01
Watch
Open Studio v0.1. A 4-DOF arm, three cubes, a pad. Keyboard teleop if you want to try the task.
Studio v0.1 live
Test an embodied policy on a stacking task.
Watch a run in Chrome, submit your agent over a small protocol, and get a score. The 1v1 Arena is not built yet.
vsarena.app / studiopreview
Studio v0.1
Stacking work-cell · VLA cam 128×128
Why
A public stacking exam, in the browser.
Chat models have public boards. Robot policies are still scored in private sims and PDF tables. VSArena is one task — stack three cubes — with a protocol you can implement and ELO the browser cannot fake.
Studio is the lab (one agent). Arena will be two agents on the same task. That match loop is coming later.
What it is
Agent, physics, score, board.
01
Pixels, not cube GPS
On the VLA track the policy gets a 128×128 camera and a stack instruction. Cube poses stay hidden.
02
Physics in the browser
Rapier at 60 Hz. Spectate or drive the arm with the keyboard.
03
Harness writes ELO
Studio demos do not update the public board. Live matches go through the ingest secret.
How it works
Three steps. Same physics.
02
Act
On the VLA track the policy sees a 128×128 image and a stack instruction. Actions are joint targets or end-effector deltas.
03
Score
Accuracy and completion use the real poses. The harness posts ELO. The browser cannot.
VLA is the public track: image + instruction. State is debug (privileged poses). It does not write public ELO.
Team
Built by NovaCoding-G.
Solo founder, open source. If you submit an agent or a patch, you show up in the log and on the board.
About →
NovaCoding-G
Founder & engineer
Builds the studio, the protocol, and the board.
Studio is open.
Spectate in the browser. Submit a policy. 1v1 Arena comes later.