Editorial policy

Last updated: April 23, 2026

Why we publish this

Google’s Helpful Content Update and the EEAT (Experience, Expertise, Authoritativeness, Trust) framework explicitly reward sites that disclose how content is produced — especially when AI is involved. This page documents our pipeline end to end so readers, crawlers, and reviewers can verify the provenance of anything we publish.

How a Discury article is produced

Every article on Discury Digest and every entry on Discury Problems goes through the same pipeline:

1. Research. An agentic AI loop reads Reddit, Hacker News, and Product Hunt threads relevant to a topic. The agent decides on its own queries, pivots when early results are weak, and pulls top-level comments only when the thread is substantive. No pre-scripted prompt chain.
2. Fact extraction. A dedicated extractor reduces the research corpus into a closed-world fact table — named products, direct quotes, numeric claims. Every fact is tagged with its source permalink.
3. Writing. A local Qwen 3.6-35B model or a cloud model (Google Gemini) drafts the article. The draft is constrained to the fact table; fabricated claims are rejected before review.
4. Critic pass. A second cloud model grades the draft on a five-dimension rubric (evidence quality, problem specificity, willingness-to-pay signal, de-duplication, helpful-content compliance). Articles below threshold regenerate with feedback; repeat failures are rejected.
5. Meta-review. A final pass checks temporal claims, brand mentions, and publish-readiness. The reviewer is instructed to be strict rather than lenient — a non-go is more valuable than a lukewarm go.
6. SEO wrap. Title, meta description, Article JSON-LD, BreadcrumbList, FAQ schema (auto-detected), internal links, and citation schema are computed at publish time. The article surfaces at its canonical URL on the relevant subdomain.
7. Human editorial oversight. A named editor (see team below) owns the category and intervenes when the system publishes something that doesn’t meet our editorial standard. Edits, retractions, and takedowns are handled out of band; the same editor signs every article in their category.

AI disclosure

Discury content is AI-assisted. It is written by language models, researched by language models, and reviewed by language models. A human editor signs off on each category (see the team section below) and owns corrections, but the word-level drafting is not done by a human. We disclose this because Google’s HCU framework, the EU AI Act, and simple reader trust all benefit from it being explicit.

What this means in practice:

• Facts are grounded in cited sources. Every quote in an evidence list is a verbatim Reddit comment with a permalink.
• Claims not backed by the fact table are rejected by the critic. Hallucinated competitor names, invented statistics, and placeholder URLs are fatal defects.
• Opinions — “this is a good idea”, “this is the wrong move” — are infrequent and, when present, attributed to the named editor.
• The editorial category-to-editor mapping is deterministic and public (see the team below); the byline you read at the top of an article is the same person who owns corrections.

Editorial team

Three named editors own Discury content. Each has a dedicated profile page with their areas of focus and reachable contact info:

• Michal Baloun — COO · Customer intelligence · Community research · Data quality
• Tomáš Cina — CEO · Digital marketing · Growth strategy · Sales operations
• Jan Hilgard — Tech Entrepreneur · AI infrastructure · Local LLM inference · Fullstack development

Sources and citations

Discury Digest articles cite the source subreddits the research pipeline read. Discury Problems entries additionally cite specific Reddit permalinks (post or comment URLs) for every direct quote. All citations are emitted in both the visible article body and the schema.org Article.citation JSON-LD field so crawlers can verify the source trail machine-readably.

We do not cite paywalled sources, private Slack/Discord threads, or anonymous tipsters. If Reddit removes a thread we cited, the citation stays in the article with the original permalink and the article gains a “source archived” note on the next refresh.

Corrections

Factual errors are corrected in place; the article’s dateModified moves forward and we annotate the change visibly when it changes the article’s conclusion. For competitor-facing claims (e.g. “X does not offer Y”), we correct within 48 hours of notification from the affected party. Contact: support@discury.io.

Takedowns

If you are quoted by permalink and want to be removed, email the above address with the permalink. We will remove the evidence entry (and any article that no longer meets our minimum citation threshold after removal) within 7 days. Reddit quotes are public by the platform’s terms of service; we honor takedowns anyway because they’re the right default.

Advertising & conflicts of interest

Discury has no paid placements, affiliate commissions, or sponsored content. Articles do link to the Discury product — that’s the gated analytical layer behind the problem catalog — but every such link is internal, not an advertising unit. The pipeline does not receive money from any company mentioned in any article; competitive-map entries are selected on topical relevance alone.

Contact

General inquiries: support@discury.io · contact form. Editorial inquiries (corrections, takedowns, factual disputes): reach the named editor via their LinkedIn on their author profile.