Help · Guide
Advanced: inside the learning loop and prompt versions
How real engagement becomes lessons, how those lessons rewrite your content prompt, and how versioning and rollback keep it safe.
A closed feedback loop
The learning loop pulls real engagement from each connected platform's read API, turns the patterns into plain-language lessons, and proposes a rewritten version of your content prompt built from the lessons you accept. You approve the new version in one click and it becomes the active prompt for future generation. You can roll back to any earlier version at any time.
Honest coverage
Metric coverage is declared per connection, not assumed. Facebook and X report full engagement; LinkedIn organisation pages and Instagram are partial; some connection types can't report metrics at all and say so rather than guessing. Where real metrics aren't available, the loop falls back to the signals you record manually, and always tells you which is which.
Provenance and diffs
Every proposed prompt version records exactly which lessons it came from, and shows a before/after diff so you can see what changed and why before you accept it. Nothing is applied silently: the improvement is auditable, reversible, and tied back to the evidence that motivated it.