The multimodal lakehouse for AI, accelerating large-scale data curation and feature engineering so teams can build better models faster.

San Francisco, CA
Pinned Tweet
1/9 The Reverie Summit 2026 agenda is live! What does it take to build world models, train robots, and search hundreds of billions of pages? Meet the teams building the data systems behind them. November 5 in San Francisco. reveriesummit.com/
3
6
24
2,472
1 week until #Reverie2026 ⏰ Speakers from @Nvidia, @runwayml, @zoox, @DynaRobotics, @LumaLabsAI, @AppliedInt, @ExaAILabs, and more! 1 day only. If you are building world models, robotics and autonomy systems, training infrastructure, or multimodal search, this is the room to be in. Nov 5, San Francisco. 1 week left to apply: na2.hubs.ly/H08Glpm0  𝘊𝘰𝘮𝘮𝘦𝘯𝘵 𝘣𝘦𝘭𝘰𝘸 𝘸𝘩𝘪𝘤𝘩 𝘴𝘦𝘴𝘴𝘪𝘰𝘯 𝘺𝘰𝘶'𝘳𝘦 𝘭𝘰𝘰𝘬𝘪𝘯𝘨 𝘧𝘰𝘳𝘸𝘢𝘳𝘥 𝘵𝘰 𝘵𝘩𝘦 𝘮𝘰𝘴𝘵! ⤵️
1
1
6
1,007
Changed one feature. Reran the whole pipeline. Sound familiar? @GM's Gokcen Iskender explains how feature-level caching avoids recomputing unchanged features for self-driving AI at #Reverie2026. Apply to attend → na2.hubs.ly/H08zQW20
1
1
446
Before blaming S3 for slow writes, benchmark your WAL. We used LAION to benchmark storing multimodal data with a WAL. Lance writes blobs directly using Lance Blob v2, so SSTable flushes contain only metadata. At concurrency 32, Lance delivered 11.2% more write throughput than SlateDB while persisting 48.2% fewer bytes on S3. na2.hubs.ly/H08wr9H0
4
2
26
2,985
Congrats @duckdb on 2.0.0 release 👏🦆
🦆 Announcing DuckDB 2.0.0 LTS “Cinnamon Teal” We are proud to release DuckDB 2.0.0, named after the cinnamon teal (Spatula cyanoptera), a striking reddish-brown duck found in the western Americas. 🌎 DuckDB 2.0.0 is the biggest DuckDB feature release so far. It ships a polished version of the Quack protocol, which turns DuckDB into a full-fledged server. It improves performance and lowers costs through supporting asynchronous I/O, a first-class VARIANT type, and a dedicated “agent mode”. It also empowers extension developers with a new SQL parser that allows adding new statements, clauses, and keywords, a new API to build on, and a new mechanism to distribute extensions more easily. In our announcement blog post, we summarize the highlights and then walk through the key changes and give guidance on how to migrate to 2.0. For the complete release notes, see the release page on GitHub⁠. duckdb.org/2026/10/21/announ… #DuckDB
7
637
Road logs. Simulation. Training. Evals. The arrows are doing a lot of work. @XPENG_Global, @AppliedInt, and @LanceDB unpack the autonomous driving data flywheel at #Reverie2026. Apply to attend → na2.hubs.ly/H08wd7C0
1
2
358
Good morning from Long Horizon! Swing by our booth, take a photo with our shield, and chat about accelerating your training at petabyte scale 💪
1
1
9
595
Find the rare driving scenarios. Put them to work in training. @XPENG_Global Bo Tian brings the data infrastructure behind VLA 2.0 to #Reverie2026, from searching billions of clips to evolving datasets. Apply to attend → reveriesummit.com/?utm_mediu…
1
2
4
4,818
Optimizers that scale. Models that teach each other. A decade of lessons learned the hard way. Rohan Anil @_arohan_, Gemini text co-lead turned Core Automation co-founder, is keynoting #Reverie2026. Bring your hardest problems.
Bit more than a decade in training, no pain no train. 💪 @BEBischof nerd sniped me into this talk with offer to have a Cat to give the keynote. Still Training: A Decade of Learning the Hard Way @ Reverie by LanceDB on Nov 5 in SF 🌉 You get to hear me about distributed training from the trenches: Hardware codesign, async and sync training -- when we almost decided to put async into chips From big batch gradient descent with parallel boosting from stochastic optimization with duals to online learning where we had regrets, to higher order methods that scale that was almost second order that nobody believed it.
1
6
528