VR WEBLINK / POLICY 2026-08-20
Simulator data
& privacy.
In short
We record your runs to check times for the leaderboard. We may also use them, labelled as sim data, in robot learning research.
This policy applies to Homebrew VR Weblink at homebrew.build/sim. It explains what the simulator records, what appears publicly, how long data is kept, and the controls available to you.
What we collect
If you sign in with Google, we receive your Google account identifier, email address, display name, and profile image. We use these details to authenticate you and create your simulator account. Your email and Google name are not shown on the public leaderboard; you pick the name that appears there.
Every completed cup-stacking run records simulator-only data: task and difficulty, timestamps, robot joint state and controls, headset and controller poses and inputs, mapped teleoperation targets, magnet state, cup poses and velocities, completion checks, client/version labels, and technical metadata. A bounded local browser queue may temporarily hold a failed episode upload for up to 24 hours so it can retry. We also store session and attempt records, scores you register, and limited error diagnostics such as an error message, stack trace, route, browser type, and simulator state.
To see where players get stuck during setup, we log when a browser tab reaches each step: visiting the page, opening the control room, pairing a headset, starting a run, finishing a run, and signing in. Each entry carries a random ID that lasts only as long as the tab, the device role, and the difficulty, plus your account when you are signed in.
Public information
Your chosen display name or X handle, best time, recent registered scores, difficulty, run count, and leaderboard rank may be public. You choose the name once and cannot change it afterward; deleting the account removes it. If you press Share on X for a saved time, the result card you shared (your name, time, rank, and a picture of the simulated robot at the end of the run) is stored and shown publicly at its share link so X can display it. Your Google name, Google email, Google identifier, raw session data, episode files, setup step logs, and diagnostic events are not public.
Why we use it
We use the data to operate the PC-to-Quest simulator, verify submitted times, detect abuse, maintain leaderboards, diagnose failures, improve the experience, build simulation datasets, and develop or evaluate future robotic-control and related machine-learning models.
Consent and future training
Before opening the PC control room, you must accept the current simulator data notice. That consent covers recording completed simulator runs and their possible future use for model development, training, evaluation, and research. We version and timestamp this consent with each stored episode. If the purpose changes materially, we will update this notice and ask again.
Retention
Account details and public scores remain until you delete them. A shared result card stays until you share that difficulty again, which replaces it, or delete your scores or account. Raw completed episode archives, setup step logs, and anonymous or de-identified session records are retained for up to 12 months. Minimal attempt records that support a score remain with that score. Error diagnostics are retained for up to 90 days. Backups may retain deleted identity or score data for up to 30 additional days before rotation. We may retain aggregated statistics that no longer identify an account.
Deletion and control
After signing in, the PC control room provides separate controls to sign out, delete all public scores, or delete your simulator account. Score deletion removes leaderboard entries, statistics, and shared result cards but keeps non-public session and episode records under the stated retention period. Account deletion removes the Google account identifier, email, display name, public username, avatar, public scores, leaderboard statistics, shared result cards, and account-linked diagnostics. Completed cup-stack sessions, attempts, and episode archives are retained without the account link until their normal retention expiry and may still be used for model development, training, evaluation, and research.
Services and security
We use Google for authentication, Vercel to host the application, Neon Postgres for account and simulator records, Upstash Redis for short-lived signaling and rate limits, and Cloudflare TURN when a relayed WebRTC connection is needed. We limit public data, protect internal exports with an administrator allowlist, rate-limit sensitive endpoints, and hash completed episode contents. No internet service can be guaranteed perfectly secure.