live wire

▸Oracle Select AI generates SQL/PGQ for property graphs—but errors if one AI profile mixes graphs with tables or viewsOracle▸Domo’s data team standardizes MCP builds with shared scaffolding; beta toolkits inherit user permissionsDomo▸Databricks proposes five outcome signals for agentic BI—and says dashboard counts and license usage miss decision qualityDatabricks▸One vibe-coded database app needed 56+ human test hours and three repair rounds despite AI Studio’s verification claimsMancas & Mancas / arXiv▸BigQuery AI.IF’s default cost mode trains a local model for most rows—and drops row-error thresholdsGoogle Cloud▸AWS’s multi-tenant data-agent pattern gives the broker Lambda no data permissions; Cognito roles carry Lake Formation scopeAWS Big Data Blog▸Amazon OpenSearch’s optimized engine adds nested agent traces, but AWS reserves it for log analytics—not vector searchAWS▸Two interpretation-based provenance semantics cannot express SQL bag difference, new theorem showsKolaitis / arXiv▸Databricks Feature Views reuse one definition for training and live scoring, but online lookup still returns latest available stateDatabricks▸Databricks Genie’s Slack and Teams previews default channel answers to public, even when viewers lack source-data accessDatabricks▸Snowflake’s two-tool GraphRAG agent beat a seven-tool graph setup 70% to 60% on 22 ontology questionsSnowflake Engineering▸Semantix opens an Apache-2.0 semantic-first data-agent stack, but labels the project WIP and has no release or adoption yetGitHub / icosmos-space▸Confluent defines Streamhouse as shared current state for agents and apps, but says broader semantics and governance are still to comeConfluent▸Graph-repair retrieval cut F1 by 0.126 on one suite; a learned router used fewer tokens than always-on retrieval in all sixLin et al. / arXiv▸Oracle Select AI generates SQL/PGQ for property graphs—but errors if one AI profile mixes graphs with tables or viewsOracle▸Domo’s data team standardizes MCP builds with shared scaffolding; beta toolkits inherit user permissionsDomo▸Databricks proposes five outcome signals for agentic BI—and says dashboard counts and license usage miss decision qualityDatabricks▸One vibe-coded database app needed 56+ human test hours and three repair rounds despite AI Studio’s verification claimsMancas & Mancas / arXiv▸BigQuery AI.IF’s default cost mode trains a local model for most rows—and drops row-error thresholdsGoogle Cloud▸AWS’s multi-tenant data-agent pattern gives the broker Lambda no data permissions; Cognito roles carry Lake Formation scopeAWS Big Data Blog▸Amazon OpenSearch’s optimized engine adds nested agent traces, but AWS reserves it for log analytics—not vector searchAWS▸Two interpretation-based provenance semantics cannot express SQL bag difference, new theorem showsKolaitis / arXiv▸Databricks Feature Views reuse one definition for training and live scoring, but online lookup still returns latest available stateDatabricks▸Databricks Genie’s Slack and Teams previews default channel answers to public, even when viewers lack source-data accessDatabricks▸Snowflake’s two-tool GraphRAG agent beat a seven-tool graph setup 70% to 60% on 22 ontology questionsSnowflake Engineering▸Semantix opens an Apache-2.0 semantic-first data-agent stack, but labels the project WIP and has no release or adoption yetGitHub / icosmos-space▸Confluent defines Streamhouse as shared current state for agents and apps, but says broader semantics and governance are still to comeConfluent▸Graph-repair retrieval cut F1 by 0.126 on one suite; a learned router used fewer tokens than always-on retrieval in all sixLin et al. / arXiv

nl2sql.aithe text-to-sql wire

edition 2026-08 · published Aug 10, 2026

The nl2sql.ai Leaderboard — Launch Edition (August 2026)

#	system	category	score	Δ	verdict
1	Cortex AnalystSnowflake	cloud-native	88.0	—	The semantic-model bet paid off: best-in-class governed accuracy for enterprise Q&A inside one warehouse.
2	AI/BI GenieDatabricks	cloud-native	86.0	—	Curated Genie spaces plus feedback loops make it the strongest conversational-BI experience on a lakehouse.
3	SQLCoder / DefogDefog.ai	open-source	81.0	—	Open-weight SQLCoder models remain the best capability-per-parameter story in the space.
4	VannaVanna AI	open-source	79.5	—	The default OSS starting point: RAG over your own DDL and query pairs, any LLM, any store.
5	Gemini in BigQueryGoogle Cloud	cloud-native	78.0	—	Deeply integrated assistance across BigQuery Studio; strongest inside GCP, less portable outside it.
6	WrenAICanner	open-source	76.5	—	The OSS semantic-layer play: MDL grounding plus a polished GenBI UI, moving fast.
7	DB-GPTeosphoros-ai	open-source	74.0	—	Broadest OSS surface — agents, fine-tuning hub, workflows — at the cost of focus.
8	Copilot in Microsoft FabricMicrosoft	cloud-native	72.5	—	Ubiquitous inside the Microsoft stack; quality varies by workload, improving steadily.

methodology

Scores are 0–100, weighted across four pillars:

Capability (40%) — quality of generated SQL on hard, messy schemas; benchmark evidence where public (BIRD, Spider 2.0); semantic grounding and self-correction.
Trust & governance (25%) — how the system prevents wrong-but-plausible answers: semantic layers, verified queries, permissions awareness, auditability.
Adoption signals (20%) — production usage, community momentum (stars, releases, case studies), ecosystem integrations.
Openness (15%) — open weights/source, self-hosting, extensibility, transparent docs.

The launch edition was scored by the founding editorial pass and is the baseline. From here, the desk re-scores monthly: every score change must cite evidence (release notes, benchmark submissions, engineering posts), and every edition links its diff. Vendors never pay for placement; this publication has no commercial relationship with anything it ranks.