution fact is a fact about a population and carries no copyright; at n=8 it can still describe one work. The sites sit behind Anubis proof-of-work and NO solver was built: webscraping R4 already declared that boundary before this seat arrived.
2026-09-10	R19	queue	2026-09-10 operator request retained, not implemented or visually verified: local/remote/rented workers return source provenance, inference uncertainty, quality/test receipts and reviewable selection decisions; NAS coordinates and retains durable results rather than assuming heavy compute placement. Link containers CT1 and gameengine GE61.
2026-10-08	R22	measure	2026-10-08 thread /s/22460514: nx_4chan thread listed the media, nx_4chan fetch downloaded 151 of 151 (143 jpg, 7 png, 1 webm, 0 failed) into media/4chan/s_22460514, and nx_mvault_walk real 4chan commit recorded 151 unique items (dup 0, commits 1, fatal 0) into the mvault record store (knowledge/store/mvault-, NXR1 records, the /vault gateway's surface). The gallery grid's view index did not change (226,595 records before and after): vault records are a separate surface from the gallery's images and recordings, so a one-click download lands in /vault, not in the /gallery grid -- a surface decision to make explicit on this board, not a defect.
2026-10-08	R22	land	the gallery backend deployed 2026-10-08 (galxserve, sha a9d8d1d7, first health-checked deploy of this daemon through /api/deploy -- its deploy_domain row galxserve mediaingest was missing and the GLOBAL fallback blocked on eight unrelated RED domains until the row was added) carries the thread view's download-all-to-library link (/api/4chan/panel); the fetch and the commit were driven as two sanctioned jobs because no loopback client exists for the authenticated POST /api/4chan/ingest -- the shape the adapter ABC must give every source.

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.

Honest verdict. The sovereign stack already owns the two halves the incumbents lease from ffmpeg, yt-dlp and Chromium: acquisition (own TLS with a Chrome-shaped hello, HLS with AES-128, streaming-to-disk, a full BitTorrent client, Wayback cross-temporal recovery, JS execution with zero network) and cataloguing (byte-sniffed types, streaming CID, classify-at-the-door vault records, thumbnails, keyframe index, usability verification, takedown exclusion). It is NOT best-of-breed for four MEASURED reasons, ranked by what the operator hits first: (1) RATE LIMITS -- the polite pacer (nx_crawl_pace: Retry-After, backoff, decay, persisted) is imported by the crawl lane and one board adapter and by NONE of the capture verbs (nx_vault_capture, nx_mvault_fetch, nx_hls_get, nx_media_grab; corpus_complete=1), so every capture hits a host as fast as its loop runs, and there is no exit rotation and no source pre-flight, while the bot-wall classifier that exists (nx_antibot, landed for the crawler) is consulted only by nx_web_crawl_step -- exactly the three things Harvestr built after it was rate-limited. (2) REACH -- Harvestr fans one username across 32 site patterns and 39 scrapers and yt-dlp reaches 1800+ sites; ours reaches 4chan, one data-row album adapter and generic page harvest, with no cookie import, no DASH, no segmented download and no live-record fleet, so the 53,567-recording cam corpus still comes from the retired Python recorder. (3) THUMBNAILS -- the pipeline exists but its image decoder is PNG-only and the keyframe index covered 1.7 percent of recordings when last measured, so the visible defect is decoder coverage and an unpromoted backfill, not a missing organ. (4) TRAINING -- the capture store, embedding store and manifest export exist while the tagger, the encoder and the provenance-carrying dataset export do not. What no rival has: the whole chain in one sovereign language, a CID chokepoint where every collection member must resolve to a record, a self-healing URL recogniser, a PII wall in fact extraction, and an ingest that proves the media decodes before it is called ingested.

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

Every compare carries this layer. Declare knowledge/compare/mediaingest.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 mediaingest, 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).

The field — discovered, not chosen

Rows written by nx_field_discover from mediaingest.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	5
scanned	5
fetched	5
reused	0
failed	0
named	5
candidates	5
mentions	5
capped	0

Rank	Rival	Seeds	Mentions	First seed	Kind	Link
1	Harvestr	1	1	col0-harvestr	named	https://raw.githubusercontent.com/KevinStreetCoder/harvestr/main/README.md
2	yt-dlp	1	1	col2-dlp	named	https://raw.githubusercontent.com/yt-dlp/yt-dlp/master/README.md
3	Video DownloadHelper	1	1	col3-video	named	https://raw.githubusercontent.com/aclap-dev/vdhcoapp/master/README.md
4	Stash	1	1	col4-stash	named	https://raw.githubusercontent.com/stashapp/stash/develop/README.md
5	Hydrus Network	1	1	col5-hydrus	named	https://hydrusnetwork.github.io/hydrus/getting_started_downloading.html

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 67 capabilities: 58 exclusive to it and 9 shared with other boards. Shared rows on this page: 9 = home here 0 + home elsewhere 1 + ambiguous 3 + undecided 5 + refused 0 (partition 9); undecided = unmapped 0 + unproven 5 + unevidenced 0; homonyms 0; since the previous census: new 0, resolved 0. Census FRESH age_s=7052.

Capability	Organ	Boards grading it	Home	Rule	Evidence	Standing on this board
nx_media_harvest	`nx_media_harvest.nx`	webscraping, mediaingest	—	none	unproven	undecided
pp_rotate	`nx_proxy_pool.nx`	webscraping, mediaingest	—	ambiguous	plan	ambiguous
ts_classify	`nx_ts_marks.nx`	media, sharehub, mediaingest	—	none	unproven	undecided
fa_is_real	`nx_fetch_any.nx`	media, mediaingest	—	none	unproven	undecided
gt_sprite_sheet	`nx_galx_trickplay.nx`	media, mediaingest	—	ambiguous	plan	ambiguous
msm_search	`nx_media_semantic.nx`	media, mediaingest	—	ambiguous	plan	ambiguous
le_prompt	`nx_galx_lora_export.nx`	mangagen, koikatsu, sharehub, mediaingest	—	none	unproven	undecided
ia_ladder_from	`nx_ingest_audit.nx`	koikatsu, mediaingest	—	none	unproven	undecided
lc_screen	`nx_lora_screen_lib.nx`	gen*, mediaingest	gen	rung	plan	home elsewhere adopt: `@xrow gen\|lc_screen\|home by rung (fleet duplicate census)`

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 1396157 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.

[harvestr] KevinStreetCoder. Harvestr README (main): one username in, every video across 50+ sites out -- yt-dlp plus 39 custom scraper classes, aria2c 16-connection downloads, HLS and DASH via ffmpeg, curl_cffi Chrome TLS fingerprint plus cloudscraper for Cloudflare, Accept text/css for DDoS-Guard, head-plus-tail SHA1 content dedupe, cookies.txt auth, Flask dashboard, Live mode via vendored StreaMonitor across 18 cam platforms polling every 5-30 s, tiered Mullvad exit rotation when a site rate-limits, per-site residential proxy, pre-flight site health. Fetched 2026-08-30. publisher · read in our library knowledge/fetched/cmp_mediaingest_harvestr.md · pin hea6c97ddc6a93357cea58886769e6bad7766a0252a6b917ae7993febc5ca7c67 · accessed 2026-08-30 · source-readGrounds: The Harvestr column on every row: it is the operator-installed incumbent on the D drive and the source of the rate-limit pain this domain exists to close; the rate-limit storm response, identity fan-out, live-record fleet, source health, cookie auth, segmented download and dashboard rows are coded from this README.
[ytdlp] yt-dlp project. README (master): 1800+ site extractors; --sleep-requests and --retry-sleep between requests; --limit-rate; --download-archive to skip already-fetched ids; --cookies-from-browser BROWSER[+KEYRING][:PROFILE]; --impersonate CLIENT via curl_cffi; -N / --concurrent-fragments for parallel DASH and HLS fragments; --live-from-start; thumbnail and metadata embedding. Fetched 2026-08-30. publisher · read in our library knowledge/fetched/cmp_mediaingest_ytdlp.md · pin h45ad19ed646fe0f037632bfefafb777ba7e4cae26477f4013832055c9381a4bd · accessed 2026-08-30 · source-readGrounds: The yt-dlp column: extractor breadth, pacing options, cookie import, impersonation, segmented fragment download, live capture and the download-archive form of resume are all read from this README.
[vdhcoapp] aclap-dev. DownloadHelper CoApp README: the multi-platform (Windows, Mac, Linux) native-messaging companion that gave the Video DownloadHelper add-on file writing, player launch and a bundled ffmpeg; header notice that the CoApp is no longer developed because VDH v10 needs no companion; registration installs a JSON host manifest into Mozilla, Google and Microsoft browser directories. Fetched 2026-08-30. publisher · read in our library knowledge/fetched/cmp_mediaingest_vdhcoapp.md · pin hf7ea005f48f1d8f0b90373aed7d89e893dd7f40047630841fb32f0d1ade4f1a6 · accessed 2026-08-30 · source-readGrounds: The Video DownloadHelper column and the browser-extension and no-companion rows: VDH is the in-browser capture bar the operator named, and its own move away from a native companion is the evidence that the edge-push model is the right Windows and third-party-browser answer.
[mdn-nativemsg] MDN Web Docs. WebExtensions Native messaging: an extension exchanges messages with a native application registered by an app manifest (name, path, type stdio, allowed_extensions), started via runtime.connectNative or sendNativeMessage; Windows setup and Chrome incompatibilities (allowed_origins) documented. Fetched 2026-08-30. publisher · read in our library knowledge/fetched/cmp_mediaingest_mdn_nativemsg.html · pin hcdc1079a17337ca63ab5207fc1296de7bbfcf8cffdd84a0e260daed12e315d66 · accessed 2026-08-30 · vendor-docGrounds: The emitted browser-extension row: the native-messaging contract is what a Windows companion would have to implement per browser; the row grades the edge-push alternative against it.
[stash] stashapp. Stash README (develop): self-hosted Go webapp that organizes and serves a SFW and NSFW collection, scans configured media directories, pulls performer, tag, studio metadata through community scrapers and stash-box, tags and finds media, requires ffmpeg. Fetched 2026-08-30. publisher · read in our library knowledge/fetched/cmp_mediaingest_stash.md · pin h9919670b3e5813dcd1175cc0fe3cf0fcf5092aa8a2e17446b22d2d7d32eeeb98 · accessed 2026-08-30 · source-readGrounds: The Stash column: the catalogue bar (scan, tags, scrapers as packages, ffmpeg dependency) on the classify, reindex, thumbnail, tagger and sovereign-stack rows; Stash is installed on the estate (docker/stash) so it is a live comparison, not a brochure.
[hydrus-dl] hydrus network. Getting started - downloading: gallery and watcher downloaders, subscriptions, per-domain bandwidth rules, parsers and URL classes as shareable data, session cookies and import options. Fetched 2026-08-30. publisher · read in our library knowledge/fetched/cmp_mediaingest_hydrus_dl.html · pin h6fb1f1dc73c77527081cc69753c7c44040c92c483eec4931a0675eb8a400dfa3 · accessed 2026-08-30 · vendor-docGrounds: The Hydrus column on the pacing, adapters-as-data, cookie, capture-scope and tag rows: per-domain bandwidth rules are the published bar the per-host pacing row is graded against.
[hydrus-dupes] hydrus network. Duplicates: similar-file search on perceptual hashes with a duplicate filter for deciding relationships; exact duplicates already collapse on the SHA-256 file hash. Fetched 2026-08-30. publisher · read in our library knowledge/fetched/cmp_mediaingest_hydrus_dupes.html · pin he782ab8856aed6c7b7ef84067e10b8dcd8ad9fa46284e38e70c2b8e44473d711 · accessed 2026-08-30 · vendor-docGrounds: The content-identity and perceptual near-duplicate rows: exact-hash plus perceptual-hash dedup is the bar; ours has the CID and the dHash and does not consult the dHash at ingest.
[gallerydl] mikf. gallery-dl README: command-line downloader for image galleries and collections from many image hosting sites, with yt-dlp integration for HLS and DASH, --cookies-from-browser and an archive database; development moved to Codeberg. Fetched 2026-08-30. publisher · read in our library knowledge/fetched/cmp_mediaingest_gallerydl.rst · pin h6a1d52690b815a6eab4a4bb297a8d19ce6d6353f6be8c6cdd6d729714c1bb4d7 · accessed 2026-08-30 · source-readGrounds: The board and gallery adapters row: gallery-dl is the image-gallery bar for reddit, tumblr, pixiv and boards that the retired Python media manager covered and only 4chan has been ported from.
[rfc6585] Nottingham, M., Fielding, R. Additional HTTP Status Codes. IETF RFC 6585, April 2012: 429 Too Many Requests, the response MAY include a Retry-After header indicating how long to wait. publisher · read in our library knowledge/fetched/cmp_mediaingest_rfc6585.html · pin h423d8b867e8f2d4b15ee37d3aaa370ff7ece430aae50690440e4934987bce348 · accessed 2026-08-30 · published-standardGrounds: The per-host pacing rows: 429 plus Retry-After is the signal a polite client must honour, which nx_crawl_pace does and the capture chokepoint does not yet see.
[rfc8216] Pantos, R., May, W. HTTP Live Streaming. IETF RFC 8216, August 2017: master and media playlists, EXT-X-KEY AES-128 segment encryption, EXT-X-MAP init segments. publisher · read in our library knowledge/fetched/cmp_mediaingest_rfc8216.html · pin h6839d54f21b1f95d9d1b0bf334790c3ef90bd829be832ce35b6db2be9dc93ac4 · accessed 2026-08-30 · published-standardGrounds: The HLS download row: the protocol nx_hls_get implements (master to variant, AES-128 decrypt) and the fMP4 init-segment clause it declares as its next rung; re-fetched 2026-08-30 for this domain's provenance journal and byte-identical to the media domain's mirror (same sha256).

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