MotherDuck

The data warehouse built for getting answers from your data. Works with AI agents and SQL. Built in collab with @ducklabs_com

[MotherDuck @motherduck]

We used AI agents to build a semantic layer. The surprise was that it became disposable.
Q&A pairs proved unreasonably effective at capturing business intent and steering agents.
If code can be regenerated, where should business knowledge live?
https://t.co/MIPAUk6SPG

[MotherDuck @motherduck]

The data stack in an AI world looks nothing like it did two years ago.

The hard part of shipping AI on your own data is everything underneath the model.

Data Outpost is your chance to catch up and get ahead. Join us in SF for workshops (Nov 4) and talks (Nov 5) that focus on the new data infrastructure required for today's world.

Meet the conference sponsors building the future of the AI data stack:
• @AirbyteHQ
• @braintrust
• @EstuaryDev
• @_hex_tech
• @reductoai
• @sigmacomputing

Early bird pricing is live until September 2nd. Don't wait: https://t.co/A3FG8XMRZn

[MotherDuck @motherduck]

What are the chances your next flight leaves on time? You'll need a lot of data for that... Alex on our team gave @replit 3 prompts: its agent signed itself up for MotherDuck, loaded >30 million rows, and shipped a live app.

Try it out & build your own! https://t.co/WsK258zvti

[MotherDuck @motherduck]

Beyond Copilots: AI That Builds the Whole Data Stack https://t.co/LMO23e9Ncd

[MotherDuck @motherduck]

Return to a world where data lakes never get used properly. Or, take the red pill, and find out how deep the DuckLake goes.

Chapters 1 - 4 are available for free download: https://t.co/hj1KYYH1ZH

[MotherDuck @motherduck]

Async I/O lands in v2.0 this fall.

Separate ASYNC thread pools and a read-ahead queue took TPC-H Q6 on S3 from 8.23s to 2.84s, and one CSV query from 878s to 45s.

[MotherDuck @motherduck]

DuckDB Monthly #44 is out. 10 links, and a community spotlight worth your time.

https://t.co/gpxdxA525J

[MotherDuck @motherduck]

The new ADBC extension turns one DuckDB connection into a door to 30+ external databases.

Snowflake, Databricks, Postgres, all over Arrow with zero-copy transfers.

[MotherDuck @motherduck]

Flights just landed in Asia Pacific (lol). You can now run your data pipelines in AWS Sydney (ap-southeast-2) and Tokyo (ap-northeast-1), keeping pipelines close to where your data actually lives.

Flights let you run any pip-installable Python in a built-in runtime on MotherDuck. Sync from a source system to your MotherDuck database, build an automated analytics agent, and much more. Check out our cookbook for examples.

Get started with Flights: https://t.co/Tw4dOr2rof

[MotherDuck @motherduck]

Introducing Guides: Warehouse-Native Context for AI Agents https://t.co/NslUgyx6wM

[MotherDuck @motherduck]

"Everyone's an admin" stops being funny somewhere around employee 20. In case you missed it, MotherDuck now has role-based access controls. Admins assign roles that decide what each user can do across the platform and which shares they can read.

Default roles include Admin, Builder, and Explorer–define custom roles to tailor access controls even further.

Details in the docs: https://t.co/qpioNWXd5W

[MotherDuck @motherduck]

We're scaling a ton, so is our GTM team. Come join us in SF on August 25th- whether you want a MotherDucking drink, to meet some of the sales team, or to chat with people in the industry about how to navigate the fast-moving landscape of GTM in an AI world, come quack. Register here: https://t.co/9Ajvg8vasM

[Alex Monahan @__AlexMonahan__]

Hot take from @Ai4Conferences last week: way too many companies building "enterprise context for agents." Context matters, but there isn't room for 100s of businesses whose whole product is context engineering.

Modern Data Stack era all over again.

Gimme those AI hot takes!!

[MotherDuck @motherduck]

Thousands of apps get vibe coded every day. Almost none reach prod.

Vercel V0 & MotherDuck collapse the gap to 10 prompts. My agent signed up for MotherDuck itself, loaded the "don't die" guy's health data, built the app, and deployed.

Full example app: https://t.co/etTTITbhLe

[MotherDuck @motherduck]

be honest: how much of the SQL your team shipped last month was AI-written?

now the real question— is that number high or low? nobody knows. we're finding out, by team size. answer anonymously, see everyone else's number first: https://t.co/vWG2MKv8pN

[MotherDuck @motherduck]

Your agent can't know that subscriptions_v1 stopped updating in March after the Stripe migration. Nothing in the schema says so.

Guides were meant for this. Here's a complete one–just a few lines of discoverable context that lives in the warehouse

Guides are available now, making every agent query smarter. Start here: https://t.co/KuN620dnJQ

[MotherDuck @motherduck]

LLMs have only been reliably good at SQL for 6–9 months.

Everyone's asking why agents haven't hit data yet. That's the answer — it's not late, it's early.

The infra question is the interesting one: agents go from zero to a flood of queries instantly. Warehouses were built for people who query on a schedule.

https://t.co/F7G5z9AOBm

[MotherDuck @motherduck]

Watch out what version AI is recommending you, and that's of course not only true for Python but any other active programming language!
Watch the full episodes of Explain ANALYZE pod : https://t.co/5XPWoGeJfC