You.com

A major tech company burned through its entire 2026 AI coding budget in four months with 5,000 engineers—some running up $2,000/month in tokens. Then, leadership was asked if any of it was actually improving the product. The COO's answer: "That link is not there yet, right?" That's an architecture problem dressed like a budget problem. 🤡 Token waste often isn't from usage—it's baked into how agentic systems are built. Every turn re-sends the full trajectory. Instead of claiming the fix is to "use AI less," try some architectural fixes. — Simple tool abstractions instead of parameter-heavy schemas — Dense payloads instead of raw HTML/JSON dumps — Compact link IDs instead of 150-character URLs repeated every turn — Budget as a first-class signal the model can see, not an afterthought in billing Are you measuring cost per task? What about cost per business outcome? Read more at the link in the comments. 👇

Excited to see the innovation at Nimble. Can't wait to see where they go from here.

What do an Engineering Manager from Meta, a Head of Product from Composio, an AI Research PM from Stanford University, and investors from Cota Capital, Mavka Capital, Future Energy Ventures, Entelechy Ventures, and Seventrain Ventures all have in common? They're on the judging panel for tomorrow's AI Builders Pitch Night at Brex in San Francisco. 🔥 We're bringing together people who build and people invest to answer one question: does this hold up outside of a demo? Founders pitch live and get direct feedback. The top three walk away with a Founders Bay membership and admission into the You.com Startup Program, including $1,000 in platform credits. 7/30 | SF Sign up in the comments 👇

HAPPENING TODAY: You.com is hosting an AI Breakfast at Tesla in San Francisco. Join founders, operators, and AI builders and start your day talking about what's actually breaking when you put agents into production. The panel will cover infrastructure choices, evaluation frameworks, and the gap between "it works in the sandbox" and "it works at 2am when the model vendor changes something." Coffee and breakfast first, panel at 11, then time in the Tesla showroom. It's a small group—practitioners talking to practitioners. If you're building with AI in production and want to compare notes with people solving the same problems, join us today at 10AM. Register in the comments 👇

Sales teams keep trying to bolt AI onto a process that was built for a world where buyers had way less information and way more time. As a result, we're seeing a lot of GTM AI initiatives fail because nobody rethought the workflow underneath it. Sound familiar? Luckily today (10am PT/1pm ET) Seyar Karimi, Enterprise AE at You.com, is running a workshop on exactly this. He's exploring what modern selling actually looks like when buyer behavior and sales automation are both moving at once. He'll get into: → How AI is resetting buyer expectations and cycle length → Where automation helps prospecting—and where it doesn't → Common mistakes teams make rolling out AI in sales → Why human connection still decides the close If you're building your first sales motion or trying to keep your team's process from getting automated into mush, don't miss out. Live Q&A at the end. Register the comments 👇

Your terminal agent's ability to know about this morning's CVE depends on which model you picked. That's the true fragility inside "AI agents with web access." To make agentic search more reliable, we shipped the You.com MCP Server into Warp's agent stack: you-search, you-contents, and you-research, plus an opt-in you-finance tool, over standard MCP. It works with any model Warp supports because it's a tool call instead of a model feature. You can scope tools per Agent Profile. A "quick lookup" profile only sees you-search while a "deep research" profile gets the full set. That's the difference between checking a version number and burning a multi-minute synthesis pass on it. Writeup in the comments 👇

You.com is now a one-click MCP server in Replit's Agent integrations list. Add it to a project and Agent starts pulling real-time, cited web search and research into whatever you're building, without you standing up a separate grounding layer or babysitting an API key. Click the badge in the Integrations pane and the app you're shipping can answer questions about this week, instead of only relying on its training data. For anything touching current events, pricing, or other time-sensitive info, that's the gap between a demo that impresses and a product that holds up. MCP list in the comments 👇

Reporting live from the field! The AWS Builder Loft in SF is wall-to-wall laptops right now, with teams racing the clock across three tracks—real-time intelligence, deep research, and multi-agent systems. 🛠️ Half these teams are wiring the Deep Research API into agent loops that used to take a human three hours of tab-switching, and testing. The real test today is whether or not the build survives contact with information that isn't static, served through the You.com Web Search APIs. Judging tonight—excited for what's to come.