How is Seltz's search engine different from existing AI search products?

Seltz owns the entire search stack—web crawler, search index, retrieval models, and ranking—rather than wrapping Google, Bing, or Brave's APIs. It scores individual passages and extracts the specific table, text, or image an agent needs, rather than returning full pages or summaries.

Who is Seltz's founder and what is his background?

Antonio Mallia is the founder and CEO. His PhD work in computer science at New York University focused on information retrieval, and he previously worked as an applied scientist on Amazon's artificial general intelligence team and as a research scientist at the vector-database company Pinecone.

How many people work at Seltz?

Seltz currently employs 15 people, only a half dozen of whom are full-time employees. The team works fully remote, split between the San Francisco Bay Area and hubs near European universities in Pisa, Italy, and Leipzig, Germany.

Back to articlesLarge Language Models

Large Language Models

Seltz, an AI-native search startup, raises $12.5 million（約20億円） to build an alternative search engine designed for AI agents rather than human users.

Fortune AI4h ago6 min read

Key takeaway

Seltz, a startup building search infrastructure for AI agents, has raised $12.5 million（約20億円） in seed funding. Unlike most AI search products that rely on Google or other large companies' indexes, Seltz owns its entire search stack and is designed specifically for how AI agents query and consume information—targeting an area where even the largest AI labs have struggled to build truly independent alternatives.

Summaries like this, in your inbox every morning.

3 Key Points

What happened
Seltz announced $12.5 million（約20億円） in seed funding led by Speedinvest and B Capital, with participation from Italian Founders Fund, United Ventures, and Future Back Ventures. The startup, founded last October with 15 employees, is building its own search stack—including web crawler, index, retrieval models, and ranking—rather than relying on Google, Bing, or Brave APIs.
Why it matters
AI agents and chatbots issue long, precise queries and need machine-ready information from tables, images, and body text—not human-optimized snippets. Seltz's founder argues traditional search engines were built for keyword queries and link-clicking, making them poorly suited to AI workflows. Most AI search products today depend on large tech companies' search infrastructure, leaving room for a purpose-built alternative.
What to watch
Seltz's system crawls hundreds of millions of pages daily and returns results in under 200 milliseconds. The startup enters a crowded field where competitors Parallel raised $100 million（約160億円） at a $2 billion（約3200億円） valuation, Exa pulled in $85 million（約140億円）, and Tavily was acquired by Nebius for up to $400 million（約640億円）. Funding will support continued search-stack development, hiring, and the start of enterprise sales.

FAQ

How is Seltz's search engine different from existing AI search products?: Seltz owns the entire search stack—web crawler, search index, retrieval models, and ranking—rather than wrapping Google, Bing, or Brave's APIs. It scores individual passages and extracts the specific table, text, or image an agent needs, rather than returning full pages or summaries.
Who is Seltz's founder and what is his background?: Antonio Mallia is the founder and CEO. His PhD work in computer science at New York University focused on information retrieval, and he previously worked as an applied scientist on Amazon's artificial general intelligence team and as a research scientist at the vector-database company Pinecone.
How many people work at Seltz?: Seltz currently employs 15 people, only a half dozen of whom are full-time employees. The team works fully remote, split between the San Francisco Bay Area and hubs near European universities in Pisa, Italy, and Leipzig, Germany.

Discussion

No comments yet. Be the first to share your thoughts!

OpenAI and Broadcom have jointly developed Jalapeño, a custom chip designed specifically for running large language models, with deployment planned for late 2026 at gigawatt scale.

THE DECODER1h ago

AI text detector Pangram says language models are detectable because they produce uniform arguments, clustering in narrow bands unlike the diversity of human reasoning.

THE DECODER4h ago

Tecan integrates AI agents into its lab analytics platform using NVIDIA's toolkit, enabling laboratories to prevent operational issues proactively rather than react to them after they occur.

Yahoo Finance AI7h ago

Anthropic launched Claude Tag, a Slack integration that lets teams delegate work to Claude as a persistent team member rather than a chat interface.

Latent Space7h ago

Mistral AI releases OCR 4, a document-reading model that outperformed competitors in blind tests and is now available through multiple platforms at $4 per 1,000 pages.

THE DECODER7h ago

Anthropic launches Claude Tag, embedding its AI directly into Slack for team collaboration, with internal testing showing the tool already generates 65 percent of code on the company's product team.

THE DECODER7h ago

Stay ahead with AI news

Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.

Get Started Free

Free · takes 30 seconds · unsubscribe anytime

5 minutes a day. The AI essentials.

200+ sources · Email / LINE / Slack

Get it free →