Methodology & Definitions
How content on 4Orbs is extracted, classified, and organized. This page defines every scoring system, content type, and editorial standard used across the site.
How It Works
Every piece of content on 4Orbs flows through a four-stage pipeline. Videos published by Ashton Forbes are analyzed using AI-assisted extraction, then normalized, deduplicated, and editorially reviewed before publication.
Video Analysis
Transcripts are extracted from YouTube videos and analyzed by LLMs to identify claims, theories, people, evidence, references, and glossary terms.
Structured Extraction
Raw analysis is converted into structured JSON with normalized categories, slugs, and cross-references between content types.
Deduplication
Duplicate and near-duplicate entries are merged. Existing theories are updated with new evidence rather than creating parallel entries.
Editorial Review
Content is reviewed for accuracy, enhanced descriptions are written, investigation statuses assigned, and cross-links verified before publication.
Who Writes This
Two different things happen on this site, and they deserve separate answers.
The video archive is machine-extracted. Every claim, theory, person and glossary entry in the library comes out of the pipeline above: a language model reads the transcript and proposes structured entries, which are then normalised, deduplicated and reviewed before they go live. Nobody hand-writes thousands of claim records.
The research deep-dives and the 101 guides are written with AI assistance as well. A human sets the question, chooses the sources, and signs off on what publishes. The drafting and the structural editing are done by a large language model working to a written editorial process. So the sentences you're reading here were, in most cases, machine-drafted and human-directed.
We'd rather tell you that than have you find out.
What it doesn't change
The evidentiary bar. Every DOI, arXiv ID, patent number and document reference on a research page is resolved and read before it ships. Author lists come from the published record. Claims attributed to Ashton Forbes are checked against the video transcript. New articles go through a fact-check pass run by a reviewer with no involvement in the draft, and nothing publishes with an open finding. When we get it wrong anyway, it goes in the corrections log.
What it should change for you
Where you put your trust. Provenance is the weakest reason to believe anything here, and that would still be true if every word were typed by hand. Trust the parts you can check: follow the DOI, pull the FOIA release, read the patent, watch the timestamp. This site is built to be verified by a reader who assumes we're wrong. That's the standard we'd want applied to anyone making these claims, so it's the one we publish under.
How Content Types Connect
Content Types
Theories
927Structured hypotheses extracted from video analyses. Each theory has a name, description, supporting points, cluster assignment, and tags. Theories are the primary unit of argumentation on the site.
Browse theories →Claims
1,890Specific factual assertions made in videos, each with cited evidence and a confidence rating. Claims are the atomic building blocks that support theories, surfaced within each theory and evidence entry.
Evidence
1,102Documents, footage, testimony, patents, data, and other artifacts cited as proof for claims. Each piece of evidence is typed and linked to its source video.
Browse evidence →References
1,306Scientific topics, papers, and research programs mentioned across videos. Each reference is categorized by field and mainstream acceptance status.
Browse references →Key Figures
759Individuals mentioned across the video archive: scientists, whistleblowers, officials, and researchers. Each entry tracks their appearances, roles, and context.
Browse key figures →Glossary
1,470Technical terms and jargon defined in plain language. Each term is tagged with its field (physics, aviation, intelligence, etc.) for cross-referencing.
Browse glossary →Confidence Levels
Every claim is assigned a confidence level indicating how well-supported it is by available evidence. This is an editorial assessment, not a truth claim.
Independently verifiable documentation, multiple corroborating sources, or public records. Facts not in dispute.
Credible evidence not independently verified to definitive standard. Single authoritative source or partially corroborated.
Inference, pattern matching, or theoretical arguments beyond what evidence directly supports.
Mainstream Status
Every scientific reference is tagged with its level of mainstream acceptance. This indicates how the broader scientific community views the research, not our editorial position.
Peer-reviewed, taught in universities, scientific consensus.
Active legitimate research with some peer-reviewed support.
Limited peer-reviewed support, extrapolates beyond established results.
Government programs known to exist but details restricted.
Evidence Types
Every piece of evidence is classified by type. These 12 categories describe the form of the evidence, not its strength.
Theory Clusters
Theories are organized into five thematic clusters. Each cluster has a semantic color used consistently across the site for visual identification.
Advanced fusion architectures, zero-point energy, quantum vacuum physics, and electromagnetic unification theories.
The disappearance of Malaysia Airlines Flight 370, from satellite footage and plasma orbs to field-reverse configuration physics and teleportation theories.
Government suppression of breakthrough technologies, classification programs, economic implications, and the disclosure movement.
The intersection of consciousness, quantum physics, holographic universe theory, and the nature of reality.
Specific engineering implementations, reactor designs, patents, and military technology applications.
Investigation Status
An editorial enhancement applied to reviewed theories. 243 of 927 theories have been assessed so far.
Theory is under active investigation with ongoing developments. New evidence is being gathered or analyzed.
Substantial corroborating evidence exists from multiple independent sources. Core claims are well-supported.
Supporting evidence exists but is indirect. The logical chain connecting evidence to conclusions has gaps.
Limited direct evidence. The theory relies primarily on inference, pattern matching, or extrapolation.
Significant disagreement exists within the research community. Credible arguments on multiple sides.
Data Provenance
Source
The archive starts from a curated video corpus, primarily the published work of Ashton Forbes on YouTube. From there, longer-form research deep dives cross-reference and verify against independent primary sources: SEC filings, patents, government contracts, FOIA-released documents, peer-reviewed papers, court records, and official press releases.
Method
AI-assisted extraction from video transcripts, followed by normalisation, deduplication, and editorial review. Research deep-dive articles are AI-drafted and human-directed, and cite independent primary sources directly; see Who Writes This. A weekly currency-drift check re-verifies time-sensitive claims against third-party sources.
Limitations
The video corpus carries a curatorial bias: coverage reflects one investigator's research priorities, even when individual claims inside research articles are independently sourced. AI extraction may introduce errors or misattributions. Content is not peer reviewed. Confidence ratings are editorial assessments, not verified truth claims.
Purpose
A structured research archive designed to make the body of work searchable, cross-referenced, and accessible for independent investigation and analysis.
Related Topics
Browse all content types: videos, theories, claims, evidence, references, glossary, and key figures.
Interactive force-directed graph showing how theories connect through shared tags and evidence.
Detailed evidence evaluation for the MH370 investigation, the most thoroughly assessed cluster.