This is where it started
“Press-release facts are scattered across team documents, so even a fast first draft still demands lengthy fact and tone review.”observed → Started 2025.12.01
— So we built
brieFFlow
Livebriefflowweb service
An AI workspace for grounded press-release generation and quality evaluation.
Screenshot 1 / 3
Developer evidence
What I owned in this project
- Period
- 2025.12 – Operating now
- Team
- Solo developer
- Contribution
- Planning · design · development · deployment · operations, 100%
- Owned scope
- Connects document parsing, chunking, embeddings, pgvector/FTS hybrid retrieval, evidence policy, agent generation, citation storage, and evaluation gates within team scope.
The document workflow that began as PressTuner has grown into the brieFFlow AI workspace.
Introduction
brieFFlow retrieves launch facts and team knowledge to draft grounded press releases, then manages output quality with citation and claim checks plus evaluation gates. Career-draft support remains available, while the core product is the team press-release workflow.
User flow
- 01
Collect launch or event facts and reusable team documents in the workspace.
- 02
Hybrid retrieval finds relevant evidence and AI produces a press brief and draft.
- 03
A person reviews citation, claim, and evaluation evidence before approving and refining the draft.
Highlights
- Builds startup press-release drafts quickly so teams do not start from a blank page.
- Retrieves team knowledge with hybrid RAG and connects citable evidence to the draft.
- Checks quality regressions before release with versioned evaluation sets and baseline-versus-candidate comparisons.