Every Continue user is a LanceDB user.

Thanks to Ayush Chaurasia and Nate Sesti for making this blog post finally happen!

Existing closed-source API-based LLM developer tools are opaque and the developers deserve transparency, hackability, and control over their tools.

Learn how Continue's open-source approach, powered by LanceDB, is changing the game for LLM-powered dev tools ⚒️🚀

Read the

I heard attendees get $250 free credits for YAPPING Cloud Dr. Donut ☕️

Time to put on the convergence trade? Nuclear powers don’t default right?

OpenAI launched GPT-4o last week with six benchmarks on which they have performed comparisons with GPT4 Turbo, Gemini Pro 1.5, and Claude-3 Opus.

Here I have Benchmarked GPT-4o on NIAN(Needle in a Needlestack)

Blog - blog.lancedb.com/benchmarking-g…
LanceDB #openai
Other Benchmark👇

If you squint hard enough, Rust <> gymnastics starts to look like a Lisp variant

Doing #RAG ? Using Hugging Face ? We have some updates this week:

sf-hamilton==1.63.0:
* One Hugging Face data loader & two data savers [ #parquet , LanceDB]
* Example & blog post doing #NER with LanceDB using data & models from Hugging Face

I’m very excited about the TypeScript embedding function registry being baked in LanceDB . In python, this feature abstract away embedding generation, just declare the function as part of the table schema! Comments/requests are welcome!

github.com/lancedb/lanced…

Is oat milk ok? No, no it’s not

This is going to be 🐐 very excited about the conference

I re

R is short for Rust right? Not sure I went to the right conference.

For those of you asking: yes She is the correct spelling for my last name. Just remember: He is my pronoun. She is my proper noun.

Building LanceDB with Lei Xu has been such a fun journey. It's about to get even awesome-r with our new funding round led by CRV. Huge thanks to our customers and our community for their trust!

blog.lancedb.com/new-funding-an…

Key Findings:
• Dictionary encoding is effective for most data types.
• Simple encoding schemes are preferable.
• Don't use block compression (Snappy, zstd).
• Don't require deserializing all meta-data to read one column.
• GPUs struggle with Parquet/ORC.

Impeccable timing x.com/andy_pavlo/sta…

Great post from Weston Pace to continue the discussion from x.com/criccomini/sta…

Fun read if you like nerding out about data formats / performance: blog.lancedb.com/file-readers-i…

As an introvert prone to procrastination this is *litearally* The Dream

People have been trying to create a super app a la WeChat in the US market for years. Seems like OpenAI will become that super app but with way more capabilities?

That’s a bold move Cotton