Nishi Family › Compare › Third-Party Reliance

Nishi Compare · capabilities, evidence and delivery

Third-Party Reliance

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

Nishi sovereign twins for every outside class the assistant reaches for (measured per action, full population, estate beats excluded) vs Claude Code builtin tools and the GNU userland and CPython 3.13 and Node.js 24 -- Sept-2026 field

Overview

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

Where we are. MEASURED 2026-09-05 against the sovereign action journal, the catalogue and the session record. The record begins 2026-07-20 (the day the transcript harvest started; the laptop retains transcripts only from 2026-08-05 and no laptop backup holds earlier sessions, so the estate's May-to-July build of the OS, browser and toolchain is not in it). THREE MEASUREMENT DEFECTS WERE FIXED TODAY BEFORE ANY NUMBER WAS PUBLISHED: the gauge and the miner both stopped at a 4 MB read cap on a 54 MB journal, so every published figure sampled the oldest frames (the live /outsidegap read 592 permil, a July-head sample); the journal mixes the assistant's tool calls with the estate's own autonomous beat runs, which diluted the denominator; and the ingest lane holds 79k of 757k captured frames. nx_claude_harvest now reads the file whole and partitions verb=run beats out (44 of 44 teeth GREEN on the promoted binary, three of them new and one a non-vacuity tooth), nx_actlog now reads whole (15 of 15). THE NUMBER: 685,665 frames, 274,104 estate beats excluded, 411,561 assistant actions, 47,230 outside, claude_outside_permil=114 -- 88.6 percent sovereign on the assistant's own actions, from 651 in July. The residual by class from nx_actlog freq over the full journal: Bash 12,584, PowerShell 11,588, Edit 9,398, Read 4,348 (80 percent of the outside actions); the remaining classes sit below the frequency cap of 30 (SV2). The incumbents: nx_claude_harvest (the gauge and the beat), nx_actlog (the O(F) miner), nx_selfsuff (outside services and sources per domain), the minergap- plane MG00-MG13 (the residual catalogue with owners and routes), nx_sovereignty_audit and nx_dependency_audit (source-level contaminants). What was ABSENT until today: a sovereignty domain on the compare surface (109 domains, none of them this), any SOTA axis against the field stacks, a head-to-head receipt per class, and a gate that could prove the miner (its 44-tooth selftest lived only in a verb; nx_claude_harvest_gate now forks it and pins artifact identity).

Where we need to go. OPERATOR ASK (2026-09-05): mine ALL the work since the beginning with the estate's own miner for the functionality and capability gaps that required third-party help INCLUDING the assistant's own tools -- shell, Python, JavaScript, the LLM itself, whatever the Nishi OS, browser and toolchain should own -- to get a plan that drives third-party reliance to ZERO as a stretch goal; a clear compare-style measurement that stays up to date automatically; where we stand against Sept-2026 SOTA by research the estate pulls, not a half-baked grep replacing grep; every capability to EXCEED the assistant's; records reaching far older than July. RUBRIC, verbatim: head-to-head benchmark receipts, and if we cannot get head-to-head then we need to build the capabilities -- no more non-evidence non-tested fire-and-forget. Deliverable: this board, produced by the estate, ranked by the estate, journaled by the estate. OPERATOR CLARIFICATION 2026-09-07: ALL operations are sovereign capability investment opportunities: grep and search, read, write, build, test, site emission, publication, recovery and AI execution. Repeated outside work must map to an existing owning rung and canonical organ, with the missing capability and a head-to-head acceptance receipt; graphics is one workload within this estate-wide scope. MCP and APIs are themselves investment domains: measure their correctness, discoverability, protocol compatibility, latency, resource cost and recovery against dated best-in-class comparators, then frontier targets. Claimed ownership never substitutes for a witnessed usable capability.

The unit. 1 u = one measured session-leg (organ plus gate plus live flip), the standing calibration. Local evidence: the honest-gauge fix (edit, local proof, byte-exact landing, queued build, promote, 44 of 44 on the NAS) was one leg on 2026-09-05; the sibling nx_actlog fix was half a leg the same hour.

Where we are: 16 present rows, 12 open rungs. The gauge is honest and live, the residual is named per class with its owner, the field stacks are mirrored and pinned, and no class yet carries a head-to-head receipt against the assistant. Counts measured at emit below this line.

Cost to a gauge that produces itself on this page: 1 u. M0: SV1 the journal beat and SV2 the class breakdown. Version 0.1 is a number on this board that moves without a seat.

Cost to the first head-to-head receipts: 4 u more. M1: SV3 the shell interlock (shared with engineshift), SV4 search vs ripgrep, SV6 fetch vs curl, SV5 the LLM battery vs the assistant. 5 u cumulative = version 1.0: every class carries a receipt or a build rung, never a guess.

Cost to zero on the program path: 6 u more. M2: SV7 Python, SV8 JavaScript, SV9 ssh, SV10 the laptop runner, SV12 the editor's source of record. 11 u cumulative = version 2.0.

Cost to a whole record: 0.5 u more. M3: SV11 recovers the pre-July history from an off-box backup or declares the horizon permanent. 11.5 u cumulative = version 3.0.

Research bar. ripgrep, the search oracle is measured on BurntSushi. Theirs: a line-oriented search tool that recursively searches a directory for a regex pattern [ripgrep]. Ours: SV4 -- nx_shelltool grep must match its result set on a fixed corpus and the receipt carries both wall times.

Research bar. SWE-bench, the code-agent oracle is measured on SWE-bench leaderboards. Theirs: repository-level issue resolution is the field's grade for a code-writing model [swebench]. Ours: SV5 -- the sovereign dispatcher's local and external engines graded on the same tasks as the assistant.

Research bar. CPython 3.13, the scripting runtime to retire from the program path is measured on Python Software Foundation. Theirs: the experimental free-threaded build (PEP 703) and the experimental JIT [cpython313]. Ours: SV7 -- an oracle that may be installed as a comparator, never a dependency on the build or run path.

Research bar. Node.js 24 on V8 13.6, the page runtime to retire is measured on OpenJS Foundation. Theirs: the v24.0.0 release ships V8 13.6 [node24]. Ours: SV8 -- page logic lowers from NishiLang to wasm; hand JS is the residual.

Research bar. OpenSSH 10.0, the remote-shell field with post-quantum key exchange by default is measured on OpenSSH project. Theirs: mlkem768x25519-sha256 became the default key exchange in 10.0 [openssh]. Ours: SV9 -- the sovereign upload door replaces the byte moves; ssh stays the recovery lane only.

Research bar. Model Context Protocol 2025-06-18, the boundary this board keeps is measured on MCP specification. Theirs: the tool-call protocol the assistant reaches sovereign organs through [mcpspec]. Ours: SV0 -- the gauge counts a tools/call as sovereign because the organ behind it is; the protocol is the door, not a reliance.

JSON interoperability and investment: Benchmark owed; no performance ratio measured. JSON is a language-independent text interchange standard, not a JavaScript runtime dependency. Strength: portable structured values; limits: implementation-dependent size/depth/number precision and escaping overhead require declared tests. RFC 8259 sections 8-9 require interoperable UTF-8 exchange and permit parser limits; the standard does not impose our 65536-byte connector limit. Compare native formats on exact round-trip fidelity, bytes, parsing cost, latency and failure recovery through reversible JSON adapters before claiming benefit. Primary external oracle retrieved 2026-09-07, not estate-pinned: https://www.rfc-editor.org/rfc/rfc8259 . Owners: existing SV4/SV6/SV12 and EC41.

WSL host dependency and investment: Current adapter; replacement unproven. Microsoft documents WSL as Windows/Linux interoperability with Linux tools and ELF execution; WSL2 uses a Linux kernel in a managed utility VM. Strength: access to an established Linux execution environment; tradeoff: the estate retains a Windows/WSL host dependency. Measure our actual launch, filesystem, memory, networking and recovery costs before judging or replacing it. Current configuration invokes the sovereign connector through WSL; its 64 KiB rejection is connector policy, not evidence of a WSL limit. Existing SV10 owns the runner climb. Primary external oracle retrieved 2026-09-07, not estate-pinned: https://learn.microsoft.com/en-us/windows/wsl/about .

Latest recorded release

2026-09-16 · SV13

treat row declared for SV13 (mcp nx_assets_gate GREEN on the allowlist, api /api/tools/call name=nx_assets_gate, prim nx_assets_lib in the tree, wf costcal = the plan_run plane plan-costcal- whose step 2 runs the gate, agent costcal = the daily clock row that runs the plan): nx_compare_rank sovereignty reads TREATMENT rows=1 complete=1 verdict=TREATED against allowlist 165424 B, roster 69230 B, clock plane 11935 B; the emitter renders the Treatment verdict and the three-step table on this page from the same ruler (ecosystem EC49)

Release history and work log

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

17 of 28 capability source symbols found|4 of them authored stronger-claims|11 open|coverage 607/1000|adoption 9 full / 8 partial

Evidence profile — what the gaps on this board actually are

Unstamped. No evidence profile is published for this domain yet: run nx_swcompare_evidence sovereignty and this section fills itself on the next beat. No numbers are shown, deliberately — a zero-filled profile would read as a board with no gaps, which is the one wrong answer nobody would question.

Production map

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

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

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

Recorded priority estimates

Order from nx_compare_rank (nx_dr_ocm: (deficit + cost-of-delay + option + enables) x sponsor x self-sufficiency x momentum / cost). FINISH rows are rungs whose symbol is present but whose organ is short of full adoption: listed before new work by this heuristic. Priority is not measured delivery cost or execution readiness. Stamp: # asof=1790082167 domain=sovereignty target_version=1.0 rungs=14 done=4 open=10 finish=4 ranker=nx_dr_ocm

#	Stage	Rung	Priority	Derivation
F	FINISH	Outside-reliance gauge, whole journal, beats excluded (SV0) ch_gauge_calc dossier	REGISTERED-DARK	callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_claude_harvest_gate -- a mention, not a run
F	FINISH	The number journals itself onto this board (SV1) ch_journal dossier	REGISTERED-DARK	callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_claude_harvest_gate -- a mention, not a run
F	FINISH	The residual by class, whole population, no frequency cap (SV2) ch_classes dossier	REGISTERED-DARK	callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_claude_harvest_gate -- a mention, not a run
F	FINISH	Seat shell interlock: every harness shell call names its estate verb (SV3) tg_shell_interlock dossier	REGISTERED-DARK	callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked)
#1	1.0	Head-to-head receipt: search (SV4) sv_h2h_search dossier	600	v=6 m=1 c=10
#2	1.0	Head-to-head receipt: fetch (SV6) sv_h2h_fetch dossier	400	v=4 m=1 c=10
#3	1.0	Head-to-head receipt: the frontier seat itself (SV5) sv_h2h_llm dossier	100	v=2 m=1 c=20
#4	later	A source of record for the editor (SV12) sv_fswrite_source dossier	800	v=4 m=1 c=5
#5	later	Remote bytes and remote shell to zero (SV9) sv_ssh_zero dossier	200	v=2 m=1 c=10
#6	later	The record before July (SV11) sv_history_pre_july dossier	200	v=1 m=1 c=5
#7	later	Laptop host runner (SV10) sv_laptop_runner dossier	133	v=2 m=1 c=15
#8	later	Python to zero on the program path (SV7) sv_py_zero dossier	100	v=2 m=1 c=20
#9	later	JavaScript to zero on the page path (SV8) sv_js_zero dossier	100	v=2 m=1 c=20
#10	later	Every board's foreign data assets are listed with licence, bytes, replacement rung and lift, and the run-path rows with no replacement rung fall to zero (SV13) sv_assets_ledger dossier	0	v=0 m=2 c=20

Declared roadmap — contract, acceptance, executor, effort

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Outside-reliance gauge, whole journal, beats excluded (SV0)	ch_gauge_calc	LANDED 2026-09-05: nx_claude_harvest gauge reads the journal whole (sys_read_file, no cap), partitions verb=run estate beats out of the denominator, emits beats, claude_actions, claude_outside and claude_outside_permil beside the legacy figure; three teeth incl. the non-vacuity tooth (500 versus the diluted 200); promoted a4f43b12, 44 of 44 on the NAS; nx_claude_harvest_gate forks the subject's own selftest and pins artifact identity	Organ	0 u
The number journals itself onto this board (SV1) after SV0	ch_journal	nx_claude_harvest journal <journal> <plan> <rung> appends ONE log measure row per UTC day carrying claude_outside_permil, claude_actions, claude_outside, estate_beats_excluded, frames_total and journal_bytes, idempotent by a day marker in the row; refuses to write a -1 (insufficient data) and refuses to create a plan; wired as a plan-wsmine- row so the hourly beat journals with zero seat tokens; gate teeth J1-J3: exact permil written, second call the same day refused with the plan byte-identical, missing journal refuses and writes nothing	Organ	0.5 u
The residual by class, whole population, no frequency cap (SV2) after SV0	ch_classes	the gauge emits per-outside-tool counts over assistant actions only (Bash, PowerShell, Edit, Read, Write, Grep, Glob, WebFetch, WebSearch, Agent, ToolSearch) with the cap named and the partition summing to claude_outside; today nx_actlog freq shows the top 30 of 1,635 distinct tool:verb pairs and the four largest outside classes cover 80 percent; neg-control: a synthetic journal whose outside frames all carry one tool name yields one class equal to claude_outside	Organ	0.5 u
Seat shell interlock: every harness shell call names its estate verb (SV3) after SV1	tg_shell_interlock	SHARED CONTRACT with engineshift ES22 (declared there 2026-09-05, operator verbatim: stop using its own tools and use the estate only, build any gap in from the first byte): a laptop PreToolUse organ, sibling of nx_memnew and nx_organnew, fires on every Bash and PowerShell call and answers with the sovereign verb that does the same work, or prints GAP and the rung to declare; advisory never blocking; the one sanctioned launcher is allow-listed by path; done when the Bash plus PowerShell share of assistant actions falls in the journaled series	Organ	1 u
Head-to-head receipt: search (SV4)	sv_h2h_search	nx_shelltool grep versus ripgrep versus the builtin Grep on ONE fixed corpus and pattern set [ripgrep]: identical match sets on every pattern or the arm is VOID, wall time per arm, both organs run outside the estate as comparators only; the receipt lands as compare data rendered on this page; CAPABILITY-MISSING is declared by name if the sovereign arm cannot produce the match set. Scope clarification 2026-09-07: cover exact text/regex grep and corpus search as distinct workloads; declare corpus completeness, inaccessible files, truncation and no-match semantics, then record correctness before latency and resource cost. A missing or unreliable sovereign search primitive is investment work under SV4, never an excuse to silently substitute outside search.	Organ	1 u
Head-to-head receipt: the frontier seat itself (SV5)	sv_h2h_llm	a fixed task battery with KAT judges dispatched through nx_engine_dispatch to the local no-float seat, the external U-cleared tier and the frontier seat, scored per engine on the same tasks and graded against the field's repository-level bar [swebench]; the expected result today is a LOSS for the local seat (its ceiling is 96 new tokens by data) and that loss is the build target named on engineshift ES12 and modelwright, not a reason to hide the receipt; neg-control: an engine that never answered is recorded UNREACHABLE and never scored	Organ	2 u
Head-to-head receipt: fetch (SV6)	sv_h2h_fetch	nx_research_fetch versus curl versus the builtin fetcher on ONE fixed URL set: body bytes identical to the curl arm or the arm is VOID, provenance row present only for the sovereign arm, a PDF quote check against pdftotext where the builtin summariser fabricated on 2026-08-20; receipt rendered here. Scope clarification 2026-09-07: reading and research retrieval are capability investments; disclose source identity, byte range, completeness and transport failure separately from content. Pin rival versions and retrieval dates; a summary cannot substitute for exact source bytes when the task requires exact edits or evidence.	Organ	1 u
Python to zero on the program path (SV7)	sv_py_zero	nx_sovereignty_audit enumerates every .py in the PRODUCT zone as a PRODUCT-CONTAMINANT worklist with the organ that replaces each (oracle generators become KAT organs, patch scripts become nx_fs_write edits, pdftotext is already nx_pdf_text); the count is a ratchet that may only fall; CPython stays installable as a comparator [cpython313]; done at zero with the audit GREEN	Organ	2 u
JavaScript to zero on the page path (SV8)	sv_js_zero	every page whose logic is hand JS is listed with the NishiLang source that lowers to its wasm twin through the backend that refuses unloweredable opcodes; the count is a ratchet; Node stays a comparator [node24]; done when the wasm page census reports zero hand-JS logic pages	Organ	2 u
Remote bytes and remote shell to zero (SV9)	sv_ssh_zero	every laptop-to-NAS byte move rides nx_content_put and every remote control action rides the mgmt API; ssh and scp are counted in the journal as outside and the series must fall to the recovery lane only; the field's transport keeps its post-quantum default as the bar the sovereign door must also meet [openssh]	Organ	1 u
Laptop host runner (SV10)	sv_laptop_runner	a cap-gated sovereign runner organ on the laptop so memory-plane organs, the build farm and scheduled laptop work are API calls rather than wsl launcher lines; rule-26 never-brick fence from birth; done when the WSL launcher share of assistant actions falls in the journaled series	Organ	1.5 u
The record before July (SV11)	sv_history_pre_july	search every off-box backup the operator names for Claude transcript jsonl dated before 2026-07-20 and ingest through nx_claude_harvest frames plus ingest (marker-idempotent) so the gauge and this board cover the May-to-July build; if none exists the horizon is declared permanent on this row and the gauge prints it beside every number	Operator and organ	0.5 u
A source of record for the editor (SV12)	sv_fswrite_source	nx_fs_write and the MCP reader ship inside the tools daemon and have no source file under buildroot/runtime (nx_shelltool find, corpus_complete=1), so the editor row on this board measures the upload door's atomic-write primitive instead; done when the editor has a named source the compare watch can measure and the row is repointed. Scope clarification 2026-09-07: source ownership is the first acceptance dependency for reliable sovereign writes. Exact-edit evidence must include unique-match refusal, stale-source refusal, atomic replacement, repeat-safe behavior and byte-level readback; read and write boundaries belong in the compare record alongside the source watch.	Organ	0.5 u

Milestones

Milestone	Rungs	Cumulative
M0 · The gauge produces itself	SV0,SV1,SV2	1 u
M1 · Every class carries a receipt or a build rung	SV3,SV4,SV5,SV6	4 u
M2 · Zero on the program path	SV7,SV8,SV9,SV10,SV12	6 u
M3 · The record is whole	SV11	0.5 u

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Every board's foreign data assets are listed with licence, bytes, replacement rung and lift, and the run-path rows with no replacement rung fall to zero (SV13) after SV0	sv_assets_ledger	Every domain whose run path reads foreign bytes (weights, fonts, corpora, doors) carries a <dom>.assets file rendered by the compare emitter through nx_assets_lib (the one reader) and priced through nx_costest_lib (the one estimator), with hours from nx_costcal's calibration of u against the boards' own journals; the fleet ratchet counts runpath rows whose rung is - (NORUNG) across every .assets file and may only fall; done when every runpath row names a rung, the calibration is usable (calibrated rungs at or above the distinct-rank floor) and the ratchet reads zero	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. PARTIAL -- a declared row is unmet: the ranker refuses this board until every declared row verifies. The declared rows, the rungs and the evidence read are counted in the table below, the row partition re-derived beside its total. The complete effective clock view could not be read. Agent absence remains UNPROVEN; restore that observation before claiming completion.

Treatment verdict, counted
Class	Count	Note
rows	1	treat rows declared on the plan
complete	0	every declared step verifies against the evidence
partial	1	a declared step is unmet
malformed	0	a treat row missing a field
rows, summed	1	RECONCILED: the parts sum to the total 1
rungs	14	rungs on the plan
undeclared	13	rungs with no treat row
allowlist bytes read	207162	the MCP allowlist the ranker reads
roster bytes read	119881	the gate roster
clock plane bytes read	0	the clock DESIRED plane

Rung	MCP row	API route	Primitive	Workflow	Agent beat	Verdict
SV13	nx_assets_gate present	/api/tools/call name=nx_assets_gate present	nx_assets_lib present	costcal present	costcal declared	PARTIAL

Inspect a rung and its prerequisites

Declared nodes 14. Rank input binding: PLAN_MATRIX_BOUND_ONLY. Dependency order is authored. Implementation, acceptance evidence, authority and resource readiness are unverified. No action is recommended or dispatched here.

Plan SHA-256 cb8862ec6bd74aae1b9ee1bedc92aca3687a73aa9937d671f50c2582f027ca92. 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 14 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.

SV0 — Outside-reliance gauge, whole journal, beats excluded
Prerequisites: None declared; this does not establish execution eligibility.
Contract: ch_gauge_calc
Acceptance: LANDED 2026-09-05: nx_claude_harvest gauge reads the journal whole (sys_read_file, no cap), partitions verb=run estate beats out of the denominator, emits beats, claude_actions, claude_outside and claude_outside_permil beside the legacy figure; three teeth incl. the non-vacuity tooth (500 versus the diluted 200); promoted a4f43b12, 44 of 44 on the NAS; nx_claude_harvest_gate forks the subject's own selftest and pins artifact identity
Authored effort: 0. 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.
SV1 — The number journals itself onto this board
Prerequisites: SV0 (acceptance unverified)
Contract: ch_journal
Acceptance: nx_claude_harvest journal <journal> <plan> <rung> appends ONE log measure row per UTC day carrying claude_outside_permil, claude_actions, claude_outside, estate_beats_excluded, frames_total and journal_bytes, idempotent by a day marker in the row; refuses to write a -1 (insufficient data) and refuses to create a plan; wired as a plan-wsmine- row so the hourly beat journals with zero seat tokens; gate teeth J1-J3: exact permil written, second call the same day refused with the plan byte-identical, missing journal refuses and writes nothing
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.
SV2 — The residual by class, whole population, no frequency cap
Prerequisites: SV0 (acceptance unverified)
Contract: ch_classes
Acceptance: the gauge emits per-outside-tool counts over assistant actions only (Bash, PowerShell, Edit, Read, Write, Grep, Glob, WebFetch, WebSearch, Agent, ToolSearch) with the cap named and the partition summing to claude_outside; today nx_actlog freq shows the top 30 of 1,635 distinct tool:verb pairs and the four largest outside classes cover 80 percent; neg-control: a synthetic journal whose outside frames all carry one tool name yields one class equal to claude_outside
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.
SV3 — Seat shell interlock: every harness shell call names its estate verb
Prerequisites: SV1 (acceptance unverified)
Contract: tg_shell_interlock
Acceptance: SHARED CONTRACT with engineshift ES22 (declared there 2026-09-05, operator verbatim: stop using its own tools and use the estate only, build any gap in from the first byte): a laptop PreToolUse organ, sibling of nx_memnew and nx_organnew, fires on every Bash and PowerShell call and answers with the sovereign verb that does the same work, or prints GAP and the rung to declare; advisory never blocking; the one sanctioned launcher is allow-listed by path; done when the Bash plus PowerShell share of assistant actions falls in the journaled series
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.
SV4 — Head-to-head receipt: search
Prerequisites: None declared; this does not establish execution eligibility.
Contract: sv_h2h_search
Acceptance: nx_shelltool grep versus ripgrep versus the builtin Grep on ONE fixed corpus and pattern set [@ripgrep]: identical match sets on every pattern or the arm is VOID, wall time per arm, both organs run outside the estate as comparators only; the receipt lands as compare data rendered on this page; CAPABILITY-MISSING is declared by name if the sovereign arm cannot produce the match set. Scope clarification 2026-09-07: cover exact text/regex grep and corpus search as distinct workloads; declare corpus completeness, inaccessible files, truncation and no-match semantics, then record correctness before latency and resource cost. A missing or unreliable sovereign search primitive is investment work under SV4, never an excuse to silently substitute outside search.
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.
SV5 — Head-to-head receipt: the frontier seat itself
Prerequisites: None declared; this does not establish execution eligibility.
Contract: sv_h2h_llm
Acceptance: a fixed task battery with KAT judges dispatched through nx_engine_dispatch to the local no-float seat, the external U-cleared tier and the frontier seat, scored per engine on the same tasks and graded against the field's repository-level bar [@swebench]; the expected result today is a LOSS for the local seat (its ceiling is 96 new tokens by data) and that loss is the build target named on engineshift ES12 and modelwright, not a reason to hide the receipt; neg-control: an engine that never answered is recorded UNREACHABLE and never scored
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.
SV6 — Head-to-head receipt: fetch
Prerequisites: None declared; this does not establish execution eligibility.
Contract: sv_h2h_fetch
Acceptance: nx_research_fetch versus curl versus the builtin fetcher on ONE fixed URL set: body bytes identical to the curl arm or the arm is VOID, provenance row present only for the sovereign arm, a PDF quote check against pdftotext where the builtin summariser fabricated on 2026-08-20; receipt rendered here. Scope clarification 2026-09-07: reading and research retrieval are capability investments; disclose source identity, byte range, completeness and transport failure separately from content. Pin rival versions and retrieval dates; a summary cannot substitute for exact source bytes when the task requires exact edits or evidence.
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.
SV7 — Python to zero on the program path
Prerequisites: None declared; this does not establish execution eligibility.
Contract: sv_py_zero
Acceptance: nx_sovereignty_audit enumerates every .py in the PRODUCT zone as a PRODUCT-CONTAMINANT worklist with the organ that replaces each (oracle generators become KAT organs, patch scripts become nx_fs_write edits, pdftotext is already nx_pdf_text); the count is a ratchet that may only fall; CPython stays installable as a comparator [@cpython313]; done at zero with the audit GREEN
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.
SV8 — JavaScript to zero on the page path
Prerequisites: None declared; this does not establish execution eligibility.
Contract: sv_js_zero
Acceptance: every page whose logic is hand JS is listed with the NishiLang source that lowers to its wasm twin through the backend that refuses unloweredable opcodes; the count is a ratchet; Node stays a comparator [@node24]; done when the wasm page census reports zero hand-JS logic pages
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.
SV9 — Remote bytes and remote shell to zero
Prerequisites: None declared; this does not establish execution eligibility.
Contract: sv_ssh_zero
Acceptance: every laptop-to-NAS byte move rides nx_content_put and every remote control action rides the mgmt API; ssh and scp are counted in the journal as outside and the series must fall to the recovery lane only; the field's transport keeps its post-quantum default as the bar the sovereign door must also meet [@openssh]
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.
SV10 — Laptop host runner
Prerequisites: None declared; this does not establish execution eligibility.
Contract: sv_laptop_runner
Acceptance: a cap-gated sovereign runner organ on the laptop so memory-plane organs, the build farm and scheduled laptop work are API calls rather than wsl launcher lines; rule-26 never-brick fence from birth; done when the WSL launcher share of assistant actions falls in the journaled series
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.
SV11 — The record before July
Prerequisites: None declared; this does not establish execution eligibility.
Contract: sv_history_pre_july
Acceptance: search every off-box backup the operator names for Claude transcript jsonl dated before 2026-07-20 and ingest through nx_claude_harvest frames plus ingest (marker-idempotent) so the gauge and this board cover the May-to-July build; if none exists the horizon is declared permanent on this row and the gauge prints it beside every number
Authored effort: 0.5. Executor kind: Operator and 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.
SV12 — A source of record for the editor
Prerequisites: None declared; this does not establish execution eligibility.
Contract: sv_fswrite_source
Acceptance: nx_fs_write and the MCP reader ship inside the tools daemon and have no source file under buildroot/runtime (nx_shelltool find, corpus_complete=1), so the editor row on this board measures the upload door's atomic-write primitive instead; done when the editor has a named source the compare watch can measure and the row is repointed. Scope clarification 2026-09-07: source ownership is the first acceptance dependency for reliable sovereign writes. Exact-edit evidence must include unique-match refusal, stale-source refusal, atomic replacement, repeat-safe behavior and byte-level readback; read and write boundaries belong in the compare record alongside the source watch.
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.
SV13 — Every board's foreign data assets are listed with licence, bytes, replacement rung and lift, and the run-path rows with no replacement rung fall to zero
Prerequisites: SV0 (acceptance unverified)
Contract: sv_assets_ledger
Acceptance: Every domain whose run path reads foreign bytes (weights, fonts, corpora, doors) carries a <dom>.assets file rendered by the compare emitter through nx_assets_lib (the one reader) and priced through nx_costest_lib (the one estimator), with hours from nx_costcal's calibration of u against the boards' own journals; the fleet ratchet counts runpath rows whose rung is - (NORUNG) across every .assets file and may only fall; done when every runpath row names a rung, the calibration is usable (calibrated rungs at or above the distinct-rank floor) and the ratchet reads zero
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 sovereignty 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.

SV0 Outside-reliance gauge, whole journal, beats excluded · cell PRESENT · gate NO-GATE-DECLARED

Contract	`ch_gauge_calc` · executor Organ · 0 u · stands on nothing
Definition of done	LANDED 2026-09-05: nx_claude_harvest gauge reads the journal whole (sys_read_file, no cap), partitions verb=run estate beats out of the denominator, emits beats, claude_actions, claude_outside and claude_outside_permil beside the legacy figure; three teeth incl. the non-vacuity tooth (500 versus the diluted 200); promoted a4f43b12, 44 of 44 on the NAS; nx_claude_harvest_gate forks the subject's own selftest and pins artifact identity
Matrix cell	Outside-reliance gauge, full population, estate beats excluded · `runtime/_hdl_build/nx_claude_harvest.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_claude_harvest`; declare `nx_claude_harvest TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-05 measure the live /outsidegap read 592 permil over 79,634 frames with truncated=1: the 4 MB read cap had been sampling the July head of a 54 MB journal, and the journal mixes assistant calls with estate beat runs 2026-09-05 land nx_claude_harvest gauge fixed: sys_read_file whole-journal read, verb=run beats partitioned out, new keys beats/claude_actions/claude_outside/claude_outside_permil, three teeth T21-T23 incl. non-vacuity; local WSL selftest 43 of 44 with the one RED (T14) pre-existing and environmental, proven by building the banked original (40 of 41, same tooth) 2026-09-05 lesson a build refused by admission is not lost: the NAS was in an I/O storm (procs_blocked 8 of 8) and the build was QUEUED (BQ row, rc 0); forcing a compile into the storm would have been the resource bug the shipping criterion names, so the queue fired it ten minutes later 2026-09-05 land nx_claude_harvest PROMOTED a4f43b12 through nx_organ_ship: content GREEN (no run lost), offc twin refreshed, registered; prove read NO-GATE-FOUND because the proof lived only in a verb -- the gate rung was declared in the same hour 2026-09-05 measure THE HONEST NUMBER over the whole journal (54,291,072 B, truncated 0): frames_total 685665, beats 274104, claude_actions 411561, claude_outside 47230, claude_outside_permil 114; legacy all-frames 166; banked as minergap MG13 2026-09-05 measure the promoted binary's own selftest on the NAS: 44 of 44 verdict GREEN (T14 green there as its comment predicted); nx_actlog sibling cap fixed and promoted e87b8148, its selftest 15 of 15 GREEN; nx_actlog freq over 685,920 frames names the residual: Bash 12584, PowerShell 11588, Edit 9398, Read 4348, index_overflow 1 declared 2026-09-05 land nx_claude_harvest_gate written and landed (4321 B): forks the just-built or promoted subject's selftest, exit code is the verdict, T4 pins artifact identity on the partition tooth; build queued under CPU saturation (procs_running 10 of 8) 2026-09-05 lesson a self-audit for half wiring found four gaps in the same session that shipped the fix: the proof asserted from a different build than the promoted one, a possible stale third twin, the sibling miner's identical cap, and no gate for the ship loop to resolve -- each was closed by a measurement, not a sentence 2026-09-05 land THE EMITTER GREW (operator, on this page: rows read as random glyphs and a partial adoption printed as a badge). nx_swcompare_matrix now renders a derived Pros / Cons strip per row from the same cells (has / leads / in part; ahead of whom; behind whom, leads signed; open contract; no organ measured; not adopted yet with the catalog lib's own remedy; and, when nothing is measured against Nishi, none on the measured axes -- never a victory claim), and the PARTIAL adoption chip now carries the remedy text as a CON with the raw ladder token only on hover. Fleet-wide: every domain inherits it on the next beat. PROMOTED 80d97b58 through nx_organ_ship with prove GREEN on BOTH declared gates (nx_imgattr_lib_gate 42 of 42, nx_swcompare_shape_gate 11 of 11 on the NAS), content GREEN, offc fresh. The new gate is named for the PROPERTY and bound by an organ_gate.conf row; it caught two of my own mistakes before shipping: a tooth that assumed a fixture organ would read adopted (a census-less host reads it unadopted -- the tooth was wrong, not the emitter) and an exec of the copied subject from /tmp, which is noexec on the NAS (gen_exit 127) -- fixed by forking the subject by absolute path with only the CWD inside the fixture 2026-09-05 measure EMITTER VERIFIED ON THE LIVE SURFACE: regen job 1788628509 published sovereignty index.html at 93647 B (prev 84660, the delta is the per-row Pros and Cons strips), nx_page_verify status=200 a11y-issues=0 VERDICT=GREEN through the pinned sovereign edge. The regen's own verdict was RED for three other domains (observability present<5, intelmine a stray bar in a note), not this one. Two 404s from nx_https_get on two pages were ONE vantage defect (the NAS hairpins nishifamily.com to the nginx default vhost; pass 127.0.0.1:8443 or use nx_page_verify), not two missing pages 2026-09-05 land nx_restage PROMOTED-FIRST SHIPPED (sha 06a3818a88ef8683, 29184 B, live == staged, offc twin installed, registered): when the promoted serving-root twin ../<target>.elf exists it is COPIED into buildroot/_offc and the receipt says source PROMOTED; the sovereign-lane rebuild is now the fallback for a never-promoted target and the receipt says source BUILT. Reason: nx_compare_regen's fork-freshness hashes _offc against the promoted twin and names nx_restage as the remedy, but a rebuild through the sovereign lane is byte-different from the /api/build lane on identical source (measured 2026-09-05 on nx_swcompare_matrix: 332217 B rebuilt vs 368016 B promoted), so the old remedy could never read FRESH. Adopt-gate lost 2 of 52 runs (38 permil), both NAMED rewords (the /tmp error text and the JSON literal split to carry source), allowed explicitly. Caveat carried from the ship receipt: nx_restage_gate (8 of 8) ran from an artifact built BEFORE this edit because its rebuild was refused by admission and queued (BQ-nx_restage_gate); that gate proves the target sanitizer r_san, which this change does not touch 2026-09-05 lesson A SHIP LOOP THAT SENDS AN IDENTICAL BUILD BODY TWICE IN AN HOUR PROMOTES THE PREVIOUS GENERATION WITH PROVE GREEN. Measured 2026-09-05 on nx_actlog: live_sha equal to the prior promote, promoted selftest still 15 of 15 against a 65611 B source that adds T16 and T17, nx_actlog_gate 6 of 6 because its identity tooth (the confidence string) exists in both generations. Two root fixes shipped the same hour, not a memory note: nx_organ_ship now appends a per-run bust nonce to both /api/build bodies (debt 1785363417 recurrence, open since 2026-07-29), and nx_actlog_gate T5 pins the derived-index tooth that exists only in the new generation. Law: a gate whose identity tooth pins a string both generations carry certifies the predecessor 2026-09-05 retract RETRACT of the SV0 lesson row above, same day: the stale nx_actlog promote was not an idempotency replay (no client sends the header the mgmt dedup keys on; proven by a coverage-complete grep). It was nx_hostctl cmd_buildrun staging the PREVIOUS artifact after the builder refused under its own admission check -- an exit code printed and then ignored, an artifact path shared between a refused run and a successful one. Root fix in source: the builder's exit 0 plus its own success token gate the stage. The law survives with its subject corrected: a receipt that says BUILT for a run that did not build is the false-green class, and the gate whose identity tooth pins a string both generations carry cannot catch it

SV1 The number journals itself onto this board · cell PRESENT · gate NO-GATE-DECLARED

Contract	`ch_journal` · executor Organ · 0.5 u · stands on SV0
Definition of done	nx_claude_harvest journal <journal> <plan> <rung> appends ONE log measure row per UTC day carrying claude_outside_permil, claude_actions, claude_outside, estate_beats_excluded, frames_total and journal_bytes, idempotent by a day marker in the row; refuses to write a -1 (insufficient data) and refuses to create a plan; wired as a plan-wsmine- row so the hourly beat journals with zero seat tokens; gate teeth J1-J3: exact permil written, second call the same day refused with the plan byte-identical, missing journal refuses and writes nothing
Matrix cell	Journaled reliance series on this board · `runtime/_hdl_build/nx_claude_harvest.nx` · PRESENT decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_claude_harvest`; declare `nx_claude_harvest TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-05 measure [HARV:journal:day=20701] claude_outside_permil=114 claude_actions=412416 claude_outside=47230 estate_beats_excluded=274220 frames_total=686636 journal_bytes=54374219 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-05 land ch_journal SHIPPED: nx_claude_harvest journal verb PROMOTED 699388fb through nx_organ_ship with prove=GREEN via nx_claude_harvest_gate (resolved by the target_gate convention, bank CURRENT) and the ONE named lost run (the old usage line, reworded to add the verb) journaled as ADOPT-LOST under allow_loss=1; first live run on the NAS appended the measure row above from 686,636 frames; wired as plan-wsmine- row 115 so the hourly beat journals with zero seat tokens; local teeth J1-J3 PASS (46 of 47, T14 environmental) 2026-09-06 measure [HARV:journal:day=20702] claude_outside_permil=111 claude_actions=425058 claude_outside=47231 estate_beats_excluded=276593 frames_total=701651 journal_bytes=55646978 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4058 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-07 measure [HARV:journal:day=20703] claude_outside_permil=104 claude_actions=451696 claude_outside=47234 estate_beats_excluded=284047 frames_total=735743 journal_bytes=58522901 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4061 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-08 measure [HARV:journal:day=20704] claude_outside_permil=102 claude_actions=459313 claude_outside=47235 estate_beats_excluded=288011 frames_total=747324 journal_bytes=59451455 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4062 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-09 measure [HARV:journal:day=20705] claude_outside_permil=100 claude_actions=469102 claude_outside=47236 estate_beats_excluded=293927 frames_total=763029 journal_bytes=60699417 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4063 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-10 measure [HARV:journal:day=20706] claude_outside_permil=99 claude_actions=474723 claude_outside=47237 estate_beats_excluded=303386 frames_total=778109 journal_bytes=61915197 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4064 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-11 measure [HARV:journal:day=20707] claude_outside_permil=84 claude_actions=559045 claude_outside=47237 estate_beats_excluded=311631 frames_total=870676 journal_bytes=85514890 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4064 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-12 measure [HARV:journal:day=20708] claude_outside_permil=80 claude_actions=584987 claude_outside=47237 estate_beats_excluded=311631 frames_total=896618 journal_bytes=92719616 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4064 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-13 measure [HARV:journal:day=20709] claude_outside_permil=74 claude_actions=637955 claude_outside=47237 estate_beats_excluded=311631 frames_total=949586 journal_bytes=107450195 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4064 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-14 measure [HARV:journal:day=20710] claude_outside_permil=63 claude_actions=743879 claude_outside=47282 estate_beats_excluded=311718 frames_total=1055597 journal_bytes=137109959 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4109 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-15 measure [HARV:journal:day=20711] claude_outside_permil=61 claude_actions=767176 claude_outside=47283 estate_beats_excluded=320936 frames_total=1088112 journal_bytes=144297871 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4110 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-16 measure [HARV:journal:day=20712] claude_outside_permil=60 claude_actions=787979 claude_outside=47284 estate_beats_excluded=329538 frames_total=1117517 journal_bytes=150829521 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4111 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-17 measure [HARV:journal:day=20713] claude_outside_permil=57 claude_actions=816685 claude_outside=47285 estate_beats_excluded=338506 frames_total=1155191 journal_bytes=159651035 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4112 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-18 measure [HARV:journal:day=20714] claude_outside_permil=55 claude_actions=852090 claude_outside=47286 estate_beats_excluded=347908 frames_total=1199998 journal_bytes=170350529 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4113 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-19 measure [HARV:journal:day=20715] claude_outside_permil=54 claude_actions=872379 claude_outside=47287 estate_beats_excluded=357331 frames_total=1229710 journal_bytes=176745546 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4114 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-20 measure [HARV:journal:day=20716] claude_outside_permil=51 claude_actions=915877 claude_outside=47287 estate_beats_excluded=366261 frames_total=1282138 journal_bytes=189830207 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4114 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-21 measure [HARV:journal:day=20717] claude_outside_permil=48 claude_actions=970624 claude_outside=47288 estate_beats_excluded=374616 frames_total=1345240 journal_bytes=205925134 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4115 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day) 2026-09-22 measure [HARV:journal:day=20718] claude_outside_permil=45 claude_actions=1039963 claude_outside=47289 estate_beats_excluded=383181 frames_total=1423144 journal_bytes=226052090 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4116 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)

SV2 The residual by class, whole population, no frequency cap · cell LANDED · gate NO-GATE-DECLARED

Contract	`ch_classes` · executor Organ · 0.5 u · stands on SV0
Definition of done	the gauge emits per-outside-tool counts over assistant actions only (Bash, PowerShell, Edit, Read, Write, Grep, Glob, WebFetch, WebSearch, Agent, ToolSearch) with the cap named and the partition summing to claude_outside; today nx_actlog freq shows the top 30 of 1,635 distinct tool:verb pairs and the four largest outside classes cover 80 percent; neg-control: a synthetic journal whose outside frames all carry one tool name yields one class equal to claude_outside
Matrix cell	Residual by class, whole partition, no frequency cap · `runtime/_hdl_build/nx_claude_harvest.nx` · LANDED decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_claude_harvest`; declare `nx_claude_harvest TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-05 queue IN FLIGHT (crash ledger): (1) SV2 classes verb source LANDED on the NAS byte-identical to the laptop (nx_claude_harvest.nx 145114 B sha256 0b036e99d3e3f530..., five size-CAS edits, definitions first; laptop SOVEREIGN build, selftest 49/50 with the one red = T14 environmental, C1-C3 GREEN) -- ship pending on the /api/build lease via nx_organ_ship (PROVE = nx_claude_harvest_gate). (2) nx_actlog_gate (end-to-end gate for the action miner, closes the prove=NO-GATE-FOUND hole) landed 4373 B, BUILT on the NAS by the BQ queue (planrun-build row rc=0, 30951 B) -- promote+register pending (re-run nx_organ_ship nx_actlog_gate, build-cache HIT). (3) nx_restage rewritten PROMOTED-FIRST (copies ../<target>.elf into buildroot/_offc when a promoted twin exists, announces source, rebuild is the fallback) 8731 B under content-CAS; nx_organ_ship nx_restage job 1788629928 running. Recipe if this seat dies: nx_catalog the three names, then nx_organ_ship each; journal a land row per promote 2026-09-05 land SV2 SHIPPED: nx_claude_harvest promoted 6f5c4bf6459034567a461e59fc104107ed78419a8fc73c119aff721e60f445f6 through nx_organ_ship, PROVE nx_claude_harvest_gate 6 of 6 GREEN (closure bank CURRENT), adopt-gate 1 named lost run of 696 (the usage line reworded to add the verb), offc twin byte-identical. The classes verb answers the residual BY TOOL over claude-action frames with estate beats excluded, the top 12 plus an other bucket so the partition sums to claude_outside, a 512-name table whose overflow is COUNTED, and the daily journal row now carries outside_by_tool=Tool:N,...,other:K for the top 6. The contract symbol ch_classes is DECLARED (nx_symdecl_lib applies the declaration rule to .nx organs and never falls back to a substring, so ch_classes_json alone could not flip the cell) and CALLED by the verb. Teeth C1 one-tool journal yields one class equal to claude_outside with other 0, C2 beats and nishi tools never a class, C3 sorted descending with an exact sum Bash:3,Edit:2,Grep:1,other:0 of 6 -- the plan's pre-declared done-rule as written. NAS source == laptop by sha256 5c27b81111f3a802 after five size-CAS edits

SV3 Seat shell interlock: every harness shell call names its estate verb · cell LANDED · gate NO-GATE-DECLARED

Contract	`tg_shell_interlock` · executor Organ · 1 u · stands on SV1
Definition of done	SHARED CONTRACT with engineshift ES22 (declared there 2026-09-05, operator verbatim: stop using its own tools and use the estate only, build any gap in from the first byte): a laptop PreToolUse organ, sibling of nx_memnew and nx_organnew, fires on every Bash and PowerShell call and answers with the sovereign verb that does the same work, or prints GAP and the rung to declare; advisory never blocking; the one sanctioned launcher is allow-listed by path; done when the Bash plus PowerShell share of assistant actions falls in the journaled series
Matrix cell	Seat shell interlock (every harness shell call names its estate verb) · `runtime/_hdl_build/nx_tokgate.nx` · LANDED decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_tokgate`; declare `nx_tokgate TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-05 land tg_shell_interlock landed on the shared contract with engineshift ES22 (that ledger carries the measurements): a PreToolUse organ on every harness Bash and PowerShell call naming the estate verb from hooks/shellverb.conf or GAP, gate 32 of 32, laptop-installed and wired; this board's watch row had named a source path that does not exist (runtime/nx_tokgate.nx) and is now runtime/_hdl_build/nx_tokgate.nx so the cell can flip on the next emit; the done-rule half that remains is the journaled series: the Bash plus PowerShell share of assistant actions must fall, measured by nx_actlog over the coming sessions

SV4 Head-to-head receipt: search · cell WATCHING · gate NO-GATE-DECLARED

Contract	`sv_h2h_search` · executor Organ · 1 u · stands on nothing
Definition of done	nx_shelltool grep versus ripgrep versus the builtin Grep on ONE fixed corpus and pattern set [ripgrep]: identical match sets on every pattern or the arm is VOID, wall time per arm, both organs run outside the estate as comparators only; the receipt lands as compare data rendered on this page; CAPABILITY-MISSING is declared by name if the sovereign arm cannot produce the match set. Scope clarification 2026-09-07: cover exact text/regex grep and corpus search as distinct workloads; declare corpus completeness, inaccessible files, truncation and no-match semantics, then record correctness before latency and resource cost. A missing or unreliable sovereign search primitive is investment work under SV4, never an excuse to silently substitute outside search.
Matrix cell	Head-to-head receipt: search · `runtime/nx_sov_h2h.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_sov_h2h`; declare `nx_sov_h2h TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-05 measure six Sept-2026 field references mirrored over the sovereign TLS client and pinned by provenance sha256 (ripgrep, SWE-bench, CPython 3.13, Node 24, OpenSSH, MCP 2025-06-18); the specific claims cited were witnessed by grep in each mirror before being written 2026-09-07 queue OPERATOR SCOPE 2026-09-07: all operational work is sovereign capability investment, including grep/search/read/write/build/test/site emission/publication/recovery/AI. Existing goal and rung contracts expanded; this row records requested scope only, not implementation, build, publication or functional acceptance. Next: measure the verified canonical owners and their blockers, attach comparable dated benchmark receipts, and render the changes on the existing /compare boards.

SV5 Head-to-head receipt: the frontier seat itself · cell WATCHING · gate NO-GATE-DECLARED

Contract	`sv_h2h_llm` · executor Organ · 2 u · stands on nothing
Definition of done	a fixed task battery with KAT judges dispatched through nx_engine_dispatch to the local no-float seat, the external U-cleared tier and the frontier seat, scored per engine on the same tasks and graded against the field's repository-level bar [swebench]; the expected result today is a LOSS for the local seat (its ceiling is 96 new tokens by data) and that loss is the build target named on engineshift ES12 and modelwright, not a reason to hide the receipt; neg-control: an engine that never answered is recorded UNREACHABLE and never scored
Matrix cell	Head-to-head receipt: the frontier seat · `runtime/nx_sov_h2h.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_sov_h2h`; declare `nx_sov_h2h TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SV6 Head-to-head receipt: fetch · cell WATCHING · gate NO-GATE-DECLARED

Contract	`sv_h2h_fetch` · executor Organ · 1 u · stands on nothing
Definition of done	nx_research_fetch versus curl versus the builtin fetcher on ONE fixed URL set: body bytes identical to the curl arm or the arm is VOID, provenance row present only for the sovereign arm, a PDF quote check against pdftotext where the builtin summariser fabricated on 2026-08-20; receipt rendered here. Scope clarification 2026-09-07: reading and research retrieval are capability investments; disclose source identity, byte range, completeness and transport failure separately from content. Pin rival versions and retrieval dates; a summary cannot substitute for exact source bytes when the task requires exact edits or evidence.
Matrix cell	Head-to-head receipt: fetch · `runtime/nx_sov_h2h.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_sov_h2h`; declare `nx_sov_h2h TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SV7 Python to zero on the program path · cell WATCHING · gate NO-GATE-DECLARED

Contract	`sv_py_zero` · executor Organ · 2 u · stands on nothing
Definition of done	nx_sovereignty_audit enumerates every .py in the PRODUCT zone as a PRODUCT-CONTAMINANT worklist with the organ that replaces each (oracle generators become KAT organs, patch scripts become nx_fs_write edits, pdftotext is already nx_pdf_text); the count is a ratchet that may only fall; CPython stays installable as a comparator [cpython313]; done at zero with the audit GREEN
Matrix cell	Python residual ratchet on the program path · `runtime/nx_sovereignty_audit.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_sovereignty_audit`; declare `nx_sovereignty_audit TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SV8 JavaScript to zero on the page path · cell WATCHING · gate NO-GATE-DECLARED

Contract	`sv_js_zero` · executor Organ · 2 u · stands on nothing
Definition of done	every page whose logic is hand JS is listed with the NishiLang source that lowers to its wasm twin through the backend that refuses unloweredable opcodes; the count is a ratchet; Node stays a comparator [node24]; done when the wasm page census reports zero hand-JS logic pages
Matrix cell	JavaScript residual ratchet on the page path · `runtime/nx_wasmpage_census.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_wasmpage_census`; declare `nx_wasmpage_census TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SV9 Remote bytes and remote shell to zero · cell WATCHING · gate UNRECORDED

Contract	`sv_ssh_zero` · executor Organ · 1 u · stands on nothing
Definition of done	every laptop-to-NAS byte move rides nx_content_put and every remote control action rides the mgmt API; ssh and scp are counted in the journal as outside and the series must fall to the recovery lane only; the field's transport keeps its post-quantum default as the bar the sovereign door must also meet [openssh]
Matrix cell	Remote-bytes residual ratchet · `runtime/nx_content_put.nx` · WATCHING decl
Gate	`nx_cdc_gate` · UNRECORDED — no recorded verdict for this gate (it is declared for the organ but the evidence referee's last run did not record it: it is not on this board's gates map, or the referee has not run since the per-gate record shipped)
Work log	no log row names this rung yet

SV10 Laptop host runner · cell WATCHING · gate NO-GATE-DECLARED

Contract	`sv_laptop_runner` · executor Organ · 1.5 u · stands on nothing
Definition of done	a cap-gated sovereign runner organ on the laptop so memory-plane organs, the build farm and scheduled laptop work are API calls rather than wsl launcher lines; rule-26 never-brick fence from birth; done when the WSL launcher share of assistant actions falls in the journaled series
Matrix cell	Laptop host runner (the wsl launcher class) · `runtime/nx_laptop_runner.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_laptop_runner`; declare `nx_laptop_runner TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-16 measure A WITNESS FOR THE LAPTOP HOST RUNNER: a five-target stage upload written as a harness shell loop failed all five (the loop variable's dotted name parsed as a property access, the door then read the url as its cap file); five explicit calls succeeded; the durable fix is this rung's organ, and every harness shell loop until then is the class this rung retires

SV11 The record before July · cell WATCHING · gate NO-GATE-DECLARED

Contract	`sv_history_pre_july` · executor Operator and organ · 0.5 u · stands on nothing
Definition of done	search every off-box backup the operator names for Claude transcript jsonl dated before 2026-07-20 and ingest through nx_claude_harvest frames plus ingest (marker-idempotent) so the gauge and this board cover the May-to-July build; if none exists the horizon is declared permanent on this row and the gauge prints it beside every number
Matrix cell	Pre-July record recovery · `runtime/_hdl_build/nx_claude_harvest.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_claude_harvest`; declare `nx_claude_harvest TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	2026-09-10 measure 2026-09-10 bounded coverage audit: existing native harvester source runtime/_hdl_build/nx_claude_harvest.nx found171680B with frames/delta/learn/beat/journal interfaces; main runtime path was absent, so discovery must resolve actual owner. Existing nx_memory MCP default attachment denied; same installed tool succeeded immediately through existing scoped MCP connection, listing2924 store lines. Local main Claude project167 top-level JSONL files observed, not proof of complete historical access or mined requirements. NAS knowledge path-name search for transcript returned0 of15197 files; that scope does not imply no transcript archive. Existing action journal/learn owner must be reused for requirement-to-board mapping; EC2 carries audit status, exact original-request coverage remains unverified. No private transcript or legal source sent to Claude.

SV12 A source of record for the editor · cell WATCHING · gate NO-GATE-DECLARED

Contract	`sv_fswrite_source` · executor Organ · 0.5 u · stands on nothing
Definition of done	nx_fs_write and the MCP reader ship inside the tools daemon and have no source file under buildroot/runtime (nx_shelltool find, corpus_complete=1), so the editor row on this board measures the upload door's atomic-write primitive instead; done when the editor has a named source the compare watch can measure and the row is repointed. Scope clarification 2026-09-07: source ownership is the first acceptance dependency for reliable sovereign writes. Exact-edit evidence must include unique-match refusal, stale-source refusal, atomic replacement, repeat-safe behavior and byte-level readback; read and write boundaries belong in the compare record alongside the source watch.
Matrix cell	Editor source of record · `runtime/nx_fs_write.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_fs_write`; declare `nx_fs_write TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

SV13 Every board's foreign data assets are listed with licence, bytes, replacement rung and lift, and the run-path rows with no replacement rung fall to zero · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`sv_assets_ledger` · executor Organ · 2 u · stands on SV0
Definition of done	Every domain whose run path reads foreign bytes (weights, fonts, corpora, doors) carries a <dom>.assets file rendered by the compare emitter through nx_assets_lib (the one reader) and priced through nx_costest_lib (the one estimator), with hours from nx_costcal's calibration of u against the boards' own journals; the fleet ratchet counts runpath rows whose rung is - (NORUNG) across every .assets file and may only fall; done when every runpath row names a rung, the calibration is usable (calibrated rungs at or above the distinct-rank floor) and the ratchet reads zero
Matrix cell	NO-MATRIX-ROW — no matrix row carries `sv_assets_ledger`; 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	2026-09-16 land treat row declared for SV13 (mcp nx_assets_gate GREEN on the allowlist, api /api/tools/call name=nx_assets_gate, prim nx_assets_lib in the tree, wf costcal = the plan_run plane plan-costcal- whose step 2 runs the gate, agent costcal = the daily clock row that runs the plan): nx_compare_rank sovereignty reads TREATMENT rows=1 complete=1 verdict=TREATED against allowlist 165424 B, roster 69230 B, clock plane 11935 B; the emitter renders the Treatment verdict and the three-step table on this page from the same ruler (ecosystem EC49) 2026-09-16 land THE FOREIGN-ASSET LEDGER LANDED AS EMITTER CAPABILITY, NOT A SIDE PAGE (operator 2026-09-16: know via /compare what the lift would be to replace a third party we use with our own, in cost): nx_assets_lib (row grammar, states runpath candidate oracle unavailable, kinds, first-failing-rule verdicts, partitions that sum, three-point aggregates) and nx_costest_lib (deci-u three-point estimates, root-sum-square band beside the correlated sum, quantile rank rule with a derived distinct-rank floor, the 6ND and 2ND compute rules, the calibration census and the one ledger grammar) with gates nx_assets_gate 36/36 and nx_costest_gate 63/63 on the laptop lane; assets_pass in nx_swcompare_lib renders the table and the api.json object on both generators; nx_costcal writes knowledge/status/costcal.ledger from every plan's rung costs and land rows (more than a thousand land rows on the boards); first data: search.assets (11 rows, the MiniLM weights as the only runpath foreign weights, five open candidates, two unavailable rank-1 artifacts, two oracles, the GPU door) and llm.assets (the Qwen2.5 GGUFs, replaced by LM16)

dossiers=14 cells_joined=13 no_matrix_row=1 (sum 14) gates_recorded=0 gates_unrecorded=1 no_gate_declared=12 (sum 13 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. 28 feature rows mapped of 28 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/sovereignty.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 600–800, distinctiveness 200–550 of 0–1000 permil (the full range put every player in one corner)
readability of the 2D cut, self-graded by the layout ruler: label overlaps 0 · labels over marks 0 · off-canvas 0 · unresolved labels 0 · mark overlaps 1 (a fact of the data: two players that close are that close) · data spread 649 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 600–800 · distinctiveness 200–550 · breadth 200–550 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 2 · unresolved labels 0 · mark overlaps 1 (a fact of the data: two players that close are that close) · data spread 235 permil of the plot

Nishi Claude Code builtin tools GNU userland (bash, coreutils, grep, ssh, curl) CPython 3.13 Node.js 24 (V8 13.6)
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 13 (a fact of the data: two players that close are that close) · data spread 536 permil of the plot

Player	Quadrant	centrality	distinctiveness	breadth	depth	momentum	Rows held	Days on spine	First seen
Nishi	Aspirational	693	490	452	745	0 since 2026-09-23	17	10	2026-09-15
Claude Code builtin tools	Aspirational	758	339	369	738	0 since 2026-09-23	14	10	2026-09-15
GNU userland (bash, coreutils, grep, ssh, curl)	Unconventional	677	371	333	848	0 since 2026-09-23	11	10	2026-09-15
CPython 3.13	Peripheral	645	233	261	733	0 since 2026-09-23	10	10	2026-09-15
Node.js 24 (V8 13.6)	Peripheral	645	233	261	733	0 since 2026-09-23	10	10	2026-09-15

Day	Matrix rows reviewed	Players recorded
2026-09-15	26	5
2026-09-16	26	5
2026-09-17	28	5
2026-09-18	28	5
2026-09-19	28	5
2026-09-20	28	5
2026-09-21	28	5
2026-09-22	28	5
2026-09-23	28	5
2026-10-08	28	5

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 112, unassessed 0. Authored assessments are not verified outcomes.

Capability	Nishi	Claude Code builtin tools	GNU userland (bash, coreutils, grep, ssh, curl)	CPython 3.13	Node.js 24 (V8 13.6)
Outside-reliance gauge, full population, estate beats excluded Source present; authored stronger claim: `ch_gauge_calc` in `runtime/_hdl_build/nx_claude_harvest.nx`, verified at emit. MEASURED 2026-09-05: claude_outside_permil=114 over 411561 assistant actions since 2026-07-20 (47230 outside, 274104 estate beats excluded, 685665 frames, journal 54 MB read whole, truncated=0). No rival measures its own outside share, and the July-head sample that read 592 is retired. The exceed: a per-action reliance meter that runs on a beat and journals itself to this board (SV1). Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_claude_harvest_gate -- a mention, not a run. 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); only referenced by nx_claude_harvest_gate -- a mention, not a run Pros Source symbol found; stronger capability claim is authored; ahead of Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13, Node.js 24 (V8 13.6)Cons not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_claude_harvest_gate -- a mention, not a run	◉	○	○	○	○
Shell and pipeline execution (Bash, PowerShell) Measured: `s_walk` exists in `runtime/nx_shelltool.nx`, verified at emit. The largest residual class: Bash 12584 plus PowerShell 11588 of 47230 outside actions (nx_actlog freq over the full journal). The sovereign shell tier is nx_shelltool (grep, glob, find under a declared budget) plus the organs themselves; the kill rung SV3 shares engineshift ES22. 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:actlog` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind Claude Code builtin tools (leads), GNU userland (bash, coreutils, grep, ssh, curl) (leads)	●	◉	◉	●	●
Seat shell interlock (every harness shell call names its estate verb) Measured: `tg_shell_interlock` exists in `runtime/_hdl_build/nx_tokgate.nx`, verified at emit. SHARED CONTRACT with engineshift ES22 (declared 2026-09-05): a laptop PreToolUse organ that fires on every Bash and PowerShell call a seat attempts and answers with the sovereign verb, or prints GAP and the rung to declare; advisory, never blocking. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_tokgate_gate -- a mention, not a run. 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); only referenced by nx_tokgate_gate -- a mention, not a run Pros Implementing source symbol found; runtime acceptance separate; ahead of Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13, Node.js 24 (V8 13.6)Cons not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_tokgate_gate -- a mention, not a run	●	○	○	○	○
File read (Read builtin) Measured: `fs_cat` exists in `runtime/_hdl_build/nx_fs.nx`, verified at emit. Read 4348 outside actions. The digest-first twin is nx_fs (size, lines, outline, tail before any whole read); the 4 KB coreutils multitool carries fs_cat, the MCP reader ships inside the tools daemon. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIVE` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind GNU userland (bash, coreutils, grep, ssh, curl) (leads)	●	●	◉	●	●
File edit and atomic write (Edit, Write builtins) Measured: `cp_write_atomic` exists in `runtime/nx_content_put.nx`, verified at emit. Edit 9398 outside actions. The sovereign editor is nx_fs_write (exact-match edit with a uniqueness contract, atomic write, append, compare-and-swap expect=); its source ships inside the tools daemon, so the measured cell is the atomic-write primitive of the upload door. Naming a source of record for nx_fs_write is hygiene rung SV12. 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:cron` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	●	●	●	●
Search and grep (Grep, Glob, Select-String) Measured: `s_grep_file` exists in `runtime/nx_shelltool.nx`, verified at emit. Literal-substring grep, glob and find over the NAS tree under a declared byte and deadline budget, with the BM25 corpus search and the crawled-web index beside it. Oracle: ripgrep [ripgrep]. Head-to-head receipt = SV4. Adoption: `RUN-BY:actlog` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; fully adopted on the estate ladderCons behind GNU userland (bash, coreutils, grep, ssh, curl) (leads)	●	●	◉	●	●
Web fetch (WebFetch) Source present; authored stronger claim: `rf_provenance` in `runtime/nx_research_fetch.nx`, verified at emit. Sovereign TLS 1.3 fetch that writes a provenance row (url, mirror, sha256, bytes) per fetch; the six mirrors this page cites were fetched by it today. The exceed: provenance-pinned mirrors by construction, where the builtin fetcher summarises and produced quotes absent from a PDF on 2026-08-20 (measured twice). Receipt = SV6. 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:actlog` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13, Node.js 24 (V8 13.6); fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	●	●	●	●
Web search (WebSearch) Measured: `wss_urlenc` exists in `runtime/nx_websearch_cli.nx`, verified at emit. Sovereign open-web plus Common-Crawl index behind nishifamily.com/search exposed as a tool; the July window counted WebSearch 14, WebFetch 12 and Agent 50 outside research calls. 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 GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13, Node.js 24 (V8 13.6)Cons not adopted yet: source exists, never compiled: /api/build it	●	●	○	○	○
Agent and sub-task orchestration (Task, Agent) Open — no implementing organ is measured for this axis yet. Workflows as data: a plan is rows in a plane and one cap-gated call runs the sequence. The seat-side fan-out class is what nx_goalmap retired on 2026-09-05 (5.7 million subagent tokens for one goal map, then one job). rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Pros none measured yetCons behind Claude Code builtin tools; no implementing organ measured	○	●	○	○	○
LLM reasoning and code generation (the frontier seat itself) Measured: `ed_dispatch` exists in `runtime/nx_engine_dispatch.nx`, verified at emit. The headline gap. The dispatcher ladders local, external and frontier engines by declared class with a KAT judge and a shift ledger; the local no-float seat's ceiling is 96 new tokens by data, so code authoring is frontier-only today. Oracle: SWE-bench [swebench]. Head-to-head receipt = SV5; exceeding the assistant is the build target, not a claim. 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 GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13, Node.js 24 (V8 13.6)Cons behind Claude Code builtin tools (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)	●	◉	○	○	○
Python runtime on the program path Measured: `nx_pdf_extract_text` exists in `runtime/nx_pdf_text.nx`, verified at emit. pdftotext, oracle generators and patch scripts are the residual; the sovereign PDF extractor replaced poppler for citation reading. The runtime being retired is CPython 3.13 [cpython313]. Rung SV7 drives every .py on the program path to zero (nx_sovereignty_audit classifies each as PRODUCT-CONTAMINANT). Adoption: `RUN-BY:fork:nx_pdf_gate` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl), Node.js 24 (V8 13.6); fully adopted on the estate ladderCons behind CPython 3.13 (leads)	●	○	○	◉	○
JavaScript runtime on the page path Measured: `wat_emit_module` exists in `runtime/wasm.nx`, verified at emit. The same NishiLang source lowers to wasm and the backend refuses any opcode it cannot lower (LN13a); page logic still written by hand in JS is the residual. The runtime being retired is Node.js 24 on V8 13.6 [node24]. Rung SV8. 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 Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13Cons behind Node.js 24 (V8 13.6) (leads); not adopted yet: imported only by validation organs (gates, tests, benches): wire it into a shipping program	●	○	○	○	◉
Remote bytes and remote shell (ssh, scp) Measured: `v_commit` exists in `runtime/nx_content_put.nx`, verified at emit. The chunked, sha-receipted, atomic upload door replaced two same-day break-glass scp transfers; ssh remains the sanctioned recovery lane only. Oracle: OpenSSH with mlkem768x25519-sha256 by default [openssh]. Rung SV9. Adoption: `RUN-BY:cron` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Claude Code builtin tools, CPython 3.13, Node.js 24 (V8 13.6); fully adopted on the estate ladderCons behind GNU userland (bash, coreutils, grep, ssh, curl) (leads)	●	○	◉	○	○
Laptop host runner (the wsl launcher class) Open — watching `runtime/nx_laptop_runner.nx` : `sv_laptop_runner`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. WSL plumbing was 396 calls in the July window and is still how the memory-plane organs and the laptop build farm are reached; LAP01 host-runner is the kill rung SV10, with the never-brick fence from birth. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them dossierwatching sv_laptop_runner Pros none measured yetCons behind Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl); open contract; expected source declaration not established	○	●	●	○	○
Tool-call protocol boundary (MCP) Measured: `ta_mcp_call` exists in `runtime/nx_tools_api.nx`, verified at emit. The one dependency this board keeps by design: the assistant reaches every sovereign organ through tools/call [mcpspec], the sovereign server speaks it with capability tokens and also serves an OAuth 2.1 authorization server. Not a reliance to drive to zero -- the door. Adoption: `LIB-WIRED importers=26 nonval=11` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13, Node.js 24 (V8 13.6); fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	●	○	○	○
History horizon of the record Measured: `ch_bt_walk` exists in `runtime/_hdl_build/nx_claude_harvest.nx`, verified at emit. The sovereign journal begins 2026-07-20 (harvest start); the laptop retains transcripts from 2026-08-05; pre-July sessions were rotated before capture and no laptop backup holds them (measured 2026-09-05). Rung SV11 names the only recovery route, an off-box backup. Declared, never filled. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_claude_harvest_gate -- a mention, not a run. not adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_claude_harvest_gate -- a mention, not a run Pros Implementing source symbol found; runtime acceptance separate; ahead of Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13, Node.js 24 (V8 13.6)Cons not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_claude_harvest_gate -- a mention, not a run	●	○	○	○	○
Journaled reliance series on this board Source present; authored stronger claim: `ch_journal` in `runtime/_hdl_build/nx_claude_harvest.nx`, verified at emit. SV1 LANDED 2026-09-05: one log measure row per UTC day from the live journal, idempotent by a day marker, refuses -1, wired on the wsmine beat (row 115); no rival journals its own outside share onto its roadmap. The flip of this cell is the receipt of presence; the series rows on this page are the evidence. Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_claude_harvest_gate -- a mention, not a run. 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); only referenced by nx_claude_harvest_gate -- a mention, not a run Pros Source symbol found; stronger capability claim is authored; ahead of Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13, Node.js 24 (V8 13.6)Cons not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_claude_harvest_gate -- a mention, not a run	◉	○	○	○	○
Residual by class, whole partition, no frequency cap Measured: `ch_classes` exists in `runtime/_hdl_build/nx_claude_harvest.nx`, verified at emit. SV2: per-outside-tool counts over assistant actions only, partition summing to claude_outside, cap named; today the top-30 frequency list covers 80 percent (Bash, PowerShell, Edit, Read). Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_claude_harvest_gate -- a mention, not a run. 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); only referenced by nx_claude_harvest_gate -- a mention, not a run Pros Implementing source symbol found; runtime acceptance separate; ahead of Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13, Node.js 24 (V8 13.6)Cons not adopted yet: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked); only referenced by nx_claude_harvest_gate -- a mention, not a run	●	○	○	○	○
Head-to-head receipt: search Open — watching `runtime/nx_sov_h2h.nx` : `sv_h2h_search`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. SV4: nx_shelltool grep versus ripgrep versus the builtin Grep on one fixed corpus, identical match sets or VOID, wall time per arm, rendered here [ripgrep]. CAPABILITY-MISSING until the receipt exists. dossierwatching sv_h2h_search Pros none measured yetCons behind Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl) (leads), CPython 3.13, Node.js 24 (V8 13.6); open contract; expected source declaration not established	○	●	◉	●	●
Head-to-head receipt: the frontier seat Open — watching `runtime/nx_sov_h2h.nx` : `sv_h2h_llm`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. SV5: a fixed KAT battery through nx_engine_dispatch to local, external and frontier engines, scored per engine and graded against the repository-level bar [swebench]; the expected local loss is the build target. CAPABILITY-MISSING until the receipt exists. dossierwatching sv_h2h_llm Pros none measured yetCons behind Claude Code builtin tools (leads); open contract; expected source declaration not established	○	◉	○	○	○
Head-to-head receipt: fetch Open — watching `runtime/nx_sov_h2h.nx` : `sv_h2h_fetch`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. SV6: nx_research_fetch versus curl versus the builtin fetcher on one fixed URL set, body bytes identical or VOID, provenance only on the sovereign arm, a PDF quote check where the summariser fabricated. CAPABILITY-MISSING until the receipt exists. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them dossierwatching sv_h2h_fetch Pros none measured yetCons behind Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13, Node.js 24 (V8 13.6); open contract; expected source declaration not established	○	●	●	●	●
Python residual ratchet on the program path DARK — `runtime/nx_sovereignty_audit.nx` EXISTS but does not declare `sv_py_zero`: 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. SV7: every .py in the PRODUCT zone listed with its replacing organ, a count that may only fall, CPython kept as a comparator [cpython313]. dossierwatching sv_py_zero Pros none measured yetCons behind CPython 3.13 (leads); open contract; expected source declaration not established	○	○	○	◉	○
JavaScript residual ratchet on the page path Open — watching `runtime/nx_wasmpage_census.nx` : `sv_js_zero`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. SV8: every hand-JS page listed with the NishiLang source that lowers to its wasm twin, a count that may only fall, Node kept as a comparator [node24]. dossierwatching sv_js_zero Pros none measured yetCons behind Node.js 24 (V8 13.6) (leads); open contract; expected source declaration not established	○	○	○	○	◉
Remote-bytes residual ratchet DARK — `runtime/nx_content_put.nx` EXISTS but does not declare `sv_ssh_zero`: 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. SV9: ssh and scp counted in the journal as outside and driven to the recovery lane only; the field's post-quantum default is the bar the sovereign door must meet [openssh]. dossierwatching sv_ssh_zero Pros none measured yetCons behind GNU userland (bash, coreutils, grep, ssh, curl) (leads); open contract; expected source declaration not established	○	○	◉	○	○
Pre-July record recovery DARK — `runtime/_hdl_build/nx_claude_harvest.nx` EXISTS but does not declare `sv_history_pre_july`: 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. SV11: ingest any operator-named off-box backup of transcripts dated before 2026-07-20 through frames plus ingest, or declare the horizon permanent; the gauge prints the horizon beside every number. dossierwatching sv_history_pre_july Pros none measured yetCons open contract; expected source declaration not established	○	○	○	○	○
Editor source of record Open — watching `runtime/nx_fs_write.nx` : `sv_fswrite_source`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. SV12: nx_fs_write ships inside the tools daemon with no source under buildroot/runtime (find, corpus_complete=1); done when a named source exists and the editor row is repointed to it. rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them dossierwatching sv_fswrite_source Pros none measured yetCons behind Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13, Node.js 24 (V8 13.6); open contract; expected source declaration not established	○	●	●	●	●
Foreign data assets priced per board (weights, fonts, corpora, doors) Source present; authored stronger claim: `fas_load` in `runtime/nx_assets_lib.nx`, verified at emit. SV13 LANDED 2026-09-16: the one reader of <dom>.assets, rendered by the compare emitter on both generators with licence, bytes, used-by, replacement rung and a three-point lift priced through nx_costest_lib and converted to hours from nx_costcal's calibration of u against the boards' own journals; no rival prices its own third-party reliance with a replacement lift. First data on search.assets and llm.assets. Adoption: `LIB-WIRED importers=3 nonval=2` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of Claude Code builtin tools, GNU userland (bash, coreutils, grep, ssh, curl), CPython 3.13, Node.js 24 (V8 13.6); fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	○	○	○	○
Run-path foreign assets with no replacement rung, fleet ratchet DARK — `runtime/nx_sovereignty_audit.nx` EXISTS but does not declare `sv_assets_ratchet`: 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. SV13: counts NORUNG rows across every <dom>.assets file on a beat, may only fall, names the offenders; done at zero with the calibration usable. watching sv_assets_ratchet Pros none measured yetCons open contract; expected source declaration not established	○	○	○	○	○

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	20	rows carrying a Yes, Best or Part in a rival column
cited	11	a reference mark resolves to a pinned mirror
uncited	9	an observation read, badged until a reference lands
rows, summed	20	RECONCILED: the parts sum to the total 20

Delivery and evidence

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

Debt register

Id	Sev	What it is	Unblock
sovereignty-freq-index-overflownot in this scope	3	nx_actlog freq over the full 685,920-frame journal reports index_overflow=1: one of its fixed index tables (AL_IX_TVMAX 4096, AL_IX_GMAX 16384, AL_IX_ARENA 1 MiB) filled, so bigram mining over the whole journal is a floor; the top-30 frequency list is exact.	Size the index tables from the frame count instead of constants (the same law that removed the read cap), announce which table filled, and gate it on a fixture larger than every table.
sovereignty-ingest-behindnot in this scope	4	The chbeat capture spool holds 757,529 frames since 2026-07-21 while the journal holds 685,665; the ingest lane is behind by roughly 70k frames and the gauge's denominator is a floor until it catches up.	Run the ingest to the spool's high-water mark and add the spool-versus-journal delta to the journaled measure row so a lag is visible on this board.

Risk register

Risk	Likelihood x impact	Mitigation
A seat reads the legacy outside_permil (all frames) as the reliance number	probable x medium	the page prints claude_outside_permil as the headline and the legacy figure beside it with the beat count; the journal row carries both
The four largest outside classes hide the long tail below the frequency cap	probable x low	SV2 emits the whole partition summing to claude_outside; until then the board says the top four cover 80 percent and names the cap
A twin flips a watch cell green with no receipt behind it	possible x high	the flip is the receipt of presence only; every SV rung names its receipt or its neg-control and the rubric row says a class without a receipt is CAPABILITY-MISSING
The pre-July build of the estate is never recovered and the trajectory reads from a late start	possible x medium	SV11 declares the horizon on the row and the gauge prints it beside every number; a late start is stated, never hidden

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
2026-09-05	SV0	measure	the live /outsidegap read 592 permil over 79,634 frames with truncated=1: the 4 MB read cap had been sampling the July head of a 54 MB journal, and the journal mixes assistant calls with estate beat runs
2026-09-05	SV0	land	nx_claude_harvest gauge fixed: sys_read_file whole-journal read, verb=run beats partitioned out, new keys beats/claude_actions/claude_outside/claude_outside_permil, three teeth T21-T23 incl. non-vacuity; local WSL selftest 43 of 44 with the one RED (T14) pre-existing and environmental, proven by building the banked original (40 of 41, same tooth)
2026-09-05	SV0	lesson	a build refused by admission is not lost: the NAS was in an I/O storm (procs_blocked 8 of 8) and the build was QUEUED (BQ row, rc 0); forcing a compile into the storm would have been the resource bug the shipping criterion names, so the queue fired it ten minutes later
2026-09-05	SV0	land	nx_claude_harvest PROMOTED a4f43b12 through nx_organ_ship: content GREEN (no run lost), offc twin refreshed, registered; prove read NO-GATE-FOUND because the proof lived only in a verb -- the gate rung was declared in the same hour
2026-09-05	SV0	measure	THE HONEST NUMBER over the whole journal (54,291,072 B, truncated 0): frames_total 685665, beats 274104, claude_actions 411561, claude_outside 47230, claude_outside_permil 114; legacy all-frames 166; banked as minergap MG13
2026-09-05	SV0	measure	the promoted binary's own selftest on the NAS: 44 of 44 verdict GREEN (T14 green there as its comment predicted); nx_actlog sibling cap fixed and promoted e87b8148, its selftest 15 of 15 GREEN; nx_actlog freq over 685,920 frames names the residual: Bash 12584, PowerShell 11588, Edit 9398, Read 4348, index_overflow 1 declared
2026-09-05	SV0	land	nx_claude_harvest_gate written and landed (4321 B): forks the just-built or promoted subject's selftest, exit code is the verdict, T4 pins artifact identity on the partition tooth; build queued under CPU saturation (procs_running 10 of 8)
2026-09-05	SV4	measure	six Sept-2026 field references mirrored over the sovereign TLS client and pinned by provenance sha256 (ripgrep, SWE-bench, CPython 3.13, Node 24, OpenSSH, MCP 2025-06-18); the specific claims cited were witnessed by grep in each mirror before being written
2026-09-05	SV0	lesson	a self-audit for half wiring found four gaps in the same session that shipped the fix: the proof asserted from a different build than the promoted one, a possible stale third twin, the sibling miner's identical cap, and no gate for the ship loop to resolve -- each was closed by a measurement, not a sentence
2026-09-05	SV1	measure	[HARV:journal:day=20701] claude_outside_permil=114 claude_actions=412416 claude_outside=47230 estate_beats_excluded=274220 frames_total=686636 journal_bytes=54374219 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-05	SV1	land	ch_journal SHIPPED: nx_claude_harvest journal verb PROMOTED 699388fb through nx_organ_ship with prove=GREEN via nx_claude_harvest_gate (resolved by the target_gate convention, bank CURRENT) and the ONE named lost run (the old usage line, reworded to add the verb) journaled as ADOPT-LOST under allow_loss=1; first live run on the NAS appended the measure row above from 686,636 frames; wired as plan-wsmine- row 115 so the hourly beat journals with zero seat tokens; local teeth J1-J3 PASS (46 of 47, T14 environmental)
2026-09-05	SV0	land	THE EMITTER GREW (operator, on this page: rows read as random glyphs and a partial adoption printed as a badge). nx_swcompare_matrix now renders a derived Pros / Cons strip per row from the same cells (has / leads / in part; ahead of whom; behind whom, leads signed; open contract; no organ measured; not adopted yet with the catalog lib's own remedy; and, when nothing is measured against Nishi, none on the measured axes -- never a victory claim), and the PARTIAL adoption chip now carries the remedy text as a CON with the raw ladder token only on hover. Fleet-wide: every domain inherits it on the next beat. PROMOTED 80d97b58 through nx_organ_ship with prove GREEN on BOTH declared gates (nx_imgattr_lib_gate 42 of 42, nx_swcompare_shape_gate 11 of 11 on the NAS), content GREEN, offc fresh. The new gate is named for the PROPERTY and bound by an organ_gate.conf row; it caught two of my own mistakes before shipping: a tooth that assumed a fixture organ would read adopted (a census-less host reads it unadopted -- the tooth was wrong, not the emitter) and an exec of the copied subject from /tmp, which is noexec on the NAS (gen_exit 127) -- fixed by forking the subject by absolute path with only the CWD inside the fixture
2026-09-05	SV0	measure	EMITTER VERIFIED ON THE LIVE SURFACE: regen job 1788628509 published sovereignty index.html at 93647 B (prev 84660, the delta is the per-row Pros and Cons strips), nx_page_verify status=200 a11y-issues=0 VERDICT=GREEN through the pinned sovereign edge. The regen's own verdict was RED for three other domains (observability present<5, intelmine a stray bar in a note), not this one. Two 404s from nx_https_get on two pages were ONE vantage defect (the NAS hairpins nishifamily.com to the nginx default vhost; pass 127.0.0.1:8443 or use nx_page_verify), not two missing pages
2026-09-05	SV3	land	tg_shell_interlock landed on the shared contract with engineshift ES22 (that ledger carries the measurements): a PreToolUse organ on every harness Bash and PowerShell call naming the estate verb from hooks/shellverb.conf or GAP, gate 32 of 32, laptop-installed and wired; this board's watch row had named a source path that does not exist (runtime/nx_tokgate.nx) and is now runtime/_hdl_build/nx_tokgate.nx so the cell can flip on the next emit; the done-rule half that remains is the journaled series: the Bash plus PowerShell share of assistant actions must fall, measured by nx_actlog over the coming sessions
2026-09-05	SV2	queue	IN FLIGHT (crash ledger): (1) SV2 classes verb source LANDED on the NAS byte-identical to the laptop (nx_claude_harvest.nx 145114 B sha256 0b036e99d3e3f530..., five size-CAS edits, definitions first; laptop SOVEREIGN build, selftest 49/50 with the one red = T14 environmental, C1-C3 GREEN) -- ship pending on the /api/build lease via nx_organ_ship (PROVE = nx_claude_harvest_gate). (2) nx_actlog_gate (end-to-end gate for the action miner, closes the prove=NO-GATE-FOUND hole) landed 4373 B, BUILT on the NAS by the BQ queue (planrun-build row rc=0, 30951 B) -- promote+register pending (re-run nx_organ_ship nx_actlog_gate, build-cache HIT). (3) nx_restage rewritten PROMOTED-FIRST (copies ../<target>.elf into buildroot/_offc when a promoted twin exists, announces source, rebuild is the fallback) 8731 B under content-CAS; nx_organ_ship nx_restage job 1788629928 running. Recipe if this seat dies: nx_catalog the three names, then nx_organ_ship each; journal a land row per promote
2026-09-05	SV0	land	nx_restage PROMOTED-FIRST SHIPPED (sha 06a3818a88ef8683, 29184 B, live == staged, offc twin installed, registered): when the promoted serving-root twin ../<target>.elf exists it is COPIED into buildroot/_offc and the receipt says source PROMOTED; the sovereign-lane rebuild is now the fallback for a never-promoted target and the receipt says source BUILT. Reason: nx_compare_regen's fork-freshness hashes _offc against the promoted twin and names nx_restage as the remedy, but a rebuild through the sovereign lane is byte-different from the /api/build lane on identical source (measured 2026-09-05 on nx_swcompare_matrix: 332217 B rebuilt vs 368016 B promoted), so the old remedy could never read FRESH. Adopt-gate lost 2 of 52 runs (38 permil), both NAMED rewords (the /tmp error text and the JSON literal split to carry source), allowed explicitly. Caveat carried from the ship receipt: nx_restage_gate (8 of 8) ran from an artifact built BEFORE this edit because its rebuild was refused by admission and queued (BQ-nx_restage_gate); that gate proves the target sanitizer r_san, which this change does not touch
2026-09-05	SV2	land	SV2 SHIPPED: nx_claude_harvest promoted 6f5c4bf6459034567a461e59fc104107ed78419a8fc73c119aff721e60f445f6 through nx_organ_ship, PROVE nx_claude_harvest_gate 6 of 6 GREEN (closure bank CURRENT), adopt-gate 1 named lost run of 696 (the usage line reworded to add the verb), offc twin byte-identical. The classes verb answers the residual BY TOOL over claude-action frames with estate beats excluded, the top 12 plus an other bucket so the partition sums to claude_outside, a 512-name table whose overflow is COUNTED, and the daily journal row now carries outside_by_tool=Tool:N,...,other:K for the top 6. The contract symbol ch_classes is DECLARED (nx_symdecl_lib applies the declaration rule to .nx organs and never falls back to a substring, so ch_classes_json alone could not flip the cell) and CALLED by the verb. Teeth C1 one-tool journal yields one class equal to claude_outside with other 0, C2 beats and nishi tools never a class, C3 sorted descending with an exact sum Bash:3,Edit:2,Grep:1,other:0 of 6 -- the plan's pre-declared done-rule as written. NAS source == laptop by sha256 5c27b81111f3a802 after five size-CAS edits
2026-09-05	SV0	lesson	A SHIP LOOP THAT SENDS AN IDENTICAL BUILD BODY TWICE IN AN HOUR PROMOTES THE PREVIOUS GENERATION WITH PROVE GREEN. Measured 2026-09-05 on nx_actlog: live_sha equal to the prior promote, promoted selftest still 15 of 15 against a 65611 B source that adds T16 and T17, nx_actlog_gate 6 of 6 because its identity tooth (the confidence string) exists in both generations. Two root fixes shipped the same hour, not a memory note: nx_organ_ship now appends a per-run bust nonce to both /api/build bodies (debt 1785363417 recurrence, open since 2026-07-29), and nx_actlog_gate T5 pins the derived-index tooth that exists only in the new generation. Law: a gate whose identity tooth pins a string both generations carry certifies the predecessor
2026-09-05	SV0	retract	RETRACT of the SV0 lesson row above, same day: the stale nx_actlog promote was not an idempotency replay (no client sends the header the mgmt dedup keys on; proven by a coverage-complete grep). It was nx_hostctl cmd_buildrun staging the PREVIOUS artifact after the builder refused under its own admission check -- an exit code printed and then ignored, an artifact path shared between a refused run and a successful one. Root fix in source: the builder's exit 0 plus its own success token gate the stage. The law survives with its subject corrected: a receipt that says BUILT for a run that did not build is the false-green class, and the gate whose identity tooth pins a string both generations carry cannot catch it
2026-09-06	SV1	measure	[HARV:journal:day=20702] claude_outside_permil=111 claude_actions=425058 claude_outside=47231 estate_beats_excluded=276593 frames_total=701651 journal_bytes=55646978 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4058 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-07	SV1	measure	[HARV:journal:day=20703] claude_outside_permil=104 claude_actions=451696 claude_outside=47234 estate_beats_excluded=284047 frames_total=735743 journal_bytes=58522901 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4061 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-07	SV4	queue	OPERATOR SCOPE 2026-09-07: all operational work is sovereign capability investment, including grep/search/read/write/build/test/site emission/publication/recovery/AI. Existing goal and rung contracts expanded; this row records requested scope only, not implementation, build, publication or functional acceptance. Next: measure the verified canonical owners and their blockers, attach comparable dated benchmark receipts, and render the changes on the existing /compare boards.
2026-09-08	SV1	measure	[HARV:journal:day=20704] claude_outside_permil=102 claude_actions=459313 claude_outside=47235 estate_beats_excluded=288011 frames_total=747324 journal_bytes=59451455 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4062 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-09	SV1	measure	[HARV:journal:day=20705] claude_outside_permil=100 claude_actions=469102 claude_outside=47236 estate_beats_excluded=293927 frames_total=763029 journal_bytes=60699417 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4063 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-10	SV1	measure	[HARV:journal:day=20706] claude_outside_permil=99 claude_actions=474723 claude_outside=47237 estate_beats_excluded=303386 frames_total=778109 journal_bytes=61915197 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4064 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-10	SV11	measure	2026-09-10 bounded coverage audit: existing native harvester source runtime/_hdl_build/nx_claude_harvest.nx found171680B with frames/delta/learn/beat/journal interfaces; main runtime path was absent, so discovery must resolve actual owner. Existing nx_memory MCP default attachment denied; same installed tool succeeded immediately through existing scoped MCP connection, listing2924 store lines. Local main Claude project167 top-level JSONL files observed, not proof of complete historical access or mined requirements. NAS knowledge path-name search for transcript returned0 of15197 files; that scope does not imply no transcript archive. Existing action journal/learn owner must be reused for requirement-to-board mapping; EC2 carries audit status, exact original-request coverage remains unverified. No private transcript or legal source sent to Claude.
2026-09-11	SV1	measure	[HARV:journal:day=20707] claude_outside_permil=84 claude_actions=559045 claude_outside=47237 estate_beats_excluded=311631 frames_total=870676 journal_bytes=85514890 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4064 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-12	SV1	measure	[HARV:journal:day=20708] claude_outside_permil=80 claude_actions=584987 claude_outside=47237 estate_beats_excluded=311631 frames_total=896618 journal_bytes=92719616 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4064 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-13	SV1	measure	[HARV:journal:day=20709] claude_outside_permil=74 claude_actions=637955 claude_outside=47237 estate_beats_excluded=311631 frames_total=949586 journal_bytes=107450195 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4064 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-14	SV1	measure	[HARV:journal:day=20710] claude_outside_permil=63 claude_actions=743879 claude_outside=47282 estate_beats_excluded=311718 frames_total=1055597 journal_bytes=137109959 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4109 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-15	SV1	measure	[HARV:journal:day=20711] claude_outside_permil=61 claude_actions=767176 claude_outside=47283 estate_beats_excluded=320936 frames_total=1088112 journal_bytes=144297871 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4110 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-16	SV1	measure	[HARV:journal:day=20712] claude_outside_permil=60 claude_actions=787979 claude_outside=47284 estate_beats_excluded=329538 frames_total=1117517 journal_bytes=150829521 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4111 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-16	SV13	land	treat row declared for SV13 (mcp nx_assets_gate GREEN on the allowlist, api /api/tools/call name=nx_assets_gate, prim nx_assets_lib in the tree, wf costcal = the plan_run plane plan-costcal- whose step 2 runs the gate, agent costcal = the daily clock row that runs the plan): nx_compare_rank sovereignty reads TREATMENT rows=1 complete=1 verdict=TREATED against allowlist 165424 B, roster 69230 B, clock plane 11935 B; the emitter renders the Treatment verdict and the three-step table on this page from the same ruler (ecosystem EC49)
2026-09-16	SV13	land	THE FOREIGN-ASSET LEDGER LANDED AS EMITTER CAPABILITY, NOT A SIDE PAGE (operator 2026-09-16: know via /compare what the lift would be to replace a third party we use with our own, in cost): nx_assets_lib (row grammar, states runpath candidate oracle unavailable, kinds, first-failing-rule verdicts, partitions that sum, three-point aggregates) and nx_costest_lib (deci-u three-point estimates, root-sum-square band beside the correlated sum, quantile rank rule with a derived distinct-rank floor, the 6ND and 2ND compute rules, the calibration census and the one ledger grammar) with gates nx_assets_gate 36/36 and nx_costest_gate 63/63 on the laptop lane; assets_pass in nx_swcompare_lib renders the table and the api.json object on both generators; nx_costcal writes knowledge/status/costcal.ledger from every plan's rung costs and land rows (more than a thousand land rows on the boards); first data: search.assets (11 rows, the MiniLM weights as the only runpath foreign weights, five open candidates, two unavailable rank-1 artifacts, two oracles, the GPU door) and llm.assets (the Qwen2.5 GGUFs, replaced by LM16)
2026-09-16	SV10	measure	A WITNESS FOR THE LAPTOP HOST RUNNER: a five-target stage upload written as a harness shell loop failed all five (the loop variable's dotted name parsed as a property access, the door then read the url as its cap file); five explicit calls succeeded; the durable fix is this rung's organ, and every harness shell loop until then is the class this rung retires
2026-09-17	SV1	measure	[HARV:journal:day=20713] claude_outside_permil=57 claude_actions=816685 claude_outside=47285 estate_beats_excluded=338506 frames_total=1155191 journal_bytes=159651035 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4112 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-18	SV1	measure	[HARV:journal:day=20714] claude_outside_permil=55 claude_actions=852090 claude_outside=47286 estate_beats_excluded=347908 frames_total=1199998 journal_bytes=170350529 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4113 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-19	SV1	measure	[HARV:journal:day=20715] claude_outside_permil=54 claude_actions=872379 claude_outside=47287 estate_beats_excluded=357331 frames_total=1229710 journal_bytes=176745546 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4114 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-20	SV1	measure	[HARV:journal:day=20716] claude_outside_permil=51 claude_actions=915877 claude_outside=47287 estate_beats_excluded=366261 frames_total=1282138 journal_bytes=189830207 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4114 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-21	SV1	measure	[HARV:journal:day=20717] claude_outside_permil=48 claude_actions=970624 claude_outside=47288 estate_beats_excluded=374616 frames_total=1345240 journal_bytes=205925134 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4115 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)
2026-09-22	SV1	measure	[HARV:journal:day=20718] claude_outside_permil=45 claude_actions=1039963 claude_outside=47289 estate_beats_excluded=383181 frames_total=1423144 journal_bytes=226052090 outside_by_tool=Bash:12584,PowerShell:11588,Edit:9398,Read:4348,Write:3585,Grep:1670,other:4116 (nx_claude_harvest journal: full read, estate beats excluded, one row per UTC day)

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 Sept-2026 field, from the mirrored record: the search field's bar is a line-oriented regex search over a directory tree [ripgrep]; the code-agent field grades itself on resolving real repository issues [swebench]; the scripting runtime the estate still leans on shipped a free-threaded build and a JIT [cpython313]; the page runtime the estate still leans on ships on V8 13.6 [node24]; the remote-shell field made a post-quantum hybrid key exchange its default [openssh]; and the one boundary this board keeps by design is the tool-call protocol through which the assistant reaches every sovereign organ [mcpspec]. What none of them measures is its own outside share: the estate's miner reads the whole 54 MB action journal (685,665 frames, no cap) and, with the estate's 274,104 autonomous beat runs excluded, finds 47,230 of 411,561 assistant actions outside the estate since 2026-07-20 -- 114 permil, down from the July baseline of 651. The residual is named per class below; the largest is the shell (Bash 12,584 and PowerShell 11,588 actions), then the editor (9,398) and the reader (4,348). The frontier model itself is the headline gap: the sovereign dispatcher ladders local, external and frontier engines by declared class, but the local seat's ceiling is 96 new tokens by data, so code authoring is frontier-only today and the head-to-head receipt against the assistant is the build target, not a claim.

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

Every compare carries this layer. Declare knowledge/compare/sovereignty.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 sovereignty, 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 sovereignty.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	2
scanned	2
fetched	2
reused	0
failed	0
named	2
candidates	2
mentions	2
capped	0

Rank	Rival	Seeds	Mentions	First seed	Kind	Link
1	GNU userland (bash, coreutils, grep, ssh, curl)	1	1	col1-gnu	named	https://github.com/BurntSushi/ripgrep
2	Node.js 24 (V8 13.6)	1	1	col3-node	named	https://nodejs.org/en/blog/release/v24.0.0

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 28 capabilities: 23 exclusive to it and 5 shared with other boards. Shared rows on this page: 5 = home here 0 + home elsewhere 2 + ambiguous 1 + undecided 2 + refused 0 (partition 5); undecided = unmapped 1 + unproven 1 + unevidenced 0; homonyms 1; since the previous census: new 0, resolved 0. Census FRESH age_s=7167.

Capability	Organ	Boards grading it	Home	Rule	Evidence	Standing on this board
wat_emit_module	`-`	lang:runtime/nx_wasm.nx, gameengine:runtime/nx_wasm.nx, sovereignty:runtime/wasm.nx	—	homonym	-	homonym
nx_pdf_extract_text	`nx_pdf_text.nx`	mangagen, sovereignty	—	none	unproven	undecided
v_commit	`nx_content_put.nx`	sharehub, sovereignty, dataio+, modding	dataio	gates	gates	home elsewhere adopt: `@xrow dataio\|v_commit\|home by gates (fleet duplicate census)`
rf_provenance	`nx_research_fetch.nx`	extllm, sovereignty	—	none	unmapped	undecided
ed_dispatch	`nx_engine_dispatch.nx`	engineshift*, codeeffectiveness, sovereignty	engineshift	rung	plan	home elsewhere adopt: `@xrow engineshift\|ed_dispatch\|home by rung (fleet duplicate census)`
tg_shell_interlock	`nx_tokgate.nx`	engineshift, sovereignty	—	ambiguous	plan	ambiguous

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 1396272 s), sources read 4 of 7 declared — a BLIND census is a FLOOR: unread sources can only add rows.

kind	name	board	source	evidence
directive	`nx_claude_harvest_gate`	sovereignty	plan-queue	rung=SV2

rows shown 1|this board's directives 1|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.

[ripgrep] BurntSushi. ripgrep: a line-oriented search tool that recursively searches a directory for a regex pattern; the field's fastest widely deployed grep and the oracle for the search rung (README read 2026-09-05, the phrase witnessed in the mirror). publisher · read in our library knowledge/fetched/cmp_sovereignty_ripgrep.html · pin h8d212ccbdbd36439cbbd73a34ee30d60ae8bfd4fbc7b8b4e131ab12f55216ac2 · accessed 2026-09-05 · source-readGrounds: Search and grep
[swebench] SWE-bench. Leaderboards (Verified, Lite, Full): the repository-level bar for code-writing agents, the field's standard for whether a model resolves real GitHub issues; the oracle the LLM code rung must be graded against (site read 2026-09-05). publisher · read in our library knowledge/fetched/cmp_sovereignty_swebench.html · pin h5d031921be2267f0d210fde3c654d56e9be7fc5ee45b3bab3b9f5888cce73d5a · accessed 2026-09-05 · datasetGrounds: LLM reasoning and code generation
[cpython313] Python Software Foundation. What's New In Python 3.13: the release that introduced the experimental free-threaded build and the experimental JIT; the runtime the estate's analysis scripts and oracle generators still lean on outside the sovereign path (docs read 2026-09-05). publisher · read in our library knowledge/fetched/cmp_sovereignty_cpython313.html · pin ha2d5c7f6c38942635861cacc84c66489b28187f4bc255fd7cfdcdb9ac775cdc6 · accessed 2026-09-05 · vendor-docGrounds: Python runtime
[node24] OpenJS Foundation. Node.js v24.0.0 release announcement: the JavaScript runtime on the V8 engine that page tooling and JS oracles run on; the twin is the estate's wasm lane compiled from the same NishiLang source (release post read 2026-09-05). publisher · read in our library knowledge/fetched/cmp_sovereignty_node24.html · pin h0f633f510a060c6b70da617754a8fc74f88972e7e5cd39d27a5e2e8c3323032e · accessed 2026-09-05 · vendor-docGrounds: JavaScript runtime
[openssh] OpenSSH project. OpenSSH 10.0 release notes: the transport the estate's byte moves and remote control leaned on before the sovereign content-put door, and the release in which the post-quantum hybrid key exchange mlkem768x25519-sha256 became the default (text read 2026-09-05, the algorithm name witnessed in the mirror). publisher · read in our library knowledge/fetched/cmp_sovereignty_openssh100.txt · pin h483c8097e5f0760d6d41bcf7e0f2dbf1942ab8eb136ea715623658d07ffde872 · accessed 2026-09-05 · vendor-docGrounds: Remote bytes and shell
[mcpspec] Model Context Protocol. Specification 2025-06-18: the tool-call protocol through which the assistant reaches the estate's sovereign organs, the interop boundary every sovereign action crosses and the one dependency this board deliberately keeps (spec read 2026-09-05). publisher · read in our library knowledge/fetched/cmp_sovereignty_mcpspec.html · pin h2af38d2c8f9982afe0b11df4aaeb286b968151279d3183ccb3130a79bda0c325 · accessed 2026-09-05 · published-standardGrounds: Agent and tool protocol

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