bring spice

Problems with your text-embedding models? Filip explains the common issues with the traditional approach to search + embeddings. Superlinked has a smarter approach, using a MIXTURE of embeddings instead. Check out the video to find out more.

I put troll back into petroleum..

So what do we mean by "object level embeddings" exactly? Hamel Husain pulling it out of me bit by bit, thank you for the patience sir!

making progress on the blue app

"Custom embedding" doesn't mean a text model fine-tune but rather 👇

Let’s go!

lu.ma/4ykyhl1y let's get some tacos this Thu? earlier that day: genaisummit.ai/agenda/4718874

Please STOP modeling smooth preferences with 💩FILTERS💩

Any BOOST in ES/Open Search can be done better vectors.. see why👇

1. Pull 100 results from your database. 2. Reshuffle them, throw 90 away and use the other 10, burning compute & latency. 3. ??? 4. Profit! It's easy to slap a re-ranker on top of your IR stack.. but SHOULD YOU?! What about just pulling the 10 good results straight out of your

"BigAI" lobby wants you to think "text+image->vector" is the only way.. But you have products to sell, users to understand and jira issues to attribute. These are DATA OBJECTS, not individual pieces of strings and text. The sooner you understand that you need vectors that

Don't train encoders from scratch, build them from LEGOs 👇

Got 20 signups on the other social app in the first hour, let’s cross post on X and get ONE more :-D —— I’m hosting a free session for anyone dealing with vector search, cost vs. latency tradeoffs, or embedding model questions. Think of it as office hours with practical advice

If you don’t make a choice, you are choosing confusion or whatever the default is for everybody else.

What are you shipping this weekend?

🇺🇸

So there is now an oura ring for your “other finger”…