You.com

3x fewer tokens. 84% accuracy on the Vertical RTK benchmark. That’s what we got using Jev from TypeSafe AI via OpenRouter as a reranking layer for You.com Knowledge, which we recently launched in the Web Search API. Knowledge returns structured, real-time data from authoritative sources across domains like stocks, sports, and more. For many queries, that data answers the question directly, so much of the retrieved web context is unnecessary for the synthesis model. Jev sits between retrieval and synthesis, selecting only the most relevant results to pass downstream. We're walking through the whole build in a live workshop, if token cost is a line item in your search or agent pipeline, come build it with us.

Introducing Knowledge: Licensed real-time data, now available in the You.com web search API. Prices, rates, odds, weather, scores, government spending. The data that lives behind licensed feeds and structured databases, instead of on crawlable pages. Here's the question that convinced us to build it. How much did NASA pay out in federal contract outlays in FY2025? Every major web search API we tested, ours included, returned something different, from $19.8 billion to $28.8 billion. The answer is $20,002,053,117. Nothing was hallucinated. Every one of those is a real NASA figure from a real page. Some are a different line item entirely, and the closest is off by two hundred million dollars. A page carrying a dozen spending figures leaves the model to work out which one you asked for. So Knowledge reads the record instead. Licensed feeds are queried in parallel with the web index, and every figure comes back with its field name, its source, and an as_of timestamp. Across 155 questions like the NASA one, our accuracy goes from 64.5% to 84.2%, adding only 140ms of latency One parameter. Same endpoint. Same $5 CPM.

Research. Pay. Trade. Your agent can now do all three in one loop. We teamed up with Coinbase for their agent launch. A few minutes before the market opens, your agent pulls the latest financial news from You.com, pays for deeper data with x402, and trades on what it just found. Fresh, sourced information every morning, on repeat.

That's a wrap on Build with YOU: The Live Web Agent Hackathon, our first in New York 🗽 We brought together a great group of AI builders, founders, and engineers and asked them to make agents that catches its own failures and recovers, or one that measurably improves the more it runs. What came back was better than we expected. 🥇 LaunchGuard by Belal E.: AI launch control system using You.com Search, Research & Finance Research APIs to evaluate market demand, churn risk, and release readiness — resulting in a simple ship, hold, or revise decision. 🥈 Cleanroom by Shreyas Sreenivas: AI system that learns how to extract structured data from the live web instead of relying on brittle, hand-built scrapers. 🥉 F You Money by Forrest Pan & team: Agent team helping people find and claim money they’re already owed — from settlements and rebates to refunds and unclaimed property. Different ideas. Different approaches. Same goal: build agents that actually do something. Huge thanks to everyone who came out and built with us, and to our partners One, CrewAI, Daytona, Clean Data Alliance and Founders Bay for helping make it happen. Until the next build.

We’re excited to welcome You.com as a sponsor of The AI Conference 2026! You.com provides real-time web intelligence for AI agents and LLMs, giving developers access to APIs for web search, content retrieval, research, and citation-backed answers. Their infrastructure helps AI systems access accurate, up-to-date information from the web so they can reason and act with better context.

Curious how to make 'Slackbot' have up-to-date info on what happened this morning? Check out this session at Dreamforce featuring Saurabh Sharma and Gabe Kalal on how Slackbot got smarter with You.com powered real-time web intelligence. Worth a watch if you're thinking about what real-time web intelligence can unlock for AI agents.

We're partnering with Baseten to make the case for open-weight models. Paired with You.com web search, they become a credible, cost-effective alternative to proprietary models. Braintrust's independent evaluation backs the reason why: the web search layer predicts accuracy far more than which model you pick. Most teams reach for the biggest closed model when the actual problem is stale information. Braintrust tested it on 1,329 current-events questions. Without web search the four models ranged from 2.80% to 50.71%. With the same You.com results, 79.20% to 84.82%. That's the whole bet. The cost gap did not collapse with it. DeepSeek V4 Flash on Baseten answered 80.66% correctly at $0.0163 per question. Claude Sonnet 5 answered 79.23% at $0.0747. Roughly the same accuracy, 4.6x the price. Or put another way: which web search setup you use explained 0.0263 of the accuracy score. Which model you pick explained 0.0006. Today Baseten is launching Hosted Tools in preview, with You.com web search live at launch. If you're already running a model on Baseten, web search runs inside the inference path instead of in your app. The only change to your existing Chat Completions call is one line: tools=[{"type": "baseten__you__search"}] No separate vendor integration, no round trip per query, no resending the full conversation each turn. Fewer moving parts, up-to-date answers, open-weight economics.

You can now build a research agent in n8n without writing a polling loop. Until this release, running a frontier research query from n8n meant hand-building the plumbing: an HTTP Request node to start the job, a Wait node, an IF branch checking a status string you had to dig out of the docs, and a loop back around. Miss the timing and the workflow handed the next node a result that did not exist yet. The workflow now: 1. Research node in background mode returns a task id immediately. 2. Get Research Task polls until the status comes back "completed". What this unlocks: → Long research jobs stop dying at the execution window. Frontier effort runs to completion. → Your AI Agent picks the tool. All 7 operations run as agent sub-nodes. → Structured output. Pass an output_schema, insert straight into a database. Type You.com in the n8n node panel to begin.

It's show week. Here's where to find us at Dreamforce 2026: Booth 403, all week, with personalized live demos, the slot car racetrack, and the Dreamforce treat trailer. Come for the demo, stay for the lap time, leave with a snack. Our CPO Saurabh Sharma is also speaking twice this week on how real-time web intelligence changes what agents can actually do. Come say hi, and bring the question nobody has answered for you yet.

We’re bringing Daytona AI Builders to San Jose at SambaNova’s new office! 🚀 Join us for a special side event during WeAreDevelopers World Congress North America, dedicated to all things AI Engineering. 🎤 We have a great lineup of speakers: Vasanth Mohan — SambaNova Sako M — You.com Varshika Gambhir — Google Vashishtha Patil — Amazon Chiru B — SurrealDB Muhammad Annas Hashmi — Daytona Event partners: SambaNova, You.com & WeAreDevelopers RSVP → https://lnkd.in/ePSg8WpZ