Saved run · 2026-09-27 · replayed in the Observer, nothing runs from this page
Watch eight synthetic participants use one app at the same time
The host created a lobby of a multiplayer movie-guessing game on its live deployment. Seven more participants joined through the real Join flow on their own hosted desktops and played five rounds. Six reached the final standings. Two were blocked, and humanish analyze turned the recording into seven findings, each linked to the capture that backs it.
Click the frame to expand it. Inside, press play, open a participant, or switch to Findings.
Seven findings came out of thirteen minutes
- BlockedF1
Keyboard guessing was hard to track; one keyboard-only participant stopped in round 3 with focus stuck in the browser toolbar.
- BlockedF2
A participant who joined after the host pressed Start was told to wait for the game to end, with no sign of when that would be.
- FrictionF3
Finding and verifying a film in the picker was uncertain; one participant fell back to browser Find.
- RecoveryF4
The lobby code landed in the display-name field first; the participant blamed the focus order and corrected it.
- FrictionF5
A confirmation briefly showed the category label instead of the chosen title.
- FrictionF6
The host questioned a 74-second wait after locking in an answer.
- FrictionF7
The host saw a stale round label during a transition and raised it again in the closing report.
Two of the seven are blocked tasks; the rest are friction and one recovery. The analysis says so itself when a capture does not prove a claim, and the Observer's Findings view links every observation to its evidence.
This is one lab file and one command
A lab declares the app, eight seats with a persona each, one marked host, and the missions. The orchestrator reads the join code off the host's screen and hands it to the followers. Each seat is a real desktop; the model key stays on this machine. The run wrote itself under .humanish/runs/, and humanish analyze ran once afterwards at an $11.38 admission estimate that came to $2.29.