Most people face the same problem when it comes to file systems. Drowning in folders, forgetting what you named that document, spending 15 minutes searching for something you know exists but can’t locate. The list is endless. Knowledge workers waste roughly 19% of their time just hunting for information. Poly is trying to change that.
On Wednesday, the YC-backed startup, Poly announced it has raised $8 million in seed funding led by Felicis, with participation from Bloomberg Beta, NextView, Figma Ventures, AI Grant, Wing Ventures, and MVP Ventures.
It’s the culmination of two years in stealth and a complete product pivot from the company’s original 2022 vision. Poly is launching with a simple premise. Cloud storage built from the ground up for AI-native search, launching with 100GB of free storage at no cost.
Founders Abhay Agarwal and Sam Young started Poly in 2022 with a different mission entirely. Generating 3D assets using AI prompts. It was a reasonable bet in 2022, but the generative AI wave hit harder than anyone expected. Competitors raised massive rounds and the market exploded. Poly realized that winning in this market was next to impossible.
So the team did something harder than doubling down. “We interviewed our users and asked them what the pain points of their workflows were that could be solved by AI. Turned out that one big unmet need for users was organizing their file system,” Agarwal said.
That conversation led to a shutdown of the original product in 2023, two years in stealth rebuilding, and a complete reimagining of what Poly should be. Young has since left the company and Agarwal, who worked on vision AI for accessibility at Microsoft, is leading the new vision.
What Does Poly Do?
Poly lets you upload files and then asks it questions in plain English. “What’s the budget for Q4?” “Summarize this video.” “Find images from the New York conference.” The AI understands context across formats and returns precise answers with citations including timecodes and page numbers.
Unlike federated search tools that connect to multiple services, Poly invites you to consolidate your files in its cloud. Files sync from your computer, but they also live locally. Poly adds an intelligent layer on top like AI-powered search, auto-organization, tagging, summarization, even content generation.
You can paste a YouTube link and get an instant summary. Tag files and have the system auto-organize them. Collaborate on shared drives with teammates who can all query the same corpus.
Agarwal views it as an upgrade to Google’s NotebookLM but with better file organization and retrieval. However, Poly doesn’t currently have live web search or video generation capabilities that some competitors offer.
Poly’s competitive edge sits in a proprietary embedding model called Polyembed-v1. It’s trained to understand a vast range of content types like text, PDFs, presentations, spreadsheets, audio, video, code, and URLs.
The company claims this embedding model outperforms similar ones not only for document retrieval but also for broad search queries and multimodal requests mixing text and media.
That’s the real differentiator. Search quality, not storage quantity, will determine winners in this space. Dropbox and Google Drive have AI search tools. But Poly’s claim is that its search actually works better, at least according to early testers who found Poly outperformed Google Drive’s search in their own trials.
Poly is competing directly with Dropbox and Google Drive, both of which have added AI search features. But Poly’s bet is different. Those companies see AI search as a feature bolt-on to existing storage products. Poly sees it as the foundation. The entire product is designed around retrieval and understanding, not just syncing and sharing.
The company is targeting knowledge workers and AI-native creators, researchers, service executives analyzing customer calls, creative teams organizing reference materials, analysts digging through archives. People who measure success in insights, not megabytes.
The Roadmap
Poly launches with web and Mac apps, with Windows coming soon. The roadmap is ambitious with web search integration, stylized report generation, text and Markdown editors, custom metadata, Google Docs linking, AI agents that can analyze spreadsheets, and direct file and folder sharing outside of shared drives. The company is already offering a Model Context Protocol (MCP) server so you can use Poly within ChatGPT or Cursor.
Workers are generating more files than ever. AI is making multimodal search feasible. Cloud storage without intelligent retrieval feels increasingly antiquated. Poly is entering a market where the incumbent tools haven’t fundamentally rethought how file systems should work in an AI-first world.
“Poly was built on the belief that in an AI-first world, the file system itself has to evolve. We created a browser that replaces Finder or File Explorer with an intelligent interface that can help search, understand, and create across terabytes of information. It’s like having an LLM with infinite context from your life,” Agarwal said.
