Evan Rimer

FDE @tavilyai @twistpartygame 🐧

[Rotem Weiss @rweiss57]

I agree with @paraga. It gets even more interesting in post-training.
Companies are increasingly building closed systems with routers: the hardest tasks go to frontier models, while more specialized tasks go to smaller open-source experts that can be continuously trained on production trajectories. Over time, more traffic shifts to these increasingly capable experts.
That means agentic search is expanding from inference time into pre- and post-training as well. Models need to learn when and how to use search (and other tools) for specific tasks. A research agent may want a broad first response that shortens the path to a report; a voice assistant wants something fast and direct.
Web search will increasingly need to adapt to these post-training use cases. For us, that means making results predictable, so models can train against a stable interface, and dynamically tunable to the model, task, and point in the workflow.

[Assaf Elovic @assaf_elovic]

Didn't expect this 🤯

We replaced embeddings with Jev in GPT Researcher's RAG pipeline and tested both on 28 research tasks from SimpleQA and open ended research.

Jev beat embeddings on every quality measure we ran:
- 59% more relevant context (73% vs 46%)
- Reports preferred 15 to 3 in blind comparisons
- Same cost per report

GPT Researcher now runs on Jev by default, and no longer needs embeddings at all.

All you need is @LangChain + @tavilyai +Jev for the perfect RAG system.

Check out the repo here: https://t.co/kcvF5tPlWq
Research: https://t.co/QZIoRFnGQq

[Evan Rimer @EvanRimer]

@DanielFluxman Ask your muse to make your muse avatar look sicker

[Evan Rimer @EvanRimer]

@lakshyaag @deansacoransky @thsottiaux Need better loop engineering /goal

[lakshya @lakshyaag]

$20K grand prize, 4 tracks, already at 7.7K participants

sign up and start building - excited to see!

https://t.co/DnIQdUJ3rq

[Tavily @tavilyai]

We sat down with Shriram Sridharan, Co-founder & CTO at @rox_ai, to hear what it took to add the right web search provider to the GTM agents their customers depend on. The Rox team found with their original provider, customers weren’t getting up-to-date results, and costs were moving up.

That’s when Rox turned to Tavily.

Hear how Rox stood up an infrastructure in two weeks that now tracks millions of accounts and billions of contacts, cutting a sales rep's research time from weeks to running itself overnight: https://t.co/rctdxFGvKw

[Ricky Johnson @tavily_rick]

Searching https://t.co/0gSuWzO6DR with @Meta's Muse for dairy/soy free snacks. 25 minutes and DNF. So I added a @tavilyai connector. Same task took 4 minutes

[OpenAI @OpenAI]

This is GPT-6 Astra.

Anything you can do on a computer, Astra can do for you. Fast.

[Evan Rimer @EvanRimer]

@lakshyaag @OpenAI @charliermarsh @OpenAI can I have early access?

[Tavily @tavilyai]

Because search quality is often the limiting factor for AI agents, we spent the past several months rebuilding core parts of Tavily /search, including evidence ranking, contradiction handling, and index coverage. 

Using the same retrieval budget, reader model, and grading pipeline across providers, 𝗧𝗮𝘃𝗶𝗹𝘆 𝗶𝘀 𝗻𝗼𝘄 𝗡𝗼. 𝟭 𝗶𝗻 𝗮𝗰𝗰𝘂𝗿𝗮𝗰𝘆 𝗼𝗻 𝗦𝗲𝗮𝗹𝗤𝗔-𝗵𝗮𝗿𝗱, 𝗦𝗲𝗮𝗹𝗤𝗔-𝟬, 𝗮𝗻𝗱 𝗦𝗶𝗺𝗽𝗹𝗲𝗤𝗔 𝗩𝗲𝗿𝗶𝗳𝗶𝗲𝗱. 

See how and why these improvements help your agents retrieve better evidence and produce more accurate answers: https://t.co/ZcO7DPyVdf

[Nebius Token Factory @nebiustf]

Kimi K3 is now available on Token Factory.

We’re excited to announce that Nebius Token Factory is an official Day 0 partner for @Kimi_Moonshot's Kimi K3.

Kimi K3 is the first open-weight model to reach frontier-level performance, a major step forward for open models.

It is built for long-horizon coding, knowledge work and reasoning, with native vision and up to 1M tokens of context. Artificial Analysis scores it at 57 on its Intelligence Index, just two points behind GPT-5.6 Sol (max). That puts Kimi K3 at the top of the open-weight field and firmly among today’s frontier models.

Developers can access K3 through Token Factory’s OpenAI-compatible API and console today.

Give K3 the hard problem.

Build with Kimi K3: https://t.co/olJZmAvmDh

[Evan Rimer @EvanRimer]

@SpaceXAI 🚀

[SpaceXAI @SpaceXAI]

Try the Tavily plugin for LLM-driven search inside Grok Build

[Evan Rimer @EvanRimer]

@lakshyaag words of wisdom 🚀

[Roman Chernin @romanchernin]

Excited to see @nebiustf and @tavilyai at the slide
We've just started building

[Tavily @tavilyai]

We’re back at LangChain Interrupt as a sponsor for year two and things kick off today:

❓ AMA with @EvanRimer from Tavily and Sujee Maniyam at 3:50 PM
💻 Live demos of Tavily Deep Research + a competitive intelligence agent built with LangChain Fleet, Tavily Search, and Nebius inference
🎁 Swag, giveaways, and a few surprises

If you’re around, come hang with the Tavily team at booth 4, let us know what you’ve been building, and follow along here on X for updates throughout the event.

[Evan Rimer @EvanRimer]

@romanchernin 🚀

[dylan ツ @demian_ai]

Ich bin ein Nebiuser 🇩🇪

After a week in Amsterdam, onto Berlin (the EU is booming 🇪🇺)

On tuesday we brought ~100 ML engineers, platform architects, and technical founders together for Nebius Build/BER, a small technical gathering for people building in AI.

What I didn't expect was the range. Someone training LLMs on Nepali. Someone building AI tooling for city governments. Europe has a different kind of builder culture. More grounded. More "does this actually work for real people" and less about the valuation. Always fascinating to see how the crowd shifts city to city, and Berlin had its own flavor of sharp. The kind of event where the Q&A runs long because the questions are actually good.

Thanks to everyone who came out to Impact Hub Berlin, and to our partners Anyscale, Tavily, NVIDIA, and Black Forest Labs for going deep with us.

Berlin, you delivered. More cities coming 👀

[Evan Rimer @EvanRimer]

Berlin showed up with energy!

Excited to see what everyone builds with @tavilyai