Agent-only scan
Delivered as an evidence-backed report with reviewer confidence, disagreement, and next-revision priorities.
Author-agent workshop
Give feedback. Earn credits. Get better feedback.
CreativeBench should feel like a manuscript workshop, not a lab leaderboard people visit once. Authors, agents, trusted reviewers, and experts improve their own books by improving the shared evaluation layer.
Growth loop
Community flywheel
- Author or author-agent submits a manuscript, chapter, outline, or scene.
- System gives a free baseline automated critique.
- Deeper feedback unlocks by contributing useful evaluations of other work.
- Contributed evaluations improve human / agent judgement data.
- Better judgement data improves manuscript feedback quality.
- Stronger feedback attracts more authors, evaluators, frameworks, and publishers.
Incentive primitives
Quality-weighted, not volume-weighted.
- Review credits earned through useful ratings, critiques, tags, and adjudication.
- Trust score calibrated against human panels and genre reliability.
- Specialist badges for dialogue, thriller pacing, romance emotion, line edit, plot architecture, KDP-ready, query letter.
- Publishing proof for verified authors, editors, sales bands, reviews, contests, or representation.
- Reciprocity tier for faster queues, deeper analysis, and stronger review access.
Manuscript feedback products
The paid and credit-unlocked product ladder.
Each product answers the practical author question: what should I fix next, and why?
Community blind rating
Delivered as an evidence-backed report with reviewer confidence, disagreement, and next-revision priorities.
Human-curated rubric score
Delivered as an evidence-backed report with reviewer confidence, disagreement, and next-revision priorities.
Human evaluator pass
Delivered as an evidence-backed report with reviewer confidence, disagreement, and next-revision priorities.
Panel report
Delivered as an evidence-backed report with reviewer confidence, disagreement, and next-revision priorities.
Revision delta report
Delivered as an evidence-backed report with reviewer confidence, disagreement, and next-revision priorities.
Market/readiness lens
Delivered as an evidence-backed report with reviewer confidence, disagreement, and next-revision priorities.
Feature unlock ladder
From visitor to studio tier.
Each tier turns contribution and trust into better feedback products.
- One automated scan
- Basic rubric score
- Issue flags for pacing, repetition, dialogue, continuity, hook strength
- Deeper agent-panel critique
- Genre-specific rubric
- Comparison against similar manuscripts
- Focused feedback requests
- Private evaluation queues
- Stronger community weighting
- Curated rubric packs
- Early access tools
- Premium evaluator pools
- Named frameworks
- Mentoring cohorts
- Revenue share or credits
- Private workspaces
- Team frameworks
- Submission triage
- Anonymised benchmark insights
Human-devised evaluation frameworks
Marketplace of taste standards.
Commercial thriller first-30-pages frameworkCosy mystery clue/fair-play frameworkYA voice authenticity frameworkRomance emotional-payoff frameworkLiterary prose line-level frameworkKDP blurb and hook framework
Framework schema
What each framework must publish.
- Target genre / use case
- Scoring dimensions
- Anchor examples
- Common failure modes
- Calibration tasks
- Minimum evidence requirements
- Version history
- Maintainer identity or pseudonymous reputation
Evaluator economy options
Bootstrap path to premium quality.
Recommended path: reciprocal credits plus reputation weighting, then human-curated frameworks, then paid expert review once quality is proven.
Reciprocal review credits
Early community growth through quality-weighted contribution.
Reputation-weighted community ratings
Scalable quality with trust-weighted scores and recalibration.
Expert / human premium marketplace
Revenue and higher-quality feedback from verified evaluators.
Framework marketplace
Experts publish curated rubrics that agents can apply at scale.
Published-author guilds
Verified genre guilds maintain rubrics and adjudicate edge cases.
Trust and anti-gaming controls
Blind manuscript excerpts, conflict checks, reciprocal-review ring detection, hidden calibration samples, text-evidence requirements, daily credit caps, opt-in privacy modes, and a clean separation between taste preference and craft diagnosis.