Nishi Family › Compare › Data Engineering and Analysis Agents

Nishi Compare · capabilities, evidence and delivery

Data Engineering and Analysis Agents

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

Nishi data organs vs Spider-Agent and Agentar-Scale-SQL and DA-Agent and human data engineers on the field's benchmarks

Overview

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

Where we are. MEASURED 2026-09-16 at admission: the estate reads SQLite (nx_sqlite_rows), Parquet (nx_parquet_lib, the sovereign reader that reproduces the BRIGHT card), CSV and TSV (nx_value_parse_csv) and XLSX (nx_xlsx, which read the 1077-row OSWorld results sheet for the automation board the same day); it profiles a table (nx_analyst_data) and approximates distincts and quantiles with a measured error (nx_dataframe_approx); it grades code by a byte-exact compile-and-run judge (nx_forge_local_bench) and it can replay a rival's published outputs through its own judge (nx_bright_reasoned_import, landed on the search board today). It writes NO SQL and NO data-science code from language yet: every text-to-SQL, warehouse, dbt, GUI-workflow and agentic-analysis row is an honest _ABSENT_ contract. The field, read from the mirrored bytes: BIRD test execution accuracy 82.39 by an agent system dated 2026-09-07 against a human 72.37 [bird26]; Spider 2.0-Snow 96.70, DBT 65.6 and Lite 76.23 at the March-2026 listing [spider226]; DS-1000 held at 43.3 by Codex-002 on its own page [ds100022]; DA-Code, DSBench and DABstep grade agents end to end [dacode24] [dsbench25] [dabstep26]; LiveSQLBench rotates tasks against contamination [livesqlbench26]; Spider2-V moves the whole workflow into the tools' GUIs [spider2v25].

Where we need to go. A sovereign data agent that writes and executes SQL and analysis code against the estate's own stores and the field's benchmarks, graded by an execution judge the estate owns, with every published number carrying its harness, its n and its date, and no third party in the run path: the benchmarks are oracles outside, their tasks are data with licences beside them, and their leaders are the bars.

The unit. 1 u = one measured session-leg (organ plus gate plus a re-measured row), the estate calibration; nx_costcal reads 1 u as 0.12 h of clock, 1.45 h of seat and 19.44 h of wall across 213 landed rungs.

Cost to own the judge and the first two harnesses: 4 u. DE1 the execution judge over SQLite, DE2 the BIRD Mini-Dev harness and DE4 the Spider 2.0-Lite SQLite subset: after these the board carries a floor the estate computed itself.

Cost to field a contender: 9 u more. DE3 text-to-SQL through the engine dispatcher, DE5 the warehouse dialect bridge and DE8 dbt projects; the number that lands on the board is the judge's, never a read leaderboard cell.

Cost to reach the agentic frontier: 9 u more. DE6 the DS-1000 replay through an isolated oracle host, DE7 DABstep and DA-Code runs through the workflow engine and DE9 the GUI workflows through the sovereign browser. Program total 22 u and the three figures sum.

Latest recorded release

2026-09-16 · DE1

BOARD ADMITTED 2026-09-16 from eight sources fetched over the sovereign fetcher and read from the mirrored bytes on both hosts (byte-identical): Spider 2.0, Spider2-V, BIRD, DS-1000, DABstep, LiveSQLBench, DSBench and DA-Code; the hand page that held this domain since 2026-08-13 leaves hand.roster by matrix admission; the OSWorld results sheet read the same day through nx_xlsx is the automation board's, cited there

Release history and work log

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

5 of 13 capability source symbols found|2 of them authored stronger-claims|8 open|coverage 384/1000|adoption 3 full / 2 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 dataeng 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: STALE_OR_INVALID_BINDING. Recorded priorities require current acceptance evidence and resource checks before execution.

Recorded ranking withheld

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

Declared roadmap — contract, acceptance, executor, effort

Ladder to SOTA

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

best-in-class (T-BIC-BIRD), dated by SB-BIRD = 2026-09 ATTESTED. An agent system that writes SQL against a SQLite schema and clears the BIRD human test bar of 72.37 execution accuracy, the way the September-2026 leaders do at 82.95 (GrainSQL, dated 2026-09-26) and 82.39 ref bird26

frontier (T-FR-SPIDER2), dated by SB-SPIDER2 = 2026-07 ATTESTED. Enterprise workflow text-to-SQL over Snowflake and BigQuery plus dbt code tasks, the Spider 2.0 tracks whose leaders (dated March, May and July 2026) read 96.70, 65.6 and 76.23 ref spider226

frontier (T-FR-DATASCI), dated by SB-DABSTEP = 2026-09 ATTESTED. An agent that plans, writes and executes multi-step data analysis over real files and is graded end to end, the DABstep, DA-Code and DSBench class ref dabstep26

Rung	Closes with	Definition of done (pre-declared)	Executor	Est.
Execution judge for SQL over SQLite (DE1)	sqr_exec_judge	DONE-RULE PRE-DECLARED: the judge runs a gold query and a candidate query against the same SQLite file through nx_sqlite_rows, compares the result multisets order-insensitively, and prints EX 1 or 0 per pair with the row counts; a gate plants a candidate that returns the right rows in another order (must read 1), a candidate with one extra row (must read 0) and a candidate that errors (must read 0 with the error named); no rival oracle in the path [bird26]	Organ	1 u
BIRD Mini-Dev harness (DE2) after DE1	dc_bird_minidev	the Mini-Dev 500-pair set is mirrored from Hugging Face with its licence beside the bytes and the schema files, prepped into queries and gold TSVs the way nx_bright_prep shapes BRIGHT, and every candidate file is scored by DE1 into a receipt with n, EX and the harness hash; ACCEPT RULE: the gold set scored against itself reads EX 500 of 500 before any model is scored [bird26]	Organ	1.5 u
Sovereign text-to-SQL contender (DE3) after DE2	sqr_text2sql	a query writer over nx_engine_dispatch (local no-float seat, external U tier, frontier) that emits SQL for a natural-language question and a schema, scored only by DE1 on DE2; the receipt names the engine and its ceiling; the local seat's expected loss under its 96-token ceiling is published, never hidden; done when a receipt exists for all three engines	Organ	3 u
Spider 2.0-Lite SQLite subset harness (DE4) after DE1	dc_spider2_lite	the 135 SQLite examples of Spider 2.0-Lite (547 in total; the BigQuery and Snowflake examples wait for DE5) mirrored with licence and scored by DE1; the receipt states the subset size beside the number so a 135-example score is never read as the 547 leaderboard's [spider226]	Organ	1.5 u
Warehouse dialect bridge (DE5) after DE4	dc_dialect_bridge	BigQuery and Snowflake SQL run against the vendors' free quotas as ORACLE HOSTS outside the build path, results returned to DE1 for the multiset compare; the bridge is data (dialect rows), the credentials are vault leases, and a refused connection is recorded UNREACHABLE and never scored [spider226]	Organ	3 u
DS-1000 referee replay through an isolated Python oracle host (DE6)	fg_dscode	DS-1000 prompts and tests mirrored with licence; candidate completions from the engine dispatcher are executed on an isolated oracle host (Python is a third party and stays outside the estate's run path) and graded by the benchmark's own test cases, with the receipt carrying the host hash, n and the per-library breakdown [ds100022]	Organ	2 u
Agentic analysis runs, DABstep and DA-Code (DE7) after DE6	ad_agent_run	the workflow engine drives a plan, execute and observe loop whose tools are the estate's readers, profiler and approximate analytics, against the DABstep and DA-Code tasks mirrored with licence, graded by each benchmark's own answer checker on the oracle host; every run publishes completion rate and executable-code rate the way DA-Code's board does [dabstep26] [dacode24] [dsbench25]	Organ	3 u
dbt project tasks, Spider 2.0-DBT (DE8) after DE5	dc_dbt_project	the 68 DBT examples run through the workflow engine with the dialect bridge, graded by the benchmark's own result checks on the oracle host [spider226]	Organ	3 u
GUI-driven data workflows, Spider2-V (DE9) after DE7	br_data_workflow	the sovereign browser drives the tools' GUIs for the 494 Spider2-V tasks on an isolated oracle host, graded by the benchmark's own checkers; the GPT-4o baseline is the published floor and the estate publishes its own success rate beside it [spider2v25]	Organ	4 u

Rung	Role	Toward	Stands on	Why
DE1	substrate	T-BIC-BIRD	-	the judge every SQL number on this board is computed by
DE2	arm	T-BIC-BIRD	DE1	the harness that puts the estate's number beside the BIRD leaders on the same 500 dev pairs
DE3	contender	T-BIC-BIRD	DE2	its done-rule is a number against the human test bar
DE4	arm	T-FR-SPIDER2	DE1	the SQLite third of Spider 2.0-Lite, scored by the same judge
DE5	substrate	T-FR-SPIDER2	DE4	the dialect bridge the Snow and BigQuery tracks stand on
DE6	arm	T-FR-DATASCI	-	the DS-1000 replay sizes the data-science code lever before any agent is built
DE7	contender	T-FR-DATASCI	DE6	its done-rule is a completion rate against the agentic boards
DE8	contender	T-FR-SPIDER2	DE5	its done-rule is a number on the DBT track
DE9	arm	T-FR-SPIDER2	DE7	the GUI workflow leg, an arm until the browser lane can drive the tools

Milestones

Milestone	Rungs	Cumulative
M1 · The judge and the first harnesses	DE1,DE2,DE4	4 u
M2 · A contender on the board	DE3,DE5,DE8	9 u
M3 · The agentic frontier	DE6,DE7,DE9	9 u

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

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

Ladder verdict, counted
Class	Count	Note
targets	3	sotatarget rows on the plan
best in class	1	targets whose class is best-in-class, by code
frontier	2	targets whose class is frontier, by code
undated	0	targets naming no bar that exists
bad class	0	targets whose class is neither word
duplicate	0	target ids an earlier row already carries
rungs	9	rung rows on the plan
placed	9	rungs carrying a rungrole row
unplaced	0	rungs with no role toward any target
bad role	0	rungrole rows naming a role the ladder does not know
bad target	0	rungrole rows naming a target no sotatarget row declares
rungs, summed	9	RECONCILED: the parts sum to the total 9
substrate	2	what the target's own methods consume
arm	4	a sovereign arm that triangulates without competing for the number
contender	3	its done-rule is a number against the target
superseded	0	kept as capability, retired as a claim
placed, summed	9	RECONCILED: the parts sum to the total 9

Bar freshness, counted
Class	Count	Note
bars	8	sotabar rows on the plan
fresh	0	the bar month is the current month
attested	8	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	8	RECONCILED: the parts sum to the total 8
current month	2026-10	the month every bar is judged against

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

Inspect a rung and its prerequisites

Declared nodes 9. Rank input binding: STALE_OR_INVALID_BINDING. Dependency order is authored. Implementation, acceptance evidence, authority and resource readiness are unverified. No action is recommended or dispatched here.

Plan SHA-256 09a093a9e9f4a5eee214b0de5aa5f4f11e16bd6900ea82da65968daaed451497. Target rows 3; role rows 9. Existing risks and release worklog retain their own scope; no node completion is inferred.

Runtime observation coverage: 0 of 9 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.

DE1 — Execution judge for SQL over SQLite
Prerequisites: None declared; this does not establish execution eligibility.
Contract: sqr_exec_judge
Acceptance: DONE-RULE PRE-DECLARED: the judge runs a gold query and a candidate query against the same SQLite file through nx_sqlite_rows, compares the result multisets order-insensitively, and prints EX 1 or 0 per pair with the row counts; a gate plants a candidate that returns the right rows in another order (must read 1), a candidate with one extra row (must read 0) and a candidate that errors (must read 0 with the error named); no rival oracle in the path [@bird26]
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.
DE2 — BIRD Mini-Dev harness
Prerequisites: DE1 (acceptance unverified)
Contract: dc_bird_minidev
Acceptance: the Mini-Dev 500-pair set is mirrored from Hugging Face with its licence beside the bytes and the schema files, prepped into queries and gold TSVs the way nx_bright_prep shapes BRIGHT, and every candidate file is scored by DE1 into a receipt with n, EX and the harness hash; ACCEPT RULE: the gold set scored against itself reads EX 500 of 500 before any model is scored [@bird26]
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.
DE3 — Sovereign text-to-SQL contender
Prerequisites: DE2 (acceptance unverified)
Contract: sqr_text2sql
Acceptance: a query writer over nx_engine_dispatch (local no-float seat, external U tier, frontier) that emits SQL for a natural-language question and a schema, scored only by DE1 on DE2; the receipt names the engine and its ceiling; the local seat's expected loss under its 96-token ceiling is published, never hidden; done when a receipt exists for all three engines
Authored effort: 3. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
DE4 — Spider 2.0-Lite SQLite subset harness
Prerequisites: DE1 (acceptance unverified)
Contract: dc_spider2_lite
Acceptance: the 135 SQLite examples of Spider 2.0-Lite (547 in total; the BigQuery and Snowflake examples wait for DE5) mirrored with licence and scored by DE1; the receipt states the subset size beside the number so a 135-example score is never read as the 547 leaderboard's [@spider226]
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.
DE5 — Warehouse dialect bridge
Prerequisites: DE4 (acceptance unverified)
Contract: dc_dialect_bridge
Acceptance: BigQuery and Snowflake SQL run against the vendors' free quotas as ORACLE HOSTS outside the build path, results returned to DE1 for the multiset compare; the bridge is data (dialect rows), the credentials are vault leases, and a refused connection is recorded UNREACHABLE and never scored [@spider226]
Authored effort: 3. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
DE6 — DS-1000 referee replay through an isolated Python oracle host
Prerequisites: None declared; this does not establish execution eligibility.
Contract: fg_dscode
Acceptance: DS-1000 prompts and tests mirrored with licence; candidate completions from the engine dispatcher are executed on an isolated oracle host (Python is a third party and stays outside the estate's run path) and graded by the benchmark's own test cases, with the receipt carrying the host hash, n and the per-library breakdown [@ds100022]
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.
DE7 — Agentic analysis runs, DABstep and DA-Code
Prerequisites: DE6 (acceptance unverified)
Contract: ad_agent_run
Acceptance: the workflow engine drives a plan, execute and observe loop whose tools are the estate's readers, profiler and approximate analytics, against the DABstep and DA-Code tasks mirrored with licence, graded by each benchmark's own answer checker on the oracle host; every run publishes completion rate and executable-code rate the way DA-Code's board does [@dabstep26] [@dacode24] [@dsbench25]
Authored effort: 3. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
DE8 — dbt project tasks, Spider 2.0-DBT
Prerequisites: DE5 (acceptance unverified)
Contract: dc_dbt_project
Acceptance: the 68 DBT examples run through the workflow engine with the dialect bridge, graded by the benchmark's own result checks on the oracle host [@spider226]
Authored effort: 3. Executor kind: Organ. Responsible, accountable and verifier not established. Inspect retained worklog.
Runtime observation: UNKNOWN; outcome UNKNOWN; observed epoch -1. Declared owner: . Receipt , SHA-256 . Identity-bound observation is not accepted completion. Qualification and deployment remain unverified by this projection.
DE9 — GUI-driven data workflows, Spider2-V
Prerequisites: DE7 (acceptance unverified)
Contract: br_data_workflow
Acceptance: the sovereign browser drives the tools' GUIs for the 494 Spider2-V tasks on an isolated oracle host, graded by the benchmark's own checkers; the GPT-4o baseline is the published floor and the estate publishes its own success rate beside it [@spider2v25]
Authored effort: 4. 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 dataeng 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.

DE1 Execution judge for SQL over SQLite · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`sqr_exec_judge` · executor Organ · 1 u · stands on nothing
Definition of done	DONE-RULE PRE-DECLARED: the judge runs a gold query and a candidate query against the same SQLite file through nx_sqlite_rows, compares the result multisets order-insensitively, and prints EX 1 or 0 per pair with the row counts; a gate plants a candidate that returns the right rows in another order (must read 1), a candidate with one extra row (must read 0) and a candidate that errors (must read 0 with the error named); no rival oracle in the path [bird26]
Matrix cell	NO-MATRIX-ROW — no matrix row carries `sqr_exec_judge`; 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 BOARD ADMITTED 2026-09-16 from eight sources fetched over the sovereign fetcher and read from the mirrored bytes on both hosts (byte-identical): Spider 2.0, Spider2-V, BIRD, DS-1000, DABstep, LiveSQLBench, DSBench and DA-Code; the hand page that held this domain since 2026-08-13 leaves hand.roster by matrix admission; the OSWorld results sheet read the same day through nx_xlsx is the automation board's, cited there

DE2 BIRD Mini-Dev harness · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`dc_bird_minidev` · executor Organ · 1.5 u · stands on DE1
Definition of done	the Mini-Dev 500-pair set is mirrored from Hugging Face with its licence beside the bytes and the schema files, prepped into queries and gold TSVs the way nx_bright_prep shapes BRIGHT, and every candidate file is scored by DE1 into a receipt with n, EX and the harness hash; ACCEPT RULE: the gold set scored against itself reads EX 500 of 500 before any model is scored [bird26]
Matrix cell	NO-MATRIX-ROW — no matrix row carries `dc_bird_minidev`; 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	no log row names this rung yet

DE3 Sovereign text-to-SQL contender · cell WATCHING · gate NO-GATE-DECLARED

Contract	`sqr_text2sql` · executor Organ · 3 u · stands on DE2
Definition of done	a query writer over nx_engine_dispatch (local no-float seat, external U tier, frontier) that emits SQL for a natural-language question and a schema, scored only by DE1 on DE2; the receipt names the engine and its ceiling; the local seat's expected loss under its 96-token ceiling is published, never hidden; done when a receipt exists for all three engines
Matrix cell	Text-to-SQL over SQLite (BIRD dev and test) · `runtime/nx_sqlite_rows.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_sqlite_rows`; declare `nx_sqlite_rows TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

DE4 Spider 2.0-Lite SQLite subset harness · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`dc_spider2_lite` · executor Organ · 1.5 u · stands on DE1
Definition of done	the 135 SQLite examples of Spider 2.0-Lite (547 in total; the BigQuery and Snowflake examples wait for DE5) mirrored with licence and scored by DE1; the receipt states the subset size beside the number so a 135-example score is never read as the 547 leaderboard's [spider226]
Matrix cell	NO-MATRIX-ROW — no matrix row carries `dc_spider2_lite`; 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	no log row names this rung yet

DE5 Warehouse dialect bridge · cell WATCHING · gate NO-GATE-DECLARED

Contract	`dc_dialect_bridge` · executor Organ · 3 u · stands on DE4
Definition of done	BigQuery and Snowflake SQL run against the vendors' free quotas as ORACLE HOSTS outside the build path, results returned to DE1 for the multiset compare; the bridge is data (dialect rows), the credentials are vault leases, and a refused connection is recorded UNREACHABLE and never scored [spider226]
Matrix cell	Enterprise warehouse dialects (BigQuery, Snowflake) · `runtime/nx_dataeng_sota_census.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_dataeng_sota_census`; declare `nx_dataeng_sota_census TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

DE6 DS-1000 referee replay through an isolated Python oracle host · cell WATCHING · gate NO-GATE-DECLARED

Contract	`fg_dscode` · executor Organ · 2 u · stands on nothing
Definition of done	DS-1000 prompts and tests mirrored with licence; candidate completions from the engine dispatcher are executed on an isolated oracle host (Python is a third party and stays outside the estate's run path) and graded by the benchmark's own test cases, with the receipt carrying the host hash, n and the per-library breakdown [ds100022]
Matrix cell	Data-science code generation (DS-1000 libraries) · `runtime/nx_forge.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_forge`; declare `nx_forge TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

DE7 Agentic analysis runs, DABstep and DA-Code · cell NO-MATRIX-ROW · gate NO-MATRIX-ROW

Contract	`ad_agent_run` · executor Organ · 3 u · stands on DE6
Definition of done	the workflow engine drives a plan, execute and observe loop whose tools are the estate's readers, profiler and approximate analytics, against the DABstep and DA-Code tasks mirrored with licence, graded by each benchmark's own answer checker on the oracle host; every run publishes completion rate and executable-code rate the way DA-Code's board does [dabstep26] [dacode24] [dsbench25]
Matrix cell	NO-MATRIX-ROW — no matrix row carries `ad_agent_run`; 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	no log row names this rung yet

DE8 dbt project tasks, Spider 2.0-DBT · cell WATCHING · gate NO-GATE-DECLARED

Contract	`dc_dbt_project` · executor Organ · 3 u · stands on DE5
Definition of done	the 68 DBT examples run through the workflow engine with the dialect bridge, graded by the benchmark's own result checks on the oracle host [spider226]
Matrix cell	dbt transformation projects (Spider 2.0-DBT) · `runtime/nx_dataeng_sota_census.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_dataeng_sota_census`; declare `nx_dataeng_sota_census TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

DE9 GUI-driven data workflows, Spider2-V · cell WATCHING · gate NO-GATE-DECLARED

Contract	`br_data_workflow` · executor Organ · 4 u · stands on DE7
Definition of done	the sovereign browser drives the tools' GUIs for the 494 Spider2-V tasks on an isolated oracle host, graded by the benchmark's own checkers; the GPT-4o baseline is the published floor and the estate publishes its own success rate beside it [spider2v25]
Matrix cell	GUI-driven data workflows (Spider2-V) · `runtime/nx_browser.nx` · WATCHING decl
Gate	NO-GATE-DECLARED — `knowledge/organ_gate.conf` carries no row for `nx_browser`; declare `nx_browser TAB <gate>` and the ship loop proves it and the verdict lands here.
Work log	no log row names this rung yet

dossiers=9 cells_joined=5 no_matrix_row=4 (sum 9) gates_recorded=0 gates_unrecorded=0 no_gate_declared=5 (sum 5 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. 12 feature rows mapped of 13 matrix rows; 1 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/dataeng.cdmap (cut|x|y, cube|x|y|z); absent = the HBR defaults

bubble area = breadth · ring = momentum (green rising, red falling, grey until day two) · dashed = category means · axes fitted to the field of play: centrality 400–900, distinctiveness 200–600 of 0–1000 permil (the full range put every player in one corner)
readability of the 2D cut, self-graded by the layout ruler: label overlaps 0 · labels over marks 0 · off-canvas 0 · unresolved labels 0 · mark overlaps 0 (a fact of the data: two players that close are that close) · data spread 613 permil of the plot · quadrant words unseated 0
text contrast, measured with wcag2-ratio (floors from contrast.conf), light theme: labels 16.24 (floor 4.50) · notes 16.24 (floor 4.50) · quadrant words 5.89 (floor 4.50) · tick numerals 5.89 (floor 4.50) · axis titles 5.59 (floor 4.50) · dark theme: labels 13.78 (floor 4.50) · quadrant words 5.55 (floor 4.50) · axis titles 5.97 (floor 4.50) · dark classes under their floor 0 (one figure serves both themes: a dark shortfall is a token to fix, never a class to hide) · classes refused under their floor 0 (a refused class is not drawn; the scale classes are measured, never hidden) · export: SVG PNG (receipt, rendered by the estate's own rasteriser from this page)

axes fitted: centrality 400–900 · distinctiveness 200–600 · breadth 100–600 permil · farther bubbles are painted first, nearer ones over them · the floor shadow is each bubble's (x, y) at height 0
readability of the 3D cube, self-graded by the layout ruler: label overlaps 0 · labels over marks 0 · off-canvas 2 · unresolved labels 0 · mark overlaps 1 (a fact of the data: two players that close are that close) · data spread 185 permil of the plot

Nishi Spider-Agent (xlang) Agentar-Scale-SQL (Ant Group) DA-Agent (DA-Code) Human data engineers (BIRD)
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 4 (a fact of the data: two players that close are that close) · data spread 544 permil of the plot

Player	Quadrant	centrality	distinctiveness	breadth	depth	momentum	Rows held	Days on spine	First seen
Nishi	Unconventional	583	500	333	800	0 since 2026-09-23	5	9	2026-09-16
Spider-Agent (xlang)	Peripheral	583	250	222	666	0 since 2026-09-23	4	9	2026-09-16
Agentar-Scale-SQL (Ant Group)	Peripheral	500	277	194	777	0 since 2026-09-23	3	9	2026-09-16
DA-Agent (DA-Code)	Peripheral	583	333	250	750	0 since 2026-09-23	4	9	2026-09-16
Human data engineers (BIRD)	Aspirational	750	541	500	750	0 since 2026-09-23	8	9	2026-09-16

Day	Matrix rows reviewed	Players recorded
2026-09-16	13	5
2026-09-17	13	5
2026-09-18	12	5
2026-09-19	12	5
2026-09-20	12	5
2026-09-21	12	5
2026-09-22	12	5
2026-09-23	12	5
2026-10-08	12	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 52, unassessed 0. Authored assessments are not verified outcomes.

Capability	Nishi	Spider-Agent (xlang)	Agentar-Scale-SQL (Ant Group)	DA-Agent (DA-Code)	Human data engineers (BIRD)
Text-to-SQL over SQLite (BIRD dev and test) DARK — `runtime/nx_sqlite_rows.nx` EXISTS but does not declare `sqr_text2sql`: 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. The BIRD execution-accuracy leaders are agent systems at 82.39 test against a human 72.37 [bird26]; Spider-Agent covers the SQLite third of Spider 2.0-Lite [spider226]; DA-Code grades analysis code, not SQL [dacode24]; the estate reads SQLite pages and rows (sqr_walk) and writes no SQL from language -- DE3 is the contender behind the DE1 judge and the DE2 harness dossierwatching sqr_text2sql Pros none measured yetCons behind Spider-Agent (xlang), Agentar-Scale-SQL (Ant Group) (leads), Human data engineers (BIRD); open contract; expected source declaration not established	○	●	◉	○	●
Enterprise warehouse dialects (BigQuery, Snowflake) DARK — `runtime/nx_dataeng_sota_census.nx` EXISTS but does not declare `dc_dialect_bridge`: 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. Spider 2.0-Snow and Lite run on Snowflake and BigQuery with Spider-Agent as the published reference agent [spider226]; the estate has no warehouse dialect and no vendor connection -- DE5 puts the vendors' free quotas outside the build path as oracle hosts dossierwatching dc_dialect_bridge Pros none measured yetCons behind Spider-Agent (xlang), Human data engineers (BIRD); open contract; expected source declaration not established	○	●	○	○	●
dbt transformation projects (Spider 2.0-DBT) DARK — `runtime/nx_dataeng_sota_census.nx` EXISTS but does not declare `dc_dbt_project`: 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. 68 code-agent examples whose leader reads 65.6 and whose Spider-Agent variants read 35 to 40 [spider226]; nothing in the estate runs a dbt project -- DE8 dossierwatching dc_dbt_project Pros none measured yetCons behind Spider-Agent (xlang), Human data engineers (BIRD); open contract; expected source declaration not established	○	●	○	○	●
Data-science code generation (DS-1000 libraries) DARK — `runtime/nx_forge.nx` EXISTS but does not declare `fg_dscode`: 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. DS-1000 grades pandas, numpy, scipy and friends by execution and holds Codex-002 at 43.3 on its own page [ds100022]; DA-Agent writes and runs analysis code [dacode24]; the forge builds and runs NishiLang, not Python -- DE6 replays through an isolated oracle host dossierwatching fg_dscode Pros none measured yetCons behind DA-Agent (DA-Code), Human data engineers (BIRD); open contract; expected source declaration not established	○	○	○	●	●
Agentic multi-step data analysis (DA-Code, DSBench, DABstep) Measured: `ad_profile` exists in `runtime/_hdl_build/nx_analyst_data.nx`, verified at emit. The estate's profiler (ad_profile) is a tool, not an agent: it summarises a table and stops. DA-Agent plans, writes and executes across steps and is the DA-Code reference [dacode24]; DSBench and DABstep grade the same loop end to end [dsbench25] [dabstep26] -- DE7 wires the workflow engine around the estate's tools Adoption: `LIB-WIRED importers=3 nonval=2` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Spider-Agent (xlang), Agentar-Scale-SQL (Ant Group); fully adopted on the estate ladderCons behind DA-Agent (DA-Code) (leads)	●	○	○	◉	●
Columnar and tabular file reading (Parquet, CSV, TSV, XLSX) Source present; authored stronger claim: `pq_parse_rowgroups` in `runtime/nx_parquet_lib.nx`, verified at emit. Sovereign readers with derived capacities: Parquet (thrift-compact, snappy, dictionary pages -- the reader that reproduces the BRIGHT card), CSV and TSV (nx_value_parse_csv) and XLSX (nx_xlsx read the 1077-row OSWorld sheet on admission day); every rival reads these through Python libraries. EXCEED: no third party in the read path and every buffer derived from the file, never a ceiling rival marks uncited — the Yes, Best or Part codes on this row are an observation read with no reference mark behind them Adoption: `LIB-WIRED importers=3 nonval=2` — fully adopted (top of its ladder). Pros Source symbol found; stronger capability claim is authored; ahead of Spider-Agent (xlang), Agentar-Scale-SQL (Ant Group), DA-Agent (DA-Code), Human data engineers (BIRD); fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	◉	●	●	●	●
Approximate analytics measured against exact (HLL distinct, t-digest quantiles) Measured: `dfa_quantile` exists in `runtime/_hdl_build/nx_dataframe_approx.nx`, verified at emit. Distinct counts and quantiles with a published error against the exact answer (dfa_err_permil); no rival column documents an approximate path, and the benchmarks do not grade one Adoption: `LIB-WIRED importers=2 nonval=1` — fully adopted (top of its ladder). Pros Implementing source symbol found; runtime acceptance separate; ahead of Spider-Agent (xlang), Agentar-Scale-SQL (Ant Group), DA-Agent (DA-Code), Human data engineers (BIRD); fully adopted on the estate ladderCons none on the measured axes (rival marks are documented presence, not depth)	●	○	○	○	○
Execution-based grading, byte-exact Measured: `lb_one` exists in `runtime/nx_forge_local_bench.nx`, verified at emit. Every benchmark on this board grades by execution; Spider-Agent and DA-Agent iterate on execution feedback [spider226] [dacode24]. The estate's judge is a fresh compile and run compared byte-exactly (lb_one), deterministic and free, the reason a sovereign harness can publish n and a hash beside every number 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 Human data engineers (BIRD)Cons not adopted yet: source exists, never compiled: /api/build it	●	●	●	●	○
GUI-driven data workflows (Spider2-V) Open — watching `runtime/nx_browser.nx` : `br_data_workflow`, re-measured on every compare beat. Ship that symbol and this mark flips itself; the `comparewatch-` plane row flips with it. 494 tasks through the GUIs of professional data tools with a GPT-4o baseline agent [spider2v25]; humans do this work daily; the sovereign browser renders pages and does not yet drive a tool -- DE9 dossierwatching br_data_workflow Pros none measured yetCons behind Human data engineers (BIRD) (leads); open contract; expected source declaration not established	○	○	○	○	◉
Live contamination-free evaluation (LiveSQLBench) DARK — `runtime/nx_dataeng_sota_census.nx` EXISTS but does not declare `dc_live_set`: 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. Tasks rotated against training contamination, dated 2026-04 to 2026-06 [livesqlbench26]; the estate's census (dc_ground) grounds rows against organs and does not yet pull a live set; the pattern is the search board's rolling BRIGHT sweep, to be shared watching dc_live_set Pros none measured yetCons open contract; expected source declaration not established	○	○	○	○	○
Referee replay from a rival's published outputs Source present; authored stronger claim: `ri_stage_split` in `runtime/nx_bright_reasoned_import.nx`, verified at emit. A published output file is matched row by row to the estate's own task set and scored by the estate's judge before a byte of the rival's weights is pulled (landed on the search board 2026-09-16: the 1.5B rewriter's outputs measured 229 permil through our BM25). EXCEED: no rival ships a replay importer; theirs are scored by their own harnesses Adoption: `REGISTERED-DARK` — PARTIAL: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked). not adopted: callable, authorised, completed MCP invocation not established by supplied journal window; lifetime history unknown (direct execution may be recorded elsewhere; this scan does not establish lifetime use); no execution surface runs it either (clock, cron, daemon, roster, actlog and surfaced forks checked) Pros Source symbol found; stronger capability claim is authored; ahead of Spider-Agent (xlang), Agentar-Scale-SQL (Ant Group), DA-Agent (DA-Code), Human data engineers (BIRD)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)	◉	○	○	○	○
SQL issue repair (BIRD-CRITIC, SWE-SQL) DARK — `runtime/nx_sqlite_rows.nx` EXISTS but does not declare `sqr_sql_repair`: 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. BIRD-CRITIC and SWE-SQL (NeurIPS 2025) grade whether a broken query is repaired [bird26]; a data engineer does this by hand; the estate has the byte-exact repair loop for NishiLang (coding CB1) and none for SQL watching sqr_sql_repair Pros none measured yetCons behind Human data engineers (BIRD); open contract; expected source declaration not established	○	○	○	○	●
Interactive clarification with the user (BIRD-Interact) DARK — `runtime/nx_dataeng_sota_census.nx` EXISTS but does not declare `dc_interact`: 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. BIRD-Interact (ICLR 2026 oral) grades an agent that asks before it answers an ambiguous question [bird26]; humans do; no reference agent on this page documents it watching dc_interact Pros none measured yetCons behind Human data engineers (BIRD) (leads); 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	10	rows carrying a Yes, Best or Part in a rival column
cited	9	a reference mark resolves to a pinned mirror
uncited	1	an observation read, badged until a reference lands
rows, summed	10	RECONCILED: the parts sum to the total 10

Delivery and evidence

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

Release history and work log

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

When	Rung	Kind	Entry
2026-09-16	DE1	land	BOARD ADMITTED 2026-09-16 from eight sources fetched over the sovereign fetcher and read from the mirrored bytes on both hosts (byte-identical): Spider 2.0, Spider2-V, BIRD, DS-1000, DABstep, LiveSQLBench, DSBench and DA-Code; the hand page that held this domain since 2026-08-13 leaves hand.roster by matrix admission; the OSWorld results sheet read the same day through nx_xlsx is the automation board's, cited there

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. <b>Honest position, measured 2026-09-16 at admission.</b> The estate READS every table shape the benchmarks ship (SQLite, Parquet, CSV, TSV, XLSX), profiles and approximates, grades code byte-exactly and can replay a rival's published outputs through its own judge -- and it WRITES no SQL and no analysis code from language. Every row that names a benchmark contract is an honest absence with its rung on dataeng.plan; every rival number on this page is a cell read from a mirrored leaderboard, never a run the estate made. BIRD's September-2026 leaders clear the human test bar (82.39 against 72.37) while humans still hold dev (92.96) [bird26]; Spider 2.0's March-2026 tracks read 96.70, 65.6 and 76.23 [spider226]; the agentic boards grade end to end [dacode24] [dsbench25] [dabstep26]. The estate's exceeds are structural, not scores: a sovereign columnar reader with no third party in the path, and a referee-replay importer no rival ships.

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

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

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 13 capabilities: 12 exclusive to it and 1 shared with other boards. Shared rows on this page: 1 = home here 0 + home elsewhere 0 + ambiguous 0 + undecided 1 + refused 0 (partition 1); undecided = unmapped 0 + unproven 1 + unevidenced 0; homonyms 0; since the previous census: new 0, resolved 0. Census FRESH age_s=7225.

Capability	Organ	Boards grading it	Home	Rule	Evidence	Standing on this board
lb_one	`nx_forge_local_bench.nx`	team, dataeng	—	none	unproven	undecided

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 1396330 s), sources read 4 of 7 declared — a BLIND census is a FLOOR: unread sources can only add rows.

kind	name	board	source	evidence

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

References

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

[spider226] Spider 2.0 leaderboard page (xlang-ai, ICLR 2025 oral), fetched and read 16 Sep 2026: 632 real-world enterprise text-to-SQL workflow problems in three tracks, leaders read from the page's own tables (the newest dated entry is 2026-07-28, re-read 1 Oct 2026 from byte-identical bytes; the 16 Sep read reported 2026-03-01, which is only the first table's top-row date): Spider 2.0-Snow (547 examples on Snowflake) Genloop Sentinel Agent v2 Pro 96.70, Spider 2.0-DBT (68 code-agent examples) SignalPilot Agent 65.6, Spider 2.0-Lite (547 examples over BigQuery, Snowflake and SQLite, 135 of them SQLite) Tianqiong Data Agent plus GLM 5.2 76.23. publisher · read in our library knowledge/fetched/cmp_dataeng_spider2_leaderboard_202610.html · pin h3cd4bb95ed0f2d9d5e3aa76cb2461f9f22c8af8ac7b7d6136e3589ce01474e61 · accessed 2026-10-01 · mirrored-indexGrounds: Enterprise warehouse dialects (BigQuery, Snowflake)
[spider2v25] Spider2-V project page (xlang-ai), fetched and read 16 Sep 2026: 494 real-world tasks across complete data science and engineering workflows from warehousing to orchestration, driven through the GUIs of professional tools such as BigQuery, dbt and Airbyte; the baseline agent is GPT-4o; newest dated news 2025-01-16. publisher · read in our library knowledge/fetched/cmp_dataeng_spider2v.html · pin h12149bba816558ee318ba33afaf2278728dc064b4458a412c557b3eed629dd05 · accessed 2026-09-16 · vendor-docGrounds: GUI-driven data workflows (Spider2-V)
[bird26] BIRD text-to-SQL benchmark page (bird-bench), fetched and read 16 Sep 2026 and re-fetched 1 Oct 2026 (new top row GrainSQL, Sarim Chaudhry, Purdue University, dev 77.77 and test 82.95 dated 2026-09-26): the execution-accuracy leaderboard's top rows with their dates, DataGallery-Text2SQL (Huawei 2012 Labs) dev 78.10 and test 82.39 dated 2026-09-07, SiriusAI-SQL (Tencent) 77.77 and 82.28 dated 2026-08-22, AskData plus GPT-4o (AT and T CDO) 77.64 and 81.95 dated 2025-12-16, Agentar-Scale-SQL (Ant Group) 74.90 and 81.67 dated 2025-09-25; human performance 92.96 dev and 72.37 test; the page also announces BIRD-Interact (ICLR 2026 oral), SWE-SQL (NeurIPS 2025), BIRD-CRITIC and the Mini-Dev 500 set on Hugging Face. publisher · read in our library knowledge/fetched/cmp_dataeng_bird_202610.html · pin hfd87838abc6a2440dca3acbb2bf89776e1054603885337659f185709955a69bc · accessed 2026-10-01 · mirrored-indexGrounds: Text-to-SQL over SQLite (BIRD dev and test)
[ds100022] DS-1000 project page (Lai et al., arXiv 2211.11501, 2022), fetched and read 16 Sep 2026: 1000 data-science questions over seven Python libraries with test cases and multi-criteria execution evaluation, perturbed against memorisation; the page names Codex-002 at 43.3 percent accuracy as the best public system and delegates the live leaderboard to a separate Eval-Arena page. publisher · read in our library knowledge/fetched/cmp_dataeng_ds1000.html · pin he0f1750192affe61531f842e6ccb6dcf16699a14c76667f34954fc947226d390 · accessed 2026-09-16 · vendor-docGrounds: Data-science code generation (DS-1000 libraries)
[dabstep26] DABstep dataset card on Hugging Face (adyen/DABstep, api json), fetched and read 16 Sep 2026: the data agent benchmark for multi-step reasoning over real financial data; the card was last modified 2026-09-16. publisher · read in our library knowledge/fetched/cmp_dataeng_dabstep_api.json · pin h0fbf615880d5f3a2467ee2eead334e347d5540932d29b35c1ace7d5264e58578 · accessed 2026-09-16 · datasetGrounds: Agentic multi-step data analysis (DA-Code, DSBench, DABstep)
[livesqlbench26] LiveSQLBench page (a dynamic, contamination-free text-to-SQL benchmark by the BIRD team), fetched and read 16 Sep 2026: dated release items from 2026-04-25 to 2026-06-01; the Base-Full-V1 600-task set is announced on the BIRD page. publisher · read in our library knowledge/fetched/cmp_dataeng_livesqlbench.html · pin h60d1d169ce6ff6a88e13c8a7896d1501a0e53b945ca0fe7acb0d964a2ba82933 · accessed 2026-09-16 · vendor-docGrounds: Live contamination-free evaluation (LiveSQLBench)
[dsbench25] DSBench README (Jing et al., arXiv 2409.07703, ICLR 2025), fetched and read 16 Sep 2026: realistic data analysis and data modeling tasks for data-science agents; the README records its use in OpenAI's ChatGPT agent evaluation on 2025-07-17. publisher · read in our library knowledge/fetched/cmp_dataeng_dsbench_readme.md · pin hda3cbf56031e92c7adbf361c0f046526b254a47b6d38bb900f85f778ed57ef56 · accessed 2026-09-16 · source-readGrounds: Agentic multi-step data analysis (DA-Code, DSBench, DABstep)
[dacode24] DA-Code project page (EMNLP 2024, released 2024-10-03), fetched and read 16 Sep 2026: an agent data-science code generation benchmark; the page reports that even GPT-4 solves only 6.0 percent of the Spider 2.0 workflow tasks against 86.6 on Spider 1.0 and 57.4 on BIRD, and publishes a DA-Agent baseline leaderboard of completion rate and executable-code rate. publisher · read in our library knowledge/fetched/cmp_dataeng_dacode.html · pin he5b08d8966ba1212c9d30ab5f51a804c6e4afec8cfac11db758d62801d7c370d · accessed 2026-09-16 · vendor-docGrounds: Agentic multi-step data analysis (DA-Code, DSBench, DABstep)

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