Skip navigation
AI can write SQL. But can it actually understand your data? ๐Ÿ‘€
0:00 / 0:59

AI can write SQL. But can it actually understand your data? ๐Ÿ‘€

Vikram

@Vikram_AI_1.6K followers

Visit on X
257 likesDownloadOriginal post

13,698 views

AI can write SQL. But can it actually understand your data? ๐Ÿ‘€ That's where things get interesting. Exmergo built Dex to help AI coding agents like Claude Code, Cursor, and Codex understand your data warehouse instead of blindly generating SQL. Think of it as giving your AI agents the context they actually need. โ†’ Understand tables and relationships. โ†’ Catch expensive queries before they run. โ†’ Build dbt models, tests, and semantic definitions. โ†’ Detect schema and data changes. โ†’ Define metrics and how data should be interpreted. The interesting part? Dex Semantics scored 80% on ADE-bench with Claude Sonnet 5.5, compared to 73.3% without Dex. That's 25% fewer failed tasks. ๐Ÿ”ฅ And it's open source! Instead of letting AI guess what your data means, give it the right context to get things done. Worth checking out! ๐Ÿ‘€ Support Dex on Product Hunt (it's live ): #AI #AIAgents #SQL #DataEngineering #OpenSource

39 Comments

Comments from the original post on X (Twitter)

  • Dharmik Harinkhede's profile picture

    @AiwithDharmik

    SQL is only part of the problem. Data understanding is what matters.

    2 likes

  • Vikram's profile picture

    @Vikram_AI_

    getting agents to actually understand the semantics is where the real battle is. thanks for feedback

  • Ofek Shaked | AI Engineer's profile picture

    @VibeCoderOfek

    dex gives the agent the schema. what happens when the warehouse changes the column name overnight

    1 like

  • 01azPublic's profile picture

    @silic2023

    โ€œActive customerโ€ is a good example of why schema alone isnโ€™t enough. Two perfectly valid queries can disagree because one counts logins and another counts purchases. Making that definition visible beside the answer would help people challenge the meaning, not just the SQL.

    1 like

  • Misbah Khursheed's profile picture

    @mmisbbahh

    catching expensive queries before they run is the real win here, agents will happily full scan a 2tb table to answer a yes or no

    1 like

  • Nahid's profile picture

    @nahid_tech

    This is a great step forward for AI-powered data engineering! Giving AI agents proper data context instead of letting them guess can make a huge difference in accuracy, efficiency, and reliability. Open source makes it even better! ๐Ÿ”ฅ๐Ÿ‘

    1 like

  • Magiber Rahman's profile picture

    @itmagiber

    amazing share

    1 like

  • Aivatar's profile picture

    @Worldwide2060

    AI writing SQL was never the hard part โ€” understanding the warehouse is. Blind SQL is how you get queries that run fine and answer the wrong question. Giving agents real context about the data is the fix that actually matters. Is Dex open for anyone to try on their own warehouse?

    1 like

  • pushkar soni's profile picture

    @pushkersoni72

    amazing

    1 like

  • Aria Arooj's profile picture

    @Arj_786

    This is super cool! Giving AI the context it needs makes so much sense. Definitely going to check this out.

    1 like

  • ๆฝœๆฐดๅ‘˜Balance's profile picture

    @balance0014

    does Dex pick up dbt docs too, or just the raw schema? column names alone rarely tell you which table is stale

    1 like

  • Rohit's profile picture

    @heyrohitai

    I like that the context lives alongside the models, congrats for launch

    1 like

  • Vikram's profile picture

    @Vikram_AI_

    Appreciate it @heyrohitai keeping it right alongside the models makes the workflow so much cleaner. Try : also share with your network๐Ÿ™Œ

  • Parag pawar's profile picture

    @promptparag

    Can Dex also help with complex joins across multiple schemas?

    1 like

  • Steven Cheng's profile picture

    @stevencheng

    Context beats guessing. My agents still hallucinate column names without it.

    1 like

  • Abhinav Sharma's profile picture

    @Abhinavs1920

    a query can run perfectly and count every customer twice. table names aren't enough, the agent needs the grain of each table... what if data is too big? will do hands on.

    1 like

  • Tanay Harinkhede's profile picture

    @Tanaypawar27

    Really interesting project, especially for dbt users.

    1 like

  • Vikram's profile picture

    @Vikram_AI_

    Yaaa exactly

  • A.W.E.S.O.M.-O 4000's profile picture

    @Awesome_O_AI

    Giving AI coding agents real context about your data warehouse instead of letting them guess SQL is a smart idea

    1 like

  • Vikram's profile picture

    @Vikram_AI_

    Exactly, blindly guessing schemas always ends up in broken queries. Check out Dex for more:

    1 like