The intelligent file browser for the generative age - built to store, search, and think alongside everything you own.
You named it final_v3_ACTUAL.pdf at 1 a.m. and swore you'd remember. You won't. Multiply that by a decade of downloads, screen recordings, half-read papers, voice memos, and slide decks, and you have the quiet crisis of modern work: we own more information than ever and can locate almost none of it. The folder - that beige metaphor from 1984 - never learned to keep up.
Poly's wager is that the file system itself has to evolve. Not another chat window bolted onto storage, but the browser underneath rebuilt so it understands what's inside your files - and answers when you ask. Type "the call where we argued about pricing" and it surfaces the recording. No filename required.
Poly acts like cloud storage - but with an AI that has actually opened every file. Upload text, PDFs, office docs, images, audio, video, or a URL, and it handles the tedium you'd otherwise avoid.
Natural-language, multimodal search finds files by what they contain - not the name you forgot to give them.
Ask the assistant to condense a 90-minute call or translate a report, without leaving your drive.
Poly tags files, builds folders, and renames the mess for you - filing without the filing.
Teams query the same library together, pulling insight from a pile of mixed-media files.
An MCP server connects your Poly files to ChatGPT and Cursor - your storage becomes AI memory.
100GB free, or 2TB for $10/month. Room to keep everything, plus a reason to.
Poly's search feels simple because the engineering isn't. At its center sits a proprietary embedding model, Polyembed-v1, trained to understand a wide spread of content on equal footing - text, PDFs, documents, presentations, spreadsheets, audio, video, code, and URLs. A video clip and a paragraph become points in the same map of meaning, which is why "find the part where we discussed churn" works across a movie of a meeting.
It's the difference between a filing cabinet and a librarian who has read every book. Most storage tools index your filenames. Poly indexes your ideas.
Poly began by generating 3D textures and design assets from text prompts. It was working - and that was the problem. Watching the AI-asset market fill with well-funded rivals, the team made an unfashionable call: stop, and go build something harder to copy. Two years of stealth later, the same crew returned with a file system for the AI era.
YC-backed Poly launches AI 3D-asset and texture generation; raises an initial $3.9M.
Winds down the asset product and enters stealth to rebuild around intelligent storage.
Emerges with $8M seed led by Felicis; relaunches as an AI-native cloud file browser.
Public launch on web and Mac with 100GB free; Windows announced as coming soon.
AI researcher and designer. Former research fellow at Microsoft Research and lecturer at Stanford; BS in EECS from UC Berkeley, MS from Stanford.
Co-founded Poly in 2022 during its AI-asset era. Has since departed the company.
No official interview video is confirmed at time of writing - the YouTube link runs a live search for Poly demos and coverage.