Editorial Methodology

Who we are

SENTINEL is operated by Oblivion Technologies LLC, a Wyoming-registered limited liability company founded in 2026 by Vasile Gangal. Oblivion Technologies builds privacy-first search, intelligence, and web infrastructure — anchored by OblivionSearch, a multi-engine private search platform.

Company details

Sister products: OblivionSearch (multi-engine private search), OBLIVION Mail (private email), OBLIVION Pages (landing pages), OBLIVION Analytics (privacy-first analytics), OBLIVION Support (AI customer support), OBLIVION AI (local-only LLM answers), and ~25 other privacy-respecting products — all under one roof.

Infrastructure: Dedicated servers (64 CPU cores, 393 GB RAM, 7.7 TB storage). Zero cloud dependency — no AWS, GCP, or Azure. Your queries never leave our hardware.

The SENTINEL quality gate

Every page published on SENTINEL must pass a hard quality gate before it enters our public sitemap. Pages that fail the gate are held back for re-enrichment or permanently excluded.

The 11-point quality checklist

1. Word count ≥ 400 unique words (not counting navigation or footer)
2. Unique-content ratio ≥ 35% (non-template tokens, measured via shingled hash comparison against sibling pages)
3. Similarity to any known source < 25% (MinHash Jaccard, computed against ingested source texts and all other SENTINEL pages)
4. At least 3 unique facts that are not present in the original source
5. At least 1 derived or computed value — a cross-reference, aggregation, distance calculation, or relationship graph
6. Explicit source attribution with license, access date, and ID
7. Valid structured data (schema.org Organization + typed schema per page)
8. Unique title, meta description, H1 (enforced at the database level)
9. Flesch Reading Ease 30-80 (rejects gibberish and unreadable prose)
10. Internal link graph — at least 4 internal links and 1 external authoritative reference
11. No AI content tells — failed by a deterministic pattern scanner for phrases, em-dashes, and transition overuse

Pages that fail any criterion are not published. Pages that pass are audited weekly via MinHash dedupe sweep; anything whose near-duplicate cluster size exceeds 5 is removed.

AI disclosure (important)

SENTINEL uses large language models (LLMs) running locally on our own hardware to generate plain-English summaries and enrichment text. This is disclosed transparently on every affected page.

How we use AI:

• Plain-English paraphrase of dense source material (e.g., patent abstracts, academic paper abstracts) — always shown alongside the original
• Extraction of named entities and fact tables from structured sources
• Generation of "about this entity" summaries from pre-validated structured data
• Readability scoring and quality gating

How we do NOT use AI:

• We do not fabricate facts — the model is only allowed to restate information already present in the source
• We do not auto-translate content into other languages
• We do not generate listicles, "top 10" content, or opinion pieces
• We do not use AI to create fake reviews, testimonials, or user-generated content
• We do not generate medical, legal, or financial advice

Every AI-assisted section on SENTINEL is labelled at the bottom of the page with the model name and generation date.

Data sources & current inventory

SENTINEL aggregates information from permissively-licensed, public-domain, and open-data sources. Every source is credited on the relevant page. See our full sources & licenses page.

• Places — GeoNames (CC-BY 4.0) — currently indexing 854,104 cities, towns, and administrative areas
• Books — Project Gutenberg (public domain, US) — currently indexing 76,883 classic works
• Directory — curated entries with human-written descriptions, currently indexing 1,816,417 entities
• Q&A — editorial questions with AI + human answers
• Alternatives — programmatic comparison pages built from curated directory data

What we will never do

• We will never publish content that is pure paraphrase of a single source without adding verifiable original value
• We will never auto-translate content into new languages without human editorial review
• We will never generate fake reviews, testimonials, or ratings
• We will never show different content to search engine crawlers than to human visitors (cloaking)
• We will never fake structured data (Review, FAQ, Product schema for content that is not genuinely those things)
• We will never ingest copyrighted content without a compatible license or fair-use rationale
• We will never publish YMYL (Your Money Your Life) content — no medical, legal, or financial advice
• We will never sell, rent, or broker your personal data to third parties
• We will never serve behavioral or retargeting ads based on tracking

Advertising & sponsored content — how we keep the lights on

SENTINEL is a bootstrapped business. We are honest that running dedicated servers, paying for bandwidth, and building new features costs real money. To keep the free tier genuinely free, we may show the following — all of it clearly labeled and privacy-respecting:

1. Paid subscriptions (primary revenue model)

Our preferred way to fund the platform. See /pricing. Pro users see fewer or no ads.

2. Non-personalized contextual ads

We may display display ads on the free tier from privacy-respecting networks (such as EthicalAds, CodeFund, or direct sponsors). These ads are:

• Contextual, not behavioral — targeted only by the topic of the page you are on, never by who you are
• Clearly labeled as "Advertisement" or "Sponsored"
• Served without RTB (no real-time bidding, no Google Ads, no Meta Audience Network)
• Never involve data sales — no personal data goes to the advertiser beyond the page topic

3. Sponsored directory listings (small fraction)

A small share of directory entries may be paid placements. Every sponsored listing is:

• Visibly badged as "Sponsored"
• Marked with rel="sponsored" on every outbound link, per Google's webmaster guidelines
• Not allowed to affect organic search results
• Forbidden for YMYL content (medical, legal, financial) and adult material
• Reviewed for accuracy — we do not publish sponsored content that misrepresents facts

4. Affiliate links (fully disclosed)

Where we link to an external product or service we may, in some cases, use an affiliate link that pays us a small commission if you make a purchase. All affiliate links are disclosed in the page footer and marked with rel="sponsored". Affiliate relationships never influence our editorial ranking or quality gate.

What this does NOT mean: we are not turning into an ad network. Our hard line stays: we never sell personal data, never serve retargeting ads, never use Google/Facebook trackers, and never run content that harms users. See our Privacy Policy for the complete promise.

Correction policy

If you find a factual error on SENTINEL, email [email protected] with the URL and the correction. We aim to review within 72 hours. Corrections are logged with a timestamp on the affected page.

If you hold the copyright to content that has been misused, please see our security.txt and the DMCA procedure described there.

Growth commitment

SENTINEL follows a deliberately slow and quality-first growth curve. We do not chase raw URL counts. Our target is sustainable indexation with zero algorithmic penalties — measured in clicks per indexed page, not pages per day.

If we ever receive a Google Search Console manual action notice, we will pause all publishing, diagnose the issue publicly, and post the remediation plan on this page.