Editorial Policy

This document describes the standards for content selection, verification, and publication on the Codver.AI platform.

1. Content selection principles

• Timeliness: we only publish news no older than 24 hours
• Originality: content does not duplicate already-published articles
• Significance: minimum scoring threshold of 0.30 on the news_score scale
• Focus: AI, technology, startups, and regulation only

2. AI-generated content

Post texts are generated by the Claude language model (Anthropic) using pre-defined editorial prompts. The model has no internet access and works only with data from verified sources provided to it.

Each publication goes through:

• Predictive scoring algorithm (scroll_stop, read_depth, CTR, novelty)
• Quality review by a human operator before mass publication

3. Fact-checking

All publications are based on primary source materials (official press releases, arXiv, SEC filings, corporate blogs). Content labeled as "Insight" or "Analysis" represents editorial interpretation and does not claim the status of reported fact.

4. Corrections and removals

When a factual error or outdated information is identified:

• On Telegram: the post is edited with an [Updated] label
• On the website: the article is updated with a last-modified date
• For significant errors: the publication is removed entirely

Correction requests are accepted via the contact form.

5. AI disclosure

Every publication carries the label: "Generated by Codver AI, verified by humans." This complies with emerging AI content labeling standards (EU AI Act 2024 and equivalent).