Gate: Article Publication
What It Replaces
Multi-step editorial review chain: Katie Parrott (AI tells detection) → Eleanor Warnock (first-pass triage) → Kate Lee (final editorial review). Currently takes 2-5 days; Kate reviews every article personally.
Pass Criteria (visible to executing agent)
Tier 1: Automated First Pass (blocking)
- Thesis test: Article contains a specific, falsifiable claim or argument stated within the first 3 paragraphs. Not a recap, summary, or roundup.
- AI tells check: No formulaic transitions (“Moreover,” “Furthermore,” “In conclusion”), no hedging without substance (“It’s worth noting”), no correlative constructions adding nothing, no vague pronouns (“This approach”), no unsourced claims.
- Experience grounding: Contains at least one concrete example from first-hand experience (“I built,” “We tried,” “Here’s what happened”) — not hypothetical or theoretical.
- Voice authenticity: Written in first person. Does not contain corporate blog phrases (“We’re excited to announce,” “leverage,” “cutting-edge”). Sounds like a specific author, not a generic publication.
- Length and structure: Meets minimum depth for the topic (no thin content). Has a clear beginning (thesis), middle (evidence), and end (implication).
Tier 2: Quality Assessment (advisory, flags for human review)
- Learnable value: Reader gains a framework, technique, or insight they can apply. Evaluator asks: “After reading, what can I now DO that I couldn’t before?”
- Originality: Not substantially similar to content already published on Every or widely available elsewhere.
- Builder credibility: Claims about AI capabilities are grounded in actual demonstrated use, not speculation. If speculative, clearly labeled.
Satisfaction Metric
Target: 90% of articles passing this gate at Tier 1 should also pass Kate’s manual review without significant revisions. If the rate drops below 80%, gate criteria need recalibration.
On Fail
- Tier 1 failure (criteria 1-5): Return to author with specific feedback on which criteria failed and why. Author revises and resubmits.
- Tier 2 flags (criteria 6-8): Route to Kate Lee for human review with the flagged criteria highlighted. Kate decides: pass, revise, or reject.
- Multiple Tier 1 failures on same article: Escalate to Eleanor Warnock (Managing Editor) who triages whether the piece is viable or should be shelved.
Escalation Package
When escalated to Kate Lee, she sees:
- The full article draft
- Which specific criteria failed or were flagged
- The automated assessment with highlighted problem areas
- Author’s revision history (if resubmission)
- Suggested edits (if the agent has specific recommendations)
Architecture
- Type: Sequential, blocking at Tier 1
- Runs: Before any article enters the publication queue
- Parallel with: Nothing — this is the single entry gate
- Kate Lee’s role shift: Quality Architect — designs and evolves pass criteria. Reviews only flagged articles (estimated ~30% of submissions). Retains absolute veto power.
- Katie Parrott’s role shift: Specification Authority — defines and evolves AI tells detection criteria as models improve. Maintains the detection skill library.
Political Risk
- Kate Lee (Medium): Reframed as Quality Architect. She designs what “good” looks like at scale rather than reviewing everything personally. Key: she retains veto and reviews all borderline cases.
- Katie Parrott (Medium): Reframed as Specification Authority. Her judgment is encoded, not replaced. She evolves criteria as AI writing improves.
- Eleanor Warnock (Low): Benefits from reduced pipeline management burden.