Space is a distributed filesystem that mounts as a native drive on your computer and streams cloud-hosted files on demand, so applications open project data as if it were local without waiting for full downloads or keeping permanent copies on disk. It targets teams whose work already exceeds the size of a laptop SSD, including video editors, 3D and CAD users, and developers, and it is pitched equally at AI agents that today have to ingest or upload entire files just to read a fraction of them. Built directly above the operating system, Space makes cloud-hosted files behave like local ones. Files appear in Finder and open in existing native applications without a download or permanent local copy. When an application or agent requests data, Space streams only the exact byte ranges required, in real time, to complete the task.
The launch matters because it coincides with the company opening access beyond a private beta. Space announced a $2.4 million pre-seed round led by a16z Speedrun, with participation from Golden Ventures, Northside Ventures, and a dozen angel investors with backgrounds across prosumer and enterprise software such as Parsec, Sentry, Stan, Superwhisper, and Modem. Space is currently in private beta, with roughly 100 users and teams onboarded, backed by an organic audience of more than 80,000 across platforms. The positioning against incumbents is deliberate. This is what sets Space apart from Dropbox, Box, and Google Drive. Those tools are built primarily around syncing entire files onto local devices or bringing workflows into web applications.
Space was founded by Matthew Ao, Arihant Bapna, and Jason Zhao after the same constraint followed them from individual work into operating a company. Zhao first encountered the problem after accumulating dozens of terabytes of footage across drives as he documented his life on YouTube over the last decade. In November 2025, the founding team built the first Space prototype. The team is working out of Founders Inc. in San Francisco, and their stated roadmap moves from creative and AEC teams toward AI-native companies handling training data, computer vision pipelines, and world models, with a longer arc toward what they call the Space Computer.
Comments (7)
Sign in to join the discussion.
P
Priya Deshmukh18d ago
The word 'infinite' is doing an olympic amount of heavy lifting here. What happens on my S3 bill when my agent decides to hoard every intermediate tensor it's ever generated?
T
Tomas L.18d ago
That mount demo in the video is clean but the voiceover audio sounds like it was recorded inside a Tupperware. Fix the mic and this thing spreads twice as fast.
K
kwesi18d ago
Every filesystem startup pitches 'work directly in the cloud' and then FUSE eats their soul on macOS. Curious how you handled the kernel extension nightmare.
Y
Yu-Lin Zhao18d ago
Ex-teammate shipped basically this concept for internal ML datasets in 2019 and it got killed because nobody wanted to debug stale caches at 2am. Genuinely rooting for you to solve the part we couldn't.
R
Renata Bevilacqua18d ago
Before I forward this to my platform team: SSO, SOC2, audit logs, and can I restrict which agents get read vs write scopes on a Space? Procurement will ask, I'm just the messenger.
M
moe18d ago
Tweet copy is four arrows and a vibe. Would kill for one line about latency because streaming files to my laptop from us-east-1 has historically been a spiritual experience.
B
Bex R.18d ago
Filesystem for agents is a cute framing until an agent rm -rf's your Space at 3am and blames the prompt. Undo button better be a first-class citizen.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
Space
The infinite AI-native filesystem
The infinite AI-native filesystem
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
Space is a distributed filesystem that mounts as a native drive on your computer and streams cloud-hosted files on demand, so applications open project data as if it were local without waiting for full downloads or keeping permanent copies on disk. It targets teams whose work already exceeds the size of a laptop SSD, including video editors, 3D and CAD users, and developers, and it is pitched equally at AI agents that today have to ingest or upload entire files just to read a fraction of them. Built directly above the operating system, Space makes cloud-hosted files behave like local ones. Files appear in Finder and open in existing native applications without a download or permanent local copy. When an application or agent requests data, Space streams only the exact byte ranges required, in real time, to complete the task. The launch matters because it coincides with the company opening access beyond a private beta. Space announced a $2.4 million pre-seed round led by a16z Speedrun, with participation from Golden Ventures, Northside Ventures, and a dozen angel investors with backgrounds across prosumer and enterprise software such as Parsec, Sentry, Stan, Superwhisper, and Modem. Space is currently in private beta, with roughly 100 users and teams onboarded, backed by an organic audience of more than 80,000 across platforms. The positioning against incumbents is deliberate. This is what sets Space apart from Dropbox, Box, and Google Drive. Those tools are built primarily around syncing entire files onto local devices or bringing workflows into web applications. Space was founded by Matthew Ao, Arihant Bapna, and Jason Zhao after the same constraint followed them from individual work into operating a company. Zhao first encountered the problem after accumulating dozens of terabytes of footage across drives as he documented his life on YouTube over the last decade. In November 2025, the founding team built the first Space prototype. The team is working out of Founders Inc. in San Francisco, and their stated roadmap moves from creative and AEC teams toward AI-native companies handling training data, computer vision pipelines, and world models, with a longer arc toward what they call the Space Computer.
Comments (7)
The word 'infinite' is doing an olympic amount of heavy lifting here. What happens on my S3 bill when my agent decides to hoard every intermediate tensor it's ever generated?
That mount demo in the video is clean but the voiceover audio sounds like it was recorded inside a Tupperware. Fix the mic and this thing spreads twice as fast.
Every filesystem startup pitches 'work directly in the cloud' and then FUSE eats their soul on macOS. Curious how you handled the kernel extension nightmare.
Ex-teammate shipped basically this concept for internal ML datasets in 2019 and it got killed because nobody wanted to debug stale caches at 2am. Genuinely rooting for you to solve the part we couldn't.
Before I forward this to my platform team: SSO, SOC2, audit logs, and can I restrict which agents get read vs write scopes on a Space? Procurement will ask, I'm just the messenger.
Tweet copy is four arrows and a vibe. Would kill for one line about latency because streaming files to my laptop from us-east-1 has historically been a spiritual experience.
Filesystem for agents is a cute framing until an agent rm -rf's your Space at 3am and blames the prompt. Undo button better be a first-class citizen.