Really proud of this work, check it out if you're curious about the problems one typically encounters with shipping RAG solutions to production in an enterprise setting!

AI agents are only as useful as the tools they can reliably access, and the latest release of our Agents SDK makes it easy to connect to MCP servers. We've prepared a quick guide where we bootstrap a minimal MCP-powered chat agent with just a few lines of code: www.zeta-alpha.com/post/build-m...

Join us for the Zeta Alpha "Trends in AI" show on Friday, July 11th at 8 AM PST / 5 PM CEST live from LAB42 in Amsterdam, and online in San Francisco and around the globe. Register to receive your Zoom webinar invitation, or watch our live stream on YouTube and LinkedIn: lu.ma/trends-in-ai...

In our latest blog post, we covered how we use RAGElo, an open-source toolkit we've developed internally in Zeta Alpha, to compare multiple RAG systems head-to-head and aggregate pairwise preferences into a robust, easy-to-interpret Elo ranking. Check it out: www.zeta-alpha.com/post/robust-...

Missed last week's Trends in AI episode on evaluating AI systems? We've got you covered! Check out the highlights in our latest blog: www.zeta-alpha.com/post/trends-...

Join us next Friday!

📢 Zeta Alpha is going to OpenSearchCon Europe 2025 📢 We are excited to be contributing back to the OpenSearch community, sharing insights from our work in Enterprise AI, Neural Search, and Retrieval-Augmented Generation (RAG). Learn more about our talks at the conference below: 🧵

Join us for our monthly Trends in AI webinar on Thursday, April 3rd, at 8 AM PST / 5 PM CEST! This edition comes live from Hannover Messe, the world's leading industrial tech trade fair, and online in Amsterdam, San Francisco, and everywhere else in the world. Sign up on Luma with the link below:

A deep dive into the Gemini Embedding technical report - turns out you don't need any fancy training objectives or architectural tweaks to get SOTA on MMTEB, just a solid base model and high quality data!

Wrote a little blog piece on reasoning models as a summary of our latest webinar episode, feel free to share with your non-terminally online friends who want a gentle intro to catch up and what they mean for overall AI capabilities!

The whole world is buzzing with the performance of DeepSeek’s reasoning model, R1, and its implications for the dominance of Western tech companies in the AGI race. But what exactly are reasoning models? Did anyone see them coming? And how can they be used to create value?

Zeta Alpha recognized as a leading native AI enterprise search technology provider. www.zeta-alpha.com/post/zeta-al...

i hope 2025 is the year of good ui/ux for ai

friends don’t let friends build RAG pipelines without understanding IR theory. sounds like a solid christmas gift for someone you care about!

Drowning in the 4000 main track papers at #NeurIPS? Here’s a guide to 10 research areas and some of the spotlight papers that grabbed my... attention!

It's #NeurIPS week! With 4K+ main papers, it's OK to feel overwhelmed. Our annual NeurIPS guide is here to help! We've curated a list of 10 topics and papers to check out, whether you're attending the conference or want to stay updated on the biggest developments in AI.

wait what happened, why do i have apple intelligence on my mac... 👀

the real reason we don't have agi yet is that uv's documentation isn't in the llms' training data

"think harder" is going to become in 2025 what "shorter" has been in 2024

how about you o1-pro vide some benchmark numbers and real use cases

NeurIPS, the most prestigious conference in AI, kicks off next week! We will guide you through the top themes and highlights, tomorrow in our webinar, and next week with a detailed blog post and video. We picked 5 papers that we previously covered in our webinar which will be presented as orals: 🧵

Join us this Friday for the Trends in AI webinar, as we reflect on the breakthroughs that defined 2024 and see how our predictions held up. Plus a NeurIPS highlight reel, plus the trending AI papers of the month, like OpenScholar, M3DocRAG, DeMo, and more! lu.ma/oefbjkjl

spent the day reading the OpenScholar and Tülu 3 papers in detail. AI2 x UW has been on fire lately, this is what real “Open AI” looks like 😉. big thanks to both author teams for such thorough and well-crafted work.

sadly the unique content / news freshness on this platform fell off sharply after the hf dataset drama

happy 2 years of agi for those that celebrate

i guess elon was right about anti-censorship huh? imagine getting permabanned for posting a hf dataset

people have the chance to engrave their existence & persona into the memory of all future LLMs simply by poasting and instead they get *mad*?? posterity here i come

as sad as it is, today's meltdown over the use of an open API is probably a solid clue why the e/acc part of twitter hasn't moved here

qwenbros we are winning

may i interest you in HypeRAG, a novel methodology for RAGing on your RAG data

@​obsidian pros: i went down a rabbit hole during the weekend, so hit me with the tips you'd give to yourself if you were just getting started

one thing i enjoy on this platform is the absence of a****ic rag posts (and i know this won't last)

i had the misfortune of having to write langchain code today and i just want to punch a wall

yep, porn bots have arrived on the platform

who can cook up a "best from my followers" feed rq so i can check periodically and follow back the ones posting banger content?

really nice thread on data visualizations!

IR people: I'm trying to maintain a starter pack of other retrieval-pilled folks as they join this platform. if you're aware of others missing let me know! (& i wish there was a way to invite collaborators on starter packs & lists...) go.bsky.app/AZVJeDA

i’m not here because i’m upset with twitter or to hedge it’s decline or anything like that i’m here because i love the dopamine from digital interactions

I don't see how the "needle in the haystack" benchmark is an indicator of how models handle long context with *relevant* information (& why it's so popular). All it tests for is how well they can ignore noise and spot outliers, while most real use cases are about effectively using *all* context.