Hey YC 👋

We're Rounak, Bishal & Sourya — co-founders of ProjectX (P26). Today we're launching Infinity.

[https://youtu.be/ZYf2vdWMkDk]

TL;DR

Infinity is a new computation layer — a web-based OS where every app you open runs on its own independent compute and GPU, inside one shared workspace. Open Blender, Unreal, Isaac Sim, DaVinci, VS Code in the same browser tab. Each app is its own machine with its own GPU. They share one filesystem and one terminal. Windows and Linux apps run side by side. Runs on any browser, on any device. Open fifty apps at once — nothing slows down.

Problem

The apps people actually use to build the future — Blender, Unreal, Isaac Sim, Houdini, the CAD/CAM/CAE stack, ML pipelines — still need a $5K workstation under your desk. The cloud was supposed to fix this and didn't. Renting a cloud GPU means hours of driver installs, version matching, container debugging before you do any actual work. Streamed desktops (Citrix, Paperspace, Shadow) give you one VM with one GPU — built for one user running one app at a time. Open Blender and Unreal at the same time and you're fighting yourself for resources.

It's hitting agents too. Claude Cowork runs on your laptop, sharing your GPU with everything else. Manus runs in a CPU-bound microVM that can't open heavy apps at all. Claude can drive Blender via MCP, but Blender still runs on your machine — the agent just borrows your compute.

Solution

Infinity makes every app its own computer.

Open Blender, you get a machine with a dedicated GPU. Open Unreal in the next window, that's a second machine with its own GPU. Open Isaac Sim, that's a third. They share a filesystem and a terminal so the workspace feels like one computer, but underneath they're independent machines. Open fifty apps, you have fifty machines. Nothing slows down. Runs entirely in a browser tab — Chromebook, iPad, five-year-old laptop, doesn't matter.

Every Infinity workspace comes will soon ship with a built-in computer use agent that can access run applications without being constrained by compute, no matter how many applications you run or how heavy the workflow is. Tell it what you want, and it spins up the apps, drives them, hands files between them, and finishes the work. You can drive Infinity with a mouse and keyboard, or you can drive it with a prompt.

A concrete example. You prompt: "Make a 90-second product launch video. Write the script, generate the voiceover, edit in DaVinci, render the 3D logo in Blender, post to YouTube."

Six apps spin up at once:

• Claude writes the script
• ElevenLabs generates the voiceover
• DaVinci Resolve edits the cut
• Blender renders the 3D logo
• Chrome uploads to YouTube
• VS Code writes the publish script

They share a workspace. They hand files to each other. One finished video in an hour, not three days.

Where we are — 15 days in early preview

• 20,000+ builders signed up
• 1,160% growth in the last 14 days
• 20M+ impressions on X and Instagram
• 10,000+ signups in the last 7 days
• Out of server capacity — rate-limiting new signups while we scale

Used by engineers at Reliance Industries, CEMEX, Cognizant, Globant, Ola Electric, Lattice, Quizizz, Appsmith, Bonsai Robotics, Induction Labs, Basis Robotics, and 130+ more teams. All inbound. Our first public demo crashed our servers at midnight; we scaled the infra live at 3am and woke up to inbound from 50+ enterprises.

Future

Once every app has its own compute, agents stop being a separate problem.

A computer-use agent is just another app — except the agent drives itself. Because every agent gets its own machine on Infinity, ten agents work in parallel without stepping on each other. The thing that's impossible today — ten parallel cursors, ten parallel renders, ten parallel models — is the default on Infinity.

We're shipping the multi-agent runtime next. One prompt, a mother agent plans and coordinates, child agents execute on dedicated hardware — each routing through Claude, GPT, Gemini, or any MCP tool. Same substrate humans use, given to agents.

Ask

If you are, or know any of the following, we want to hear from you:

Heavy-GPU teams. Robotics, Physical AI, 3D/VFX, ML, post-production, CAD/CAM/CAE. If you're buying workstations or wrestling with cloud GPU setup, let's talk.

Agent builders. If you're shipping a computer-use agent and hitting the "where does this actually run" wall, we should talk early.

Free Infinity credits for all YC founders and alums. Email me directly: rounak@projectx.cloud

— Rounak, Bishal & Sourya