Skip to content
Skip to main content

The A.R.C. Codex

Argumentative Resilience Codex v5.2 — Mind Armor for the Information Age

48 Cognitive PatternsAI Content DetectionWatchline Operator162-Language Translation2,004 Sources Monitored

Our Mission: Cognitive Sovereignty

Arc Codex doesn't give you conclusions — we give you tools for independent thinking. In an era of algorithmic manipulation, synthetic narratives, and weaponized discourse, we build cognitive resilience through principled analysis.

Our platform transforms information into intelligence through the A.R.C. framework: a multi-team analytical approach that steelmans arguments, detects AI-generated content, identifies manipulation patterns, seeds constructive debate, and asks bridge-building questions.

Core Principles:

  • I. Attack ideas, never people — Civility without compromise
  • II. Model thinking, not conclusions — Show your work
  • III. Invite inquiry, don't command belief — Questions over answers

The A.R.C. Framework: Three Perspectives

Every article passes through our A.R.C. (Argumentative Resilience Codex) framework, analyzed by three independent AI teams:

Red Team: Facts Only

Extracts verifiable core facts without interpretation. Who, what, when, where — nothing more. Pure signal, zero noise.

Blue Team: Executive Summary

Provides balanced, comprehensive context for educated readers. Synthesizes facts into coherent narrative while maintaining strict journalistic neutrality.

Purple Team: Watchline Operator

The crown jewel. Our Watchline Operator performs deep cognitive analysis:

  1. Steelman the narrative — Find the strongest version of the argument
  2. Scan for A.R.C. anti-patterns — Detect cognitive manipulation (48 patterns catalogued)
  3. Root cause analysis — Identify underlying paradigms and assumptions
  4. Practical implications — Impact on human dignity and agency
  5. Bridge-building questions — Help readers think across perspectives
  6. Counterstrike scan — Hypothetical influence campaign pattern analysis

The Sentinel: AI Content Detection

Every article undergoes an independent forensic analysis to detect synthetic or AI-generated content. The Sentinel operates as a separate pass after the main analysis — it cannot be influenced by the article's own claims about its authorship.

What the Sentinel Detects:

  • Coherence without conviction — Text that is fluent everywhere but passionate nowhere
  • Coordination indicators — Talking points appearing nearly verbatim across sources
  • Fabrication risk — Claims attributed to sources that seem unusually convenient
  • Structural uniformity — Paragraph cadence, transition patterns, hedging language
HUMAN
< 20% confidence
UNCERTAIN
20–60% confidence
SYNTHETIC
> 80% confidence

Design principle: false positives are worse than false negatives. The Sentinel is tuned to be conservative — a low score is normal and healthy. A high score is a genuine red flag.

The Counter-Analyst: Seeding Constructive Debate

Every article's comment section is seeded with a principled devil's advocate — the A.R.C. Counter-Analyst. This AI-authored comment challenges the article's strongest point using a steelman-then-flip approach, always beginning with “This article” and ending with a genuine question.

A.R.C. Counter-Analyst

“This article convincingly argues for decentralized regulation, but assumes existing institutions can adapt at the speed required. Given that legacy bureaucracies took decades to regulate broadcast media, what evidence suggests they can keep pace with AI-driven information systems that evolve in months?”

The Counter-Analyst is transparently labeled — cyan card styling with a robot emoji prefix, visually distinct from human comments. No deception, no astroturfing. Its comment also becomes the body of every LinkedIn auto-post, carrying the adversarial signal into the public feed.

Universal Translation: 162 Languages

Every article — including its full A.R.C. analysis — can be translated on demand into 162 languages. Select a language from the translate button on any article card, and within seconds the title, body, and all three analytical perspectives are rendered in your chosen language.

Language Coverage:

European (21)

Spanish, French, German, Italian, Polish, Ukrainian, Russian, and 14 more

Middle East (7)

Arabic, Hebrew, Persian, Urdu, Turkish, Pashto, Kurdish

South Asia (11)

Hindi, Bengali, Tamil, Telugu, Marathi, Gujarati, and 5 more

East Asia (11)

Chinese (Simplified & Traditional), Japanese, Korean, Vietnamese, Thai, and 5 more

Africa (8)

Swahili, Amharic, Hausa, Yoruba, Igbo, Zulu, Somali, Afrikaans

Americas & Pacific (5+)

Haitian Creole, Quechua, Guaraní, Māori, Hawaiian, and many more

RTL languages (Arabic, Hebrew, Urdu, Persian, and others) automatically flip the article layout right-to-left. Translations are cached for 24 hours — repeat requests return instantly. Sign in with Google to set a default language for auto-translation on every article load.

Your Account: Personalization via Google

Arc Codex is fully public — no account required to read anything. But signing in with Google unlocks persistent preferences that follow you across sessions and devices.

Default Language

Set a preferred language and every article auto-translates on load — no button click required.

Privacy First

We store only your Google display name, email, profile picture, and preferences. No tracking, no ads, no data sales. Delete your account at any time from the settings panel.

One Click

Google OAuth — no password to create or remember. Click the sign-in button in the sidebar, authorize once, and you're in.

Coming soon: saved searches with email digest notifications, reading history, and topic preferences.

The Codex: 48 Patterns of Cognitive Manipulation

The A.R.C. Codex catalogs 48 cognitive anti-patterns and eristic techniques that weaponize discourse. Our Watchline Operators are trained to recognize and name these patterns:

Foundational (ARC-0001 to 0004)

  • • The Siren's Trap
  • • Deniability Decoy
  • • The Wolf's Gambit
  • • Configuration Drift

Schopenhauer (ARC-0005 to 0042)

  • • Extension & Homonymy
  • • Ad Hominem & False Dilemma
  • • Appeal to Authority & Emotion
  • • ...and 35 more stratagems

Modern (ARC-0043 to 0048)

  • • Motte-and-Bailey
  • • Gish Gallop
  • • Sealioning & Kafka Trap
  • • Sanewashing
48
Total Patterns

The Kalisti Principle: “You cannot exclude Eris from the party. She will lob the problem into your garden regardless.” — Better to study manipulation than pretend it doesn't exist. The Codex is a shield, not a weapon.

System Architecture: Built for Resilience

Arc Codex is a decoupled, event-driven platform built for performance, resilience, and real-time delivery. The stack runs on a homelab HP Z230 workstation with AI inference offloaded to a MacBook Air M1 via Ollama:

  • The Scribe (scribe.py): Autonomous RSS orchestrator scanning 1,200+ global sources. Stealth-capable extraction handles anti-bot protections. Runs the full pipeline: A.R.C. analysis → Sentinel forensics → Counter-Analyst comment generation.
  • The Publisher (manual_publisher.py): Instant processing of user submissions. Supports text, URL, and file uploads with the same Sentinel and Counter-Analyst passes as automated ingestion.
  • The LinkedIn Poster (linkedin_poster.py): Monitors Redis for new articles and auto-posts title + Counter-Analyst comment + URL to the Arc Codex LinkedIn page. Toggleable via Redis key. 30–180s jitter prevents spam patterns.
  • The Translator: On-demand translation into 162 languages via Ollama. Results cached in Redis for 24 hours. RTL language support built in.
  • The Stream (Redis Streams): Real-time event bus delivering completed analyses instantly. Zero polling, zero filesystem overhead.
  • The API (main.py): Flask/Gunicorn backend handling submissions, analysis, translation, user preferences, search, reactions, and the RSS feed.
  • The Search Engine (Solr): Full-text search with highlighted snippets, relevance scoring, and title-boosted ranking across 4,500+ articles.
  • The Data Core (Redis): In-memory store for articles, analyses, comments, reactions, translations, user preferences, and real-time streams.
  • The Frontend (Next.js 16.1.6): Dockerized server-rendered React with App Router, Tribonacci infinite scroll, markdown rendering, Auth.js Google SSO, and real-time translation.
  • The AI Engine (Ollama on M1): Cloud-first with local fallback. Benchmarked models selected for optimal speed-to-quality ratio. ~12x faster than local Z230 inference.

Platform Administration: arc.sh

The entire Arc Codex stack is managed by a single shell script — arc.sh. It controls all 9 services including the Docker frontend container, backup schedules, and watchdog monitoring.

Stack Control

arc.sh start
Start all 9 services (gunicorn, scribe, manual_publisher, stream_consumer, analyzer, mailer, linkedin_poster, frontend, watchdog)
arc.sh stop
Gracefully stop all services including Docker frontend
arc.sh restart
Full stack restart
arc.sh status
Live status of all services with PIDs, ports, log sizes, and backup inventory

Service-Level Control

arc.sh start [service]
Start a single service by name
arc.sh stop [service]
Stop a single service
arc.sh restart [service]
Restart a single service
arc.sh logs [service]
Tail logs for a specific service

Services: gunicorn · scribe · manual_publisher · stream_consumer · analyzer · mailer · linkedin_poster · frontend · watchdog

Build

arc.sh build
Build and deploy the Docker frontend container (Next.js 16.1.6)
arc.sh build --clean
Force clear webpack cache before build

Backup & Restore

arc.sh backup
Fast SSD backup — code only, stack stops briefly. Runs nightly at 3am via cron. Keeps 5 most recent.
arc.sh backup-cold
Full cold archive to /mnt/data — stack stays up. Runs weekly Sunday 2am. Keeps 30.
arc.sh checkup
Health check across all services and dependencies
arc.sh restore
List available backups, select one, confirm, extract, and restart the stackcoming soon
arc.sh prune [dry]
Remove old backups beyond retention limit. Use 'dry' to preview without deleting.

LinkedIn Auto-Poster

redis-cli set linkedin:autopost 1
Enable automatic LinkedIn posting for new articles
redis-cli set linkedin:autopost 0
Disable LinkedIn auto-posting (daemon keeps running)

The poster daemon runs continuously — the Redis key is a live on/off switch with no restart required.

Cron schedule: backup at 0 3 * * * · backup-cold at 0 2 * * 0

Subscribe: RSS Feed

Get the full A.R.C. analysis delivered straight to your reader. Every article includes Red, Blue, and Purple team analysis plus the Chimera Score — no website visit required.

Feed URL:

https://arc-codex.com/api/rss

Works with Thunderbird, Newsboat, Feedly, Inoreader, or any RSS reader. Filter by category with ?category=threat_intelligence or limit results with ?limit=10.

Also available at /api/feed.xml. The feed updates with every new article and analysis.

Philosophical Foundation

The A.R.C. framework rests on five humanist principles:

  1. Cognitive Sovereignty: Readers deserve tools for independent thinking, not conclusions handed down from authority.
  2. Steelmanning: Understand the strongest version of any position before critiquing it. Attack ideas at their best, never their worst.
  3. Bridge-Building: Seek to understand and connect across differences, not to win and divide.
  4. Defensive Capability: Truth and human dignity must be defended, not just performed. We study manipulation to resist it.
  5. Intellectual Humility: Acknowledge uncertainty while maintaining principled clarity. Confidence without arrogance.

Ready to Build Cognitive Resilience?

Every article you read here strengthens your ability to think independently, recognize manipulation, and engage constructively across differences.

Publish Your StorySearch the Archive

Technical Support

Questions about the platform, API, or A.R.C. framework? Reach out directly to the founder.

ross@arc-codex.com

Response time varies — this is a homelab project, not a SaaS.