OPEN LINK ↗
// 2h agoPRODUCT LAUNCH
Dex Semantics cuts agent analytics errors
Dex Semantics embeds code-level declarations into data models to capture row grains, join constraints, and business logic for AI coding agents. The framework reduces failed analytics engineering tasks by 25%, lifting Claude Sonnet to an 80% pass rate on ADE-bench.
// ANALYSIS
AI agents generate SQL with ease but routinely stumble over unspoken business logic and silent relational drift in production warehouses. Treating semantic rules as executable in-model code rather than an external metadata silo gives agents the guardrails needed to build reliable data pipelines.
- –Boosts Claude Sonnet from 73.3% to 80% on dbt Labs' ADE-bench, eliminating 25% of benchmark task failures
- –Keeps declarations as code within existing repositories instead of locking logic inside third-party metadata platforms
- –Translates semantic declarations into established formats including dbt MetricFlow, Cube, Apache Ossie, and Snowflake Semantic Views
- –Detects drift in schemas, metric definitions, and join cardinalities before faulty queries run
- –Provides safety guardrails including dev-target builds, upfront query cost quotes, and automated PII masking
// TAGS
dex-semanticsdexai-codingcoding-agentdata-toolsdevtoolopen-source
DISCOVERED
2h ago
2026-11-14
PUBLISHED
6h ago
2026-11-14
RELEVANCE
8/ 10
AUTHOR
[REDACTED]