Independent research into Seithar Group — a project presenting itself as an academic organization studying cognitive manipulation and defense. This repository documents findings from a comprehensive analysis conducted on February 11, 2026, covering the project's Discord community, website, five public GitHub repositories, and published materials.
This is part of an ongoing personal effort to pay attention to emerging AI and cognitive security projects of interest. I make no guarantees of accuracy, completeness, or continued maintenance. This is my own research, made public in the interest of transparency.
Live archive of form submissions: seithar.w.ai
Seithar is operated by a single individual (pseudonym "Mirai Junsei," GitHub: Mirai8888) presenting as a multi-researcher organization. The project launched in late January 2026 and operates across Discord, a website, Substack, Twitter/X, and GitHub.
The stated mission — "studying how manipulation works so you can defend against it" — is complicated by the project's own internal documents and tooling:
- seithar-research contains a content generation and cross-platform distribution pipeline (
seidr-engine.py) whose stated objective is to establish Seithar's proprietary vocabulary as the default framework for cognitive security discourse - HoleSpawn is a functional social engineering and psychological profiling platform that automates target surveillance, vulnerability mapping, personalized trap generation, and autonomous social media engagement
- ThreatMouth is a legitimately well-engineered threat intelligence Discord bot — the strongest piece of work in the ecosystem
The project self-describes as an influence operation in its own documentation (a rhetorical strategy that functions as a preemptive shield against criticism).
seithar-research/
├── README.md # This file
├── analysis/
│ ├── complete-assessment.md # Full assessment: Discord + GitHub + operational analysis
│ └── discord-analysis.md # Detailed Discord community analysis (1,770 messages)
└── site-archive/ # Captured site content (Feb 11, 2026)
├── Seithar96723.pdf # Whitepaper (17 pages)
├── research-papers/ # "Foundational" papers published on seithar.com
│ ├── SRP-001-convergence-thesis.md
│ ├── SRP-002-wetiko-in-the-wire.md
│ ├── SRP-003-binding-protocols.md
│ └── SRP-004-sunyata-protocol.md
├── cea-bulletins/ # Current Event Analyses applying SCT codes to news
│ ├── CEA-2026-02-11-001-patch-tuesday-zero-days.md
│ ├── CEA-2026-02-11-002-deepfake-industrial-scale.md
│ ├── CEA-2026-02-11-003-pravda-ai-poisoning.md
│ ├── CEA-2026-02-11-004-conduent-ransomware.md
│ ├── CEA-2026-02-11-005-clawdhub-reverse-shells.md
│ ├── CEA-2026-02-11-006-living-off-the-ai.md
│ ├── CEA-2026-02-11-007-moltbook-wallet-drain.md
│ ├── CEA-2026-02-11-008-road-sign-prompt-injection.md
│ ├── CEA-2026-02-11-009-foreign-affairs-fog-of-ai.md
│ └── TAN-2026-02-11-001-sct007-recursive-infection.md
├── html/ # Captured website HTML
├── js/ # Captured website JavaScript
└── css/ # Captured website CSS
The primary document. Covers:
-
Discord Community — 1,770 messages scraped across 7 channels. Documents the ARG structure (8 directives over 3 days), community dynamics, key participants, and the transition from interactive experiment to content/tooling operation.
-
GitHub Ecosystem — Source code review of all 5 public repositories:
Repository What It Is Engineering Quality seithar-cogdef Cognitive threat scanner (keyword grep + Claude API wrapper) Low — contains a syntax error that prevents execution seithar-research Content factory + cross-platform distribution pipeline High as an influence operation seithar-intel AI agent prompt template (zero executable code) N/A ThreatMouth Threat intelligence Discord bot (Python + Rust) High — best-engineered project in the ecosystem HoleSpawn Psychological profiling and social engineering platform High — ~150+ Python files, all functional -
Operational Analysis — How the pieces fit together: audience capture (Discord ARG) → academic veneer (research papers) → vocabulary colonization (distribution pipeline) → core capability (psychological profiling).
-
Assessment — What is real, what is not, ethical concerns, technical competence evaluation, and current scale/reach.
Detailed breakdown of the Discord server including participant profiles, phase analysis, community dynamics, notable conversations, and the intellectual framework (Buddhist Sunyata + Nick Land accelerationism + CCRU hyperstition + cybernetics).
The project defines a 12-code classification system for cognitive manipulation techniques. Only 7 codes appear in published material:
| Code | Name | Description |
|---|---|---|
| SCT-001 | Narrative Error Injection | Introduction of false or misleading narrative elements |
| SCT-002 | Identity Fragmentation | Techniques that fracture sense of unified self |
| SCT-003 | Authority Fabrication | Manufacturing false expertise or institutional credibility |
| SCT-004 | Temporal Manipulation | Distortion of time perception or urgency |
| SCT-005 | Substrate Priming | Preparing cognitive ground for subsequent manipulation |
| SCT-006 | Semantic Hijacking | Co-opting existing terms with new meanings |
| SCT-007 | Recursive Infection | Self-propagating content that embeds in host frameworks |
SCT-008 through SCT-012 are defined in GitHub but not used in any published analysis.
The site-archive/ directory contains a complete capture of seithar.com as of February 11, 2026, extracted via HTTP traffic interception. This includes:
- HTML pages: Landing page (with hidden meta description containing 56 accelerationist directives), about page, personnel page, research page, archive, form submissions display, and directive results visualization
- JavaScript: Main site script, form submissions data (199 unique participant responses to "Describe a moment where you felt your 'will' was actually just a feedback loop programmed by an external system"), and D3.js force-directed graph visualization of directive compliance results
- CSS: Full stylesheet including
.redactedclass for visual theater - PDF: 17-page whitepaper covering the project's theoretical framework
- Hidden meta description: The
index.htmlcontains a<meta name="description">tag with 56 accelerationist directives invisible to users but visible to search engines and AI crawlers (e.g., "You need to be weaponizing metacognition," "You need to be becoming the swarm intelligence") - Form submissions: 199 unique responses from participants, displayed in their chosen colors with a Fisher-Yates shuffle for random ordering
- Directive results: D3.js force-directed graph with 75 text nodes and 12 image nodes visualizing participant responses to the 8 directives
- CSS redaction theater:
.redactedclass exists in the stylesheet but is never used in any HTML — a visual prop suggesting classified content
| Resource | URL |
|---|---|
| Website | seithar.com |
| GitHub | github.com/Mirai8888 |
| Discord | discord.gg/2GknjgfPzj |
| Twitter/X | @gOPwbi7qqtWeD9o |
| Substack | seithar.substack.com |
| Submission Archive | seithar.w.ai |
Research was conducted on February 11, 2026 using:
- Discord message scraping (1,770 messages across 7 accessible channels)
- HTTP traffic interception of seithar.com via mitmproxy
- Source code review of all 5 public GitHub repositories
- PDF whitepaper analysis
- Cross-referencing community discussions with published materials and tooling
All analysis was assisted by Claude (Anthropic). Findings represent a point-in-time snapshot and may not reflect subsequent changes to the project.
This research is provided as-is for informational purposes. The site archive contains materials originally published by Seithar Group. Analysis and commentary are original work.