Search and inference for AI Get started → https://nitter.cf/t.co/xENay4qnSt
San Francisco, US
Joined May 2021
- Tweets4K
- Following38
- Followers24.3K
- Likes4.1K
Pinned Tweet
Introducing Inference Router: enterprise-grade inference on demand.
Kimi K3, GLM 5.3, DeepSeek V4.1 Flash, Qwen3.8 Flash, Muse Spark 1.3, Mercury 3 and more, available with an uptime SLA of at least 99.9%. US hosting and web search built in.
you.com/router
You.com retweeted
We’re hosting a one-of-a-kind hackathon and have partnered with @emergentlabs and @youdotcom 🤝
Speech to Ship is where you build an app using only your voice 🎙️
Join us:
> tons of swag
> limited spots
> $5k+ in credits & prizes
Clear your calendar on Nov 20th and join us at Emergent HQ in SF (link below).
New: earnings data in the @youdotcom Web Search API.
Quarterly net income as reported, from @Fiscal_ai. Same $5 CPM.
Try it in the playground today: you.com/platform/playground/…
You.com retweeted
Three winners at Hack With YOU: Halloween Edition, all solo builders. None of their agents stopped at an answer:
→ a call sheet posted to a film crew's Slack
→ a real phone paged through PagerDuty
→ a work order filed in Linear, only after a person approved it
On Oct 28, builders spent a full day at the AWS Builder Loft in San Francisco making AI agents with @youdotcom, @crewAIInc, @parasail_io and One. Our Head of GTM, Jacob Rissman, was on the judging panel.
🥇 Ghost Grip by Cody Wang
Give it a film script and it plans every shot for a robotic camera rig. CrewAI revises the plan until each move passes a physics check, @youdotcom pulls real film references, and Parasail runs the model. Then One posts the final call sheet to the crew's @SlackHQ.
🥈 Haunted Command Station by Jaroslav Hejlek
It reads a messy incident report, decides whether the threat is real, and builds a command station live in the browser. Press "call expert" and, through One, it opens a real @pagerduty incident, pages a phone, and speaks the alert with @ElevenLabs.
🥉 Clearbell by @SidraMiconi
Night-shift safety for care homes. It keeps conflicting handover notes word for word, lists the checks nobody has done yet, and uses @youdotcom to find real FDA recall records. Only after the duty administrator approves does One file the work order in @linear. Press it twice and the second press is blocked.
That last detail is the whole point. Agents earn trust when they act with access someone scoped, on an approval someone gave, with a record of what happened.
The winners take home a year, six months and three months of One Pro, and every builder got their first month free.
Thank you to @youdotcom for hosting, and to CrewAI, Parasail, @awscloud, UpScaleX and Founders Bay for making the day happen. Congrats Cody, Jaroslav and Sidra.
Kimi K3.1 Pro is live on Inference Router from day 0.
K3.1 is @Kimi_Moonshot's most capable open-weight model yet, with a 1M context window and a reported 65.7% on Terminal-Bench 4.0.
Run it in production with at least 99.9% uptime SLA, US hosting and ZDR available. No upfront commitment.
Excited to partner with @FireworksAI_HQ to deliver real-time web search for open models across its Frontier Ecosystem.
Point any model at the @youdotcom MCP server, and it searches the web whenever needed.
fireworks.ai/partners
The future of AI is specialized intelligence with models tuned to your data, your domain and your customers.
Today, we're proud to launch live from Forge the Fireworks Frontier Ecosystem with 60+ partners across the AI stack, helping you build and ship faster.
Check it out: fireworks.ai/partners
Beam by @reflection_ai is now available on Inference Router.
It's a US-built open-weight model for coding and agentic tasks. Teams can run it in production today, with ZDR available on every call.
you.com/router
Try Beam in the playground or talk to an engineer.
you.com/product/inference-ro…
You.com retweeted
Congrats to the winners of our 10/17 Recursive Agents Hackathon in SF, and thanks to @cerebral_valley, @nvidia and @youdotcom for helping make it happen!
📸 Attendees built agents that can take a goal and run with it on Crusoe Managed Inference.
You.com retweeted
@saahil9jain: search has to be rebuilt for agents - agents will run more searches than humans for the foreseeable future
BrightEdge's April data already put AI agent requests at 88% of human organic search activity - and it's vastly increased since then.
Saahil is CTO of @youdotcom, which he joined as a founding engineer in 2020, back when it was a consumer search engine competing with Google - now, they're helping drive the agent era of search. We got into:
- why agents care more about recall than ranking
- the life of a query, from fan-out to the index to re-ranking
- why the model is eating the harness, and how open-model routing brings it back
- why static benchmarks get memorized, and why customer evals matter more
& more
Chapters:
(0:00) Agents run a thousand searches at once
(0:25) What survives as AI models get smarter
(1:48) Is this the singularity?
(4:27) Core intelligence versus augmented intelligence
(6:33) Self-driving cars before the roads
(7:27) What belongs inside the model
(11:00) Where verifiable rewards run out
(12:16) Can agents form relationships?
(16:58) What model labs own, and what independents can build
(20:21) What a harness is
(22:39) The model is eating the harness
(23:14) Open models and routing revive the harness
(25:32) Open versus closed: what the labs keep
(31:30) OpenAI, Baseten and the enterprise model budget
(33:53) Co-evolving a search tool with the model labs
(37:42) How agent search differs from human search
(42:43) Life of a query, from fan-out to index
(47:22) Measuring retrieval by whether the agent answered
(50:21) Static benchmarks, freshness and customer evals
(54:19) Why he changed his mind on harnesses
Brought to you by:
• @WalrusProtocol - Walrus Memory gives AI agents portable, verifiable memory that carries context across apps, sessions, and other agents: walrus.xyz/cot
• @coralbricks_ai - CoralBricks serves open models through one API, fast and at low cost: DeepSeek V4.1 Flash streams at over 300 tokens a second. $5 in credits on signup, plus $5 on your first purchase: coralbricks.ai/r/Z6V2RE
• @inngest - durable execution for agents in production. Failed steps retry; completed steps are saved and skipped: inngest.link/cot-x
• @wisdomai_inc - in WisdomAI's latest benchmark, Claude Code alone topped out at 73% accuracy; WisdomAI delivered accurate answers 3x faster at about a third of the cost: wisdom.ai/benchmark
Look up the @chain_ofthought podcast on YouTube, Spotify, Apple Podcasts (links below), or wherever you listen. Thanks, Saahil!
Swapping models won't fix your accuracy if search is the bottleneck.
Your evals are measuring the retrieval layer as much as the model.
@zairah___ on running evals you can trust, from @AIconference.
Midterms are tomorrow. Your model's training data doesn't know that.
So we asked our Web Search API with Knowledge.
See the latest odds below, or run the query yourself👇
you.com/platform/playground/…
We're partnering with @FireworksAI_HQ to give open models the live web.
Kimi, DeepSeek, GLM, Qwen, MiniMax, gpt-oss: point any of them at the @youdotcom MCP server and they search when needed.
No key to start. 100 free queries a day.
Introducing web analytics data via Knowledge.
Browser, OS, search engine, device market share and more from StatCounter, updated monthly. Now live in the @youdotcom Web Search API. Same $5 CPM.
Try it: you.com/platform/playground/…
You.com retweeted
Reinforcement learning moves fast. But sometimes, training and rollout serving don’t.
Join @youdotcom, @NVIDIA, and CoreWeave on November 19th for our webinar Closing the Loop Between Inference and Post-Training. Register here: crwv.co/uuezt
Most agentic search failures are behavior problems, not knowledge problems.
Death loops. Weak tool use. Finding the answer, then doubting it.
@saahil9jain and David Li breaks down on how @youdotcom, @nvidia and @CoreWeave fixed that in Nemotron 3.5 Lightning in 3 weeks.
