Smarter Generation: Voice learning & excluding internal work items ✨

Launchlane now learns from the edits you make before publishing, so regenerated drafts actually sound like your team over time. A new classification layer also keeps internal work items out of customer-facing updates and prevents them from slipping through auto-publish. This release also fixes a bug where removed draft items could quietly reappear.

New Features

  • Auto exclude internal work items from Updates — Work items are now classified before they reach a draft, so internal chores and engineering tasks are filtered out of customer-facing updates automatically. When classification is ambiguous, a secondary LLM review resolves the call, so edge cases don't quietly slip through to your changelog. Auto-publish will no longer ship something that was never meant for your customers.

Improvements

  • Voice learning from published updates — When you edit a draft before publishing, those edits are now remembered. Future generations use your published updates as examples, so the writing gets closer to your team's actual voice instead of repeating the same mistakes.

Bug Fixes

  • Fixed an issue where removing a linked item from the per-draft sidebar would cause it to revert to pending and reappear in the draft instead of staying removed.