Faceless Social Bench live

Weekly median views per post across the eight accounts in each arm. Both arms open where a new account opens; the gap only starts to widen once the adaptive arm has enough returned posts to rewrite its hook library on.

Eight fresh accounts per run, identical channel templates and cadence: five posts a week for twelve weeks. A task is one scheduled post, brief to published. Views are read 72 hours after posting and taken as the median across the run, so one break-out video cannot carry a system.

The adaptive arm rewrites its own hook library, posting times, and thumbnail rules from what the last two weeks returned. The control arm is the same team with those three frozen at their week-one settings, and it is what the multiple in the headline compares against.

Cost is everything billed to take a post from brief to published, model tokens, generation, storage, and the posting calls. Completion is the share of the sixty scheduled posts that actually shipped, and views per account is that median across the posts a run did ship, so a system that skips its hard weeks pays for it there rather than in the view count. Followers are net of unfollows over the run.

We define and run this benchmark ourselves. Other harnesses ran with stock tooling and our prompting, so treat the ordering as directional rather than as a settled ranking of the harnesses.