Muse Wonder achieves 34.0% on Decision-Bench, our frontier benchmark for data-driven decision making. It improves substantially from Muse Spark’s 19.8%, but at over 2x the cost: Wonder uses fewer output tokens, but each one costs 3x more.

TextQL’s Ontology is now generally available. Business context and executable logic, together in files agents can read and use, and your team can review, version, and own. Business as code, built for the way agents work.

Dropbox is now connected to Ana. Ana can read the contents of your PDFs, docs, spreadsheets, and Paper docs — not just search filenames. Ask questions across Dropbox and your other connected data sources in one place. Available today to all TextQL customers.

Fable 5.1 is now live in Ana. Faster. Stronger on coding. Up to 45% cheaper on agentic workloads. That means deeper investigations and long-running agents in Ana just got faster and significantly cheaper to run at scale. Select Fable 5.1 from the model picker.

♠️ TextQL is growing our engineering team, and we'd rather meet you over cards than over a whiteboard. We're hosting a friendly poker night on Wednesday, August 26 in our new NYC office. All experience levels welcome. RSVP in comments ↓

Yesterday, we launched Data Apps to all customers. Today, Aaron Zhu walks you through how to build a working Data App from a single prompt. He shows how to build a data reconciliation workflow: a leasing intake portal that takes messy broker reports and uses an agent to validate them. The agent surfaces anomalies, brings in human input where needed, and writes clean, validated rows back to the warehouse. This is just one example. In the blog post below, we're rounding up a few more data apps that impressed us, for inspiration.

Meet Data Apps. We took what makes low-code and no-code app builders great: speed, no engineering bottleneck. And applied it directly to enterprise data. The result: fully operational apps built in one or two prompts. Accurate. And fully enterprise-ready out of the box, with version control and RBAC built in, not bolted on afterwards. We're offering free credits to Team and Enterprise organizations building their first three data apps. Try it today by describing what you want to build in a new chat. That's it. Learn more about customer examples in our blog post below 👇

Claude Opus 5 is now available in TextQL. To turn it on for your organization, head to Settings > Models and toggle Enabled next to Claude Opus 5.

🎙️ Watch the full podcast: youtu.be/PwhlxJrjQgQ?si=NdEP…

Enterprise data teams are still running modern data stacks like 1910s Detroit assembly lines. @TheEthanDing, CEO, sat down with @petesoder of @ZeroPrimeVC to talk about why that model is breaking now, and what TextQL does differently: working directly on messy enterprise data instead of demanding clean inputs first. The dirtier the data, the more this approach pays off. And because TextQL builds context over time, it gets cheaper with use instead of more expensive. They also covered the question every data leader is quietly panicking about: AI is spiking warehouse bills exponentially. The old playbook does not survive contact with that math. Thanks to Pete and the Zero Prime Ventures team for having us.

Here's a CSV upload of really messy Messi data.

Blackstone Credit & Insurance runs on 1,000s of tables, each with 100s of fields and 1,000,000,000s of rows. Until recently, knowing which tables to join, and how to join them without double-counting, was tribal knowledge. Rob Wisniewski, CTO of Credit & Insurance Technology at Blackstone, deployed TextQL's Ana on top of that data environment. → thousands of tables, billions of rows, joined accurately without analyst hand-holding → foreign key matching and entity de-duping handled natively → "time to theory and where to look" compressed from days to minutes → every answer comes with the work shown (non-negotiable in credit and insurance) "I asked Ana for a playbook on how to solve a problem... it created a scheduled playbook, configured it, and to this day every Monday I get the results in my email. That's the kind of thing you don't see in significantly more mature products." Most enterprises think faster analytics = faster queries. The bottleneck isn't query execution; it's time to theory: the gap between "I have a question" and "I know which tables to look at and how to join them without breaking the math." Ana closes that gap while showing its work for every answer. At Blackstone scale, that's not just a productivity gain; it's the difference between acting on a thesis this week and acting on it next quarter.

Your business context originates in APIs, and it's scattered across every tool that your team uses. In this walkthrough, Matt Abate, Head of AI Research at TextQL, shows how TextQL can connect to API-based sources like Salesforce, Slack, and Notion, and turn that scattered context into governed ontology inputs for Ana. No duplication, no data modeling. Ana can just reason across your business tools, as they are.

Today we're introducing Ontology Building mode in Ana. Ana builds your ontology with you, surfacing the objects, metrics, dimensions, and definitions your team argues about, and saving each one as reusable structure as you confirm it.

Every answer in TextQL now shows its work, inline. Each citation traces a number back through its entire provenance chain — a visual map of exactly how a number was derived, end to end. This includes which tables were read, the SQL that aggregated them, the Python that reshaped the result, and the answer that came out.

The insights you surface in a Thread can now be e-mailed to yourself or others in your organization. Ask Ana to share a result and she writes it up as clean HTML with charts and tables in place, tags the right colleagues, and delivers it inside your organization. That way, you no longer have to search through threads to resurface what you found.

Excited to partner with @Dropbox on bringing agentic analytics to one of the most sophisticated FP&A teams in software.

🚨BREAKING🚨 Zohran Mamdani will FORCE all NYC companies to fire any consultant charging $500/hr to build a data strategy deck. "To build a city in which all New Yorkers can thrive, we cannot allow McKinsey to sell us 47 slides that say 'become data-driven.'"

Most teams: Paying for Snowflake but only 3 people can query it TextQL: 3-minute setup, entire team gets self-service access New tutorial covers key-pair auth, schema mapping, the whole thing. Connect once. Ask questions forever. Link below.

Most teams: 3 days to write SQL queries for basic insights TextQL + PostgreSQL: 3 minutes New tutorial shows the setup. Neon, Supabase, AWS RDS—doesn't matter. Connect once. Ask questions forever. Link below.

We're throwing a happy hour during #JPM2026. Monday, Jan 12. 5:30pm. Neo Office, SF. Enterprise data and AI folks—come talk shop over drinks. No pitch decks allowed. Link Below.

santa came early: TextQL compute is now 80% cheaper turns out when AI agents query your warehouse 10x more someone has to pay for it. we decided it shouldn't be you two updates shipped this week:sandcastle lifecycles down to 4 hours → 80% lower sandbox costs token cache hit rate up from 40% to 52% → 17% lower token costs net result: 30%+ reduction in total costs for most customers our job is to drive the cost of data-driven decisions toward zero. every price cut is a feature

7. What's next. Collapsing the query engine into the sandbox (60% latency reduction estimates). Persistent sandboxes that become live applications instead of spinning down. Fine-grained network permissions so agents can call external APIs without binary "isolated vs connected" tradeoffs. (7/7)

6. Scaling with Kubernetes. We migrated to K8s. Now we scale from 10 to 1000+ sandboxes in 3 minutes. Spin-up in <300ms. Dynamic resizing from 64MB to 64GB. Multi-core support for heavy analytical workloads via gVisor secure kernel. (6/7)

5. 2023 MVP: single machine, 512GB RAM, 64 Docker containers. Paid for massive compute regardless of utilization. No dynamic resizing. Every update killed all sessions. 65th chat got rejected. (5/7)

4. Textables: the unified data layer. General-purpose sandboxes give you code execution. Sandcastles extend that with Textables, a unified interface across warehouses, BI tools, databases, and flat files. Each source speaks a different language; Textables abstract that away. Built on Apache Arrow, we stream data via zero-copy transfers at 10GB/s. (4/7)

3. The memory problem. LLMs are stateless, but data work is iterative. Snowflake returns 100K rows, APIs return nested JSON needing preprocessing. Pass that through the context window and you hit token limits immediately. Sandboxes give agents persistent, intermediate results live on disk, not in the prompt. (3/7)

2. Why agents need sandboxes. Enterprises have petabytes scattered across Snowflake, Postgres, Salesforce, Amplitude, dozens of APIs. For non-trivial analytics, agents need a place to dynamically integrate these sources. (2/7)

Most AI sandboxes are built to run code We needed one built to wrestle with petabytes of enterprise data across dozens of sources. Here's what we learned building Sandcastles (1/7)

Playbooks Now Have Memory🧠! Playbooks can now remember the context of their past runs, helping Ana refine her understanding and deliver more accurate, consistent results over time.

95% of Healthcare data lives in petabytes of SQL databases The tools for AI to use that data haven't existed Today we fix that with TextQL Healthcare 100,000+ tables. Trillions of rows. Petabytes of data. 15 minutes to insights. Epic systems with 100,000+ tables. Cerner environments. Claims databases. Clinical notes. Prior authorizations. Healthcare organizations have more data complexity than any other industry - and exactly zero AI platforms built to handle it. Until now. Here's what makes it different: 1. Direct access to ALL your systems. No migration required. Epic + Cerner + Claims + Snowflake + Databricks. Everything. At once. Other platforms: 6-month ETL projects. TextQL: Connect Monday. Query Tuesday. First insights in 15 minutes. 2. Healthcare-compliant execution environment. Autonomous agents running production code in SOC 2 Type II, HIPAA-compliant infrastructure. Full audit trails. On-premise deployment available. 3. Structured AND unstructured data. Simultaneously. Everyone else: Claims records OR clinical notes. Us: Both. At the same time. Make sense of 100,000s of tables without months of data prep. We're not launching with pilots. We're launching with Lumeris - powering their Tom™ AI platform delivering care to millions of Americans. Live partnership. Production workloads. Enterprise healthcare data at scale. Advisory Board of operators who've run organizations serving 120M+ Americans: - Varsha Rao (former CEO Nurx, COO Clover Health) - David Griffith (Trinity Life Sciences, ex-Pfizer) - Sam Mohanty (former CDO, Prime Therapeutics) - Jean-Claude Saghbini (CTO, Lumeris) - Raghu Chandra (30 year EHR Veteran) These aren't advisors. They're the people who built the systems we're now optimizing. Meet us at HLTH Conference next week - Booth #4060 Or request a demo: textql.com/hlth-conference Comment "HEALTHCARE" and we'll reach out for a customized demo!

Ana can now make playbooks from threads! Just tell Ana to turn your current conversation in a playbook.

Ana is Now Available in Slack! We've just released our Slack integration for Ana. You can now ask data questions directly in Slack channels or DMs and get instant analysis from your organization's data. Once you've setup Slack, you can just mention "@ Ana" in any channel, or send her a DM to get started. All your conversations sync seamlessly with your TextQL account.

Setup takes 10-15 minutes. Connect data sources, define analysis scope, set schedule, test output, VPC deployment. Ana handles everything else. Works with your existing warehouse, semantic layer, and communication tools from day one.

The Magic continues: Matt duplicates the Playbook for each account he owns. Suddenly every customer call has the same caliber of preparation. One Playbook saves at minimum 15-20 hours / month of manual analysis from your data team. Do that math for your company.

The Analytics Game-Changer 🎖️ Matt's got a customer call in an hour. Runs his "Customer Usage Deep Dive" Playbook. 90 seconds later he gets: Usage trends over 2 weeks Feature adoption analysis Comparison to similar cohorts Growth opportunities identified Churn signals flagged

Stored Instructions, Infinite Execution Teach Ana how to answer a business question once, she remembers forever. Not just SQL and Python, the entire analytical workflow and your visualization preference. Build you first Playbook here: app.textql.com/playbooks

The 3AM Advantage 🥱 Your customers don't stop using your product at 5 PM. Your European users are most active while you're asleep. Your weekly board reports are due Monday morning. Set customer health checks to run every Sunday at 2 AM. Ana's got you covered.

Picture this: Matt needs customer analysis for tomorrow's call. He slacks Ben at 6 PM. Ben's napping at his desk. But somehow, reports still land in Matt's inbox with insights, visualizations, and executive-ready formatting... 🧐

Get your data request done in your sleep. Playbooks are autonomous agent runs that: - Work with ANY existing infrastructure (Snowflake, BigQuery, Databricks, dbt, Looker, etc.) - Execute complex analysis workflows autonomously - Deploy entirely in your VPC with SOC2 Type II compliance - Scale from 1 playbook to hundreds without breaking your data team Here's Matt showing you how he automates all his weekly reports

Getting Started 🎬 No self-serve trials. We do 2-week proof of concepts. Connect to your warehouse, configure Ana, hammer her with real questions. Value immediately or it doesn't work. Find us here: textql.com/request-demo 🎯

What's Next 🚀 We're exploring agentic workflows where Ana handles entire analysis projects. Not just "what happened?" but "what should we do about it?" with scenario modeling, forecasting, and recommendations

Why This Matters NOW 🎯 Three things aligned: LLMs got good at code, data stack standardized (Snowflake/BigQuery + dbt), and companies realized self-serve analytics failed. The average marketer doesn't want to learn SQL. Just let Ana do it:

Security Done Right🔒 Ana runs in YOUR VPC. Data never leaves your network. LLM calls through Azure OpenAI/AWS Bedrock with your keys. We don't see, store, or train on your data. SOC2 Type II certified.

Tactical Reality 📈 Ana handles reactive analysis, releasing that 70% of your analysts' time spent on "quick questions." 🫠 Your team stops being a bottleneck. Business users stop waiting. 16 days to 4 hours average time-to-insight

How Ana Works 🌟 Ana connects to everything: warehouses, semantic layers, BI tools, that Google Sheet your CFO loves. She forms hypotheses, tests them, runs stats, creates visuals... Get complex insights in minutes, not weeks ⚡

The Spark 🧨 We watched customers with $500K Snowflake bills still take 3 days to answer "why did revenue dip?". We started building text-to-SQL, but quickly realized the best analysts are detectives, not just SQL writers 🔍

The Waiting Room 🫩 Every data team's dirty secret: a graveyard of Slack messages starting with "quick question about our numbers." By the time analysis arrives, campaigns are spent and quarters are over💀

You'll never have to write SQL again. We built Ana to: 📊 Connect to ANY data warehouse 🧑‍💻Write SQL and Python autonomously 🔐Fully adhere to Enterprise security (SOC2, VPC, etc.) ⚡️Get set up in 10 minutes Here's Matt and Ben walking through Ana's capability

🚀Shipping fast but drowning in SQL? Plug Ana (our AI data-scientist) straight into your Supabase in 60 sec: • Ask in plain English → get charts, forecasts, action items • Auto-generated docs & playbooks land in Slack/email • Free founder tier so you can keep vibe-coding

we have 0 marketing budget so we asked one of our engineers to make an ad. What is this 😭 Did he pop off?

.@sama openai did such a great job with o3

We're proud to announce that @TextQL has achieved SOC 2 Type I compliance

Do you find yourself searching for definitions of key internal terms like "CAC" or internal processes like customer onboarding, but it simply takes too long to find your answer? Now you can use TextQL to surface your Confluence or Google Docs and instantly find what you're looking for!

Did you know TextQL can integrate with your BI tools? If you have a question that can be answered by a dashboard, TextQL will surface that visualization. Here's what it looks like with Tableau. youtu.be/CM4ShutKB6c?feature…

Thanks to all your votes we’re currently top 9 on @ProductHunt ...but will we be able to surpass the next 2 competitors? The Product Hunt upvote race is on: producthunt.com/posts/ana-by…

✋Are your marketing campaigns reaching the right customer segments? Find out now at ana.textql.com.

🛑✋Don’t spend another minute on marketing analytics: Just ask your Google Analytics Data. @TextQL's Ana lets you ask data questions naturally, without using computer code. Simply ask “How do different customer segments respond to our marketing campaigns?” Game changing. Try it free on any .csv now: ana.textql.com

Just got @OpenAI plugins access; get ready to query your data straight from ChatGPT 👀

🚨 Using AI can make your data engineers 10x faster. Use our OpenAI powered data/dbt documentation generator (worth $250) for free for the next 24 hours! Like, Retweet, and Follow and I'll DM it to you for free

Proud of our CTO @amrkaahy winning the Best Use of Spellbook award at the @scale_AI Hackathon this weekend! The potential to use AI to fix errors in AI generated code unlocked a whole new frontier of accuracy