We’re excited to share that Linkup has raised a $10M seed round led by Gradient, with support from leading investors and founders from Mistral, Datadog, Deel and Dataiku.

This funding will enable us to further develop our search infrastructure, grow our team of engineers and researchers, and build what we see as the essential rails for the emerging web of AI agents.

We’re also announcing today the launch of Linkup /fast, the world's most accurate sub-second web search API.

AI's Airplane Mode Problem 

AI systems today are remarkably capable, but they have a fundamental limitation: their knowledge is frozen at the moment their training ends. Ask an AI about yesterday's news, current product pricing, or a company's latest earnings, and you'll quickly hit a wall.

Think about using your phone with airplane mode permanently enabled. You'd have access to everything already downloaded, but nothing new, nothing current. That's essentially how AI systems operate without access to real-time web information. They're incredibly smart, but operating with outdated maps of a rapidly changing world.

AI has the potential to unlock transformative outcomes: accelerating research breakthroughs, driving productivity gains, and fueling business growth. But realizing this potential hinges entirely on one thing: access to accurate, timely data.

This limitation becomes critical as AI systems take on mission-critical workflows across finance, healthcare, enterprise operations, and beyond. In these contexts, mistakes can be extremely costly, lead to misleading conclusions, and create serious reputational damage.

Why Traditional Search Fails AI

The obvious solution might seem simple: just connect AI to existing search engines. But search engines built for humans and those built for AI need to work in fundamentally different ways.

Traditional search was built to return links. AI systems need context, raw information, and current data.

Search engines were designed for human behavior: you enter a query, get a list of relevant pages, click through, skim headings, and decide what's useful. AI systems don't need that. They need the information itself: extracted, structured, and ready to process. A link is useless to an AI; it needs the text, the data, the context behind that link.

But it's not just about format. Traditional search indexes what humans want to see: popular websites, engaging content, pages optimized to keep people clicking. AI systems need something different: trusted sources, verified facts, and accurate information, not content engineered for engagement metrics. And they can go beyond web pages and leverage data on web servers, accessing the structured information that sits behind interfaces - databases, APIs, real-time feeds - rather than being limited to what's rendered on a page.

Building Search for AIs

This realization led us to rebuild web search from scratch, designed natively for how AI systems actually process information.

Our founding team brought together unique perspectives on this challenge. Denis, our CTO, previously built one of the first neural music search engines, which was acquired by Spotify. Philippe, our CEO, had built graph-based data search systems at Lyft. Boris, our COO, was previously at McKinsey, where he led digital transformation in enterprise environments.

Together, we saw an opportunity to fundamentally reimagine how information from the web should be organized and accessed by AI systems.

Indexing Information Atoms from Across the Web

Our breakthrough came from changing the basic unit of what we index. Traditional search engines index entire web pages: documents that mix extensive data points and information covering a wide range of topics with navigation, advertisements, and design elements. We do something different.

• We extract what we call "atoms of information": individual facts and discrete pieces of knowledge from across the web. A competitor's recent product update. A company's latest revenue. Yesterday's regulation change. Each becomes an atom in our index, complete with its source, timestamp, and credibility markers.

• We then generate embeddings -mathematical representations- of these information atoms. These embeddings allow AI systems to quickly find the most relevant information for any query, not by matching keywords, but by understanding the actual meaning and context of what's being asked.

• When an AI queries our platform, it doesn't receive a list of websites to visit. It receives precisely the information it needs already extracted, verified, and structured for immediate use. The AI can access facts from hundreds of sources in milliseconds, building responses based on the most current information available.

  
  

This approach has allowed us to push the boundaries of what's possible in AI search and deliver Linkup /fast, the world's most accurate sub-second web search API for AI systems that can't compromise on latency or accuracy: 

Trusted by Global Leaders

Since launching our API just months ago, we've seen rapid adoption from thousands of users globally. Leading AI companies like Artisan use our infrastructure to power their AI agents. Global enterprises like KPMG rely on us to ensure their AI systems have access to current, accurate information.

This traction validates what we've believed from the start: as AI systems become more autonomous, making increasingly independent decisions, the quality of information they access becomes critical. An AI making investment recommendations needs accurate market data. Every query to our API represents an AI making a more informed decision.

This momentum is why we’re thrilled to have Gradient’s support. Led by former founders and technical experts, Gradient has been at the forefront of AI, supporting hundreds of founders from the start. Their technical DNA and early-stage focus make them the perfect partner as we build this new foundation for the web.

What's Next

It’s been an intense & crazy year. We've gone from an idea to building AI-native web search- now powering hundreds of companies globally. But what we're most proud of isn't the product we've built, it's the team we've built it with.

Building the core infrastructure for the future of AI is the kind of work that doesn't come around often. Getting to do it alongside exceptional people makes it even better.

If you've ever wondered what it would have been like to join Google in 1998, joining us in 2026 is your chance: we're hiring across the board in New York, San Francisco, and Paris to push the boundaries of our indexing technology even further.

👉 Join us

This is just the beginning. If you think what we've built so far is impressive, wait for the next 12 months: excited for what lies ahead.

The Linkup team