We built the best, free, open-source Text2SQL tool in a day. Is it worth it? "Solving Text2SQL" was the second most common startup idea for three YC batches in a row. Everyone thought it was a no-brainer: the Big Data market was valued at roughly $330B in 2023 and the LLM tsunami was cresting. Text2SQL has lured trillion-dollar companies into billion-dollar holes. It's an extremely hard problem to solve—until the next generation of AI models comes along and... handles it. To prove this, we built Ana Small, an extremely minimal version of our Enterprise product. It has no advanced features: no hermetic Python sandbox; no visualizations; no guarantee of reproducibility via ontology; no automated runs. Ana Small is equipped merely with o3-mini and a single tool that can execute SQL, and it can connect with data warehouses in Redshift. It took one day. And yet, it still outperforms almost every major product in the market by 2-3x: Amazon Q, Databricks Genie, Hex, Snowflake Cortex Analyst. Check out the results of our tests with our internal Conglomerate benchmark below (public release coming soon). Ana Small makes one thing abundantly clear: The Bitter Lesson still casts a long shadow. More compute wins out in the end. So, was it worth the hype? The short answer is, no. Why? People don't want to talk to their data: they want their data to tell them what they need to do. Data insights are still extremely valuable, but you're going to have to work much harder than just Text2SQL to build a great company. Check out our blogpost here: https://lnkd.in/dZt5PM_J Try Ana Small here: https://small.textql.com
TextQL
数据基础架构与分析
San Francisco,California 1,209 位关注者
Building the Ontology for LLMs to talk to Data, accurately and securely.
关于我们
TextQL is a platform that simplifies the data-to-insight process for organizations. The platform indexes BI tools and semantic layers, documents data in dbt, and uses OpenAI and language models to provide self-serve power analytics. With TextQL, non-technical users can easily and quickly work with data by asking questions in their work context (Slack/Teams/email) and getting automated answers quickly and safely. The platform also leverages NLP and semantic layers, including the dbt Labs semantic layer, to ensure reasonable solutions. TextQL's elegant handoffs to human analysts, when required, dramatically simplify the whole question-to-answer process with AI. Reach out to learn more: calendly.com/ethanding
- 网站
-
textql.com
TextQL的外部链接
- 所属行业
- 数据基础架构与分析
- 规模
- 11-50 人
- 总部
- San Francisco,California
- 类型
- 私人持股
地点
-
主要
US,California,San Francisco
TextQL员工
动态
-
TextQL转发了
of course people think ai can’t write sql. it seems like most companies selling it basically can’t? we reduced ana to its essentials because we were curious how far the bare minimum gets us as is the case time and time again w/ ai, simple environments, clear goals/constraints leads to amazing results re:?The Bitter Lesson (link in comments) excited to show what we cooked up tomorrow
-
-
We’re excited to announce our new ?????????? ??????????????????????—bringing real-time data analysis right into your Slack channels and DMs. Check out our docs to learn more and see how you can add Ana to your Slack workspace: https://lnkd.in/eQdrsg_f
-
-
AI agents are transforming industries, and retail is no exception. This weekend, TextQL will be at the NRF 2025: Retail's Big Show to showcase how our technology helps retailers uncover deep insights that drive buisiness outcomes. From inventory optimization to personalized customer experiences, our AI agents integrate seamlessly with your existing tools to deliver actionable insights and drive measurable results. Check out our dedicated NRF page to learn more about our solutions: nrf.textql.com. If you're attending NRF, let's talk about AI-powered retail! #NRF2025 #AIinRetail #RetailInnovation
-
-
these best practices seem really simple if you’re heads-down in ai and using it everyday. but the truth is we’re still in the first out of the first inning in the adoption of ai products. for most people, working with ai agents is still a very new thing and definitely requires training and onboarding. whenever we onboard users to ana, our ai data scientist, we typically run workshops & office hours. these are some of the keys to success that have enabled our users to turbo-charge their productivity by 10x
-
-
One of the best ways to get started with Ana is to treat her like a human data analyst that you’re collaborating with and ask her general questions about your data. Things like: - Give me an overview of my data - What type of analysis can I do? A common theme we keep hearing from data teams who have implemented TextQL is how they are reducing their back and forth between various teams and empowering users to perform their own exploratory data analysis.
-
TextQL is proud to announce our corporate-friendly inclusion in the AI Hot 100 List by The AI Furnace ???? . We will attend The AI Hot 100 Conference in New York City on September 18th. We appreciate the assessment committee for identifying how mind-blowingly awesome we are and how we stand head and shoulders over any other product in the market. It's about time someone built a chat-with-your-data product, and we were surprised no one else was attempting it. #AIHot100 #AIConference #AIInnovation #Multimodal
-