01
Slot
Within 15 minutes of approval a piece gets a channel, a local time and a reason: the idea's timing window (rising and peak trends first), best hours per geo, spacing per channel and day, the brand's blackouts and the channel's weekly cadence target.
02
Gate
The channel's autonomy level decides what happens at slot time. Pre-checks run before anything is sent: size, duration and aspect against the platform, a public URL, caption length, hashtag count, the AI label on clone videos, opt-ins for WhatsApp.
03
Publish
A job moves queued → uploading → processing → published, or failed, or needs_manual with a plain reason. Three retries with backoff, Retry-After honoured, tokens redacted from every log and error. dm publish runs the jobs.
04
Outcomes
At 24 h, 72 h and 7 days per post: views, likes, comments, shares, saves, follows, clicks, installs — from the platform API for connected channels (a partial snapshot until the platform's own latency passes), by hand otherwise. Manual rows are never overwritten.
05
Learn
dm learn compares the idea's ICE impact with realised performance normalised by the channel's 90-day median, lists winners and losers with their loops, the observed best hours, and the two gate metrics — and appends guidance to the next ideation prompt.
06
Recycle
Cadence targets per channel push a saturated channel to next week unless the signal is rising; evergreen recycling re-queues the top posts by 7-day views on channels below cadence, 21 days apart at least. Off by default.