By Hap Nesbitt

Systems Architect & Founder, Arc Codex

Fort Collins, Colorado

We spend billions building hardware to defend physical borders, protect corporate perimeters, and intercept recognized threats in the material world. We invest heavily in threat intelligence because we understand a fundamental truth: if you do not protect your infrastructure, someone else will dismantle it.

Yet, our primary cognitive and informational infrastructure—the digital public square where human thought, discourse, and sanity actually live—is left completely unfortified.

Worse: it is actively weaponized against us.

The dominant business models of the modern internet do not profit from clarity, nuance, or resolution. They profit from engagement. And engagement is cheap to manufacture if you strip away the details, flatten every complex reality into a black-and-white culture war, and reward the loudest shouting match in the room. This kind of offense doesn't have mental health, civic stability, or truth anywhere in its top priorities. It relies on friction, outrage, and tribal division.

Every time an algorithm reduces a critical societal conversation to a binary screaming match, it precludes the fine details where actual understanding lives. It manufactures a perpetual state of low-grade civil conflict because conflict drives clicks.

We don't need another outrage engine. We need a defensive shield.
Releasing the Antidote: Free, Open-Source, and Local

Today, we are releasing the core architecture of the Arc Codex pipeline as a fully open-source project under the GNU license.

This isn't a platform designed to capture your eyeballs, track your movement, or monetize your attention. It is a defensive risk-mitigation tool for the mind. It runs on bare metal, on an oddball collection of old workstations sitting on a switch in your home office, or on a local server in a community center.

It is built to do the opposite of what the engagement economy demands:

De-escalation by Architecture: Incoming feeds are ingested, case-flattened to filter out synthetic bot noise, and analyzed not by a single, monolithic, opinionated model, but by a competing quorum of smaller, localized humanist-instructor agents.

Adversarial Cross-Examination: Using the A.R.C. framework, multiple analytical personas (Blue, Red, and Purple) interrogate incoming narratives, check for cognitive anti-patterns, and demand rigorous grounding before an article is synthesized.

Frictionless, Ad-Free Learning: Stories are rewritten at a clean, 12th-grade reading level optimized for speaking aloud—stripped of corporate jargon and hyperventilating headlines.

Sovereign Narration: Neural text-to-speech (Kokoro) renders the output locally into continuous, listenable audio bulletins. No cloud billing, no data harvesting, no third parties.

Built for the People, Not the Feed

You don't need a fiber-optic data center or a venture-capital war chest to run it. You just need a machine, an internet connection for RSS ingestion, and the willingness to take back control of your own information diet.

If Alphabet or the legacy tech giants aren't going to build a calm, educational public utility that serves human well-being rather than ad-clicks, we will build it ourselves. We will distribute the nodes so wide that no central server can be turned off, and no algorithm can gatekeep the truth.

The code is live. The architecture is proven.

Go get some.

Source and deployment instructions are available on GitHub.