Author’s Own Image

The Silent Auditor: Why I Built a “Zero-Knowledge” BBS in 2026

We broke the internet when we traded community for engagement. It’s time to trade it back.

Do you remember the “Old Internet”?

I don’t mean the dial-up noises or the waiting minutes for a JPEG to load (albeit, that is full of nostalgia!). I mean the feeling of it. The feeling that you were visiting a destination, not just scrolling through a feed. The feeling that the people you were talking to were… people. Not engagement metrics, not bots, and not “content creators” fighting for a slice of an algorithmic pie.

We lost that feeling somewhere between the invention of the “Like” button and the rise of the Infinite Scroll.

I built SimpleNet to find it again.

But I didn’t just want to clone a 1990s Bulletin Board System (BBS). I wanted to solve the two biggest problems that plague modern communities: Privacy and Safety. Usually, you have to pick one. If you want safety, you surrender your data to Big Tech moderation APIs. If you want privacy, you often end up in a cesspool of unmoderated toxicity.

I believed there was a third way. So, I built a “Safe Web” platform moderated by a Silent Auditor — an AI that lives in RAM, reads a post once, and then forgets it forever.

Here is the vision, and the code, behind SimpleNet.

The Philosophy: “The Country Club” & The Landlord

SimpleNet operates on a few radical premises that run counter to the current Silicon Valley playbook.

1. The “Country Club” Model

Most social platforms are free because you are the product. SimpleNet is paid-entry (via subscription) because you are the customer. This creates a “Country Club” effect:

• Zero Bots: The paywall is the ultimate CAPTCHA.
• High Trust: When everyone has “skin in the game,” the quality of discourse skyrockets.
• Calm UI: No notification dots. No “Who is Online.” No dopamine loops. You come here to write, read, and think.

2. The “Landlord” Model (True Data Sovereignty)

In the modern cloud, you are a digital serf. You don’t own your tweets, your posts, or your followers.

SimpleNet uses a Multi-Tenant SQLite Architecture. Every community (or “Board”) is stored in its own isolated SQLite database file. This isn’t just an implementation detail; it’s a bill of rights.

• The “Eject” Button: At any moment, a Board owner can click “Eject.” The system zips up their specific .db file and assets, and they can leave. They own the data, physically and legally.

The Tech: Under the Hood

I’m a software engineer, so let’s get technical. SimpleNet isn’t built on a precarious tower of microservices. It is built on the Boring Stack, which makes it fast, reliable, and portable.

• Language: Go (Golang) 1.24.
• Web Framework: Chi router with templ for type-safe HTML generation.
• Database: Pure Go SQLite (modernc.org/sqlite).
• Frontend: HTMX. No massive React bundles. Just HTML over the wire.

But the real engineering challenge was the moderation system.

The “Silent Auditor”: Privacy by Design

How do you moderate a community without reading everyone’s private messages?

The standard industry answer is: “Send everything to OpenAI or Google Perspective API.” I refused to do that. I did not want user data leaving the server, and I certainly didn’t want it training the next version of GPT.

My solution is the Silent Auditor.

How It Works

The Silent Auditor is a local AI agent that sits between the user’s POST request and the database. It is an "Ephemeral Witness".

1. The Intercept: When you hit “Post,” the backend holds your data in RAM. It does not write to the database yet.
2. The Audit: The content is sent to a local, open source AI Sidecar running on the same network (using models like Llama 3 for text and NudeNet/YOLO for images).
3. The Decision:

• If Safe: The AI returns a 200 OK. The content is immediately encrypted, written to disk, and the RAM is wiped. The Admin never sees the raw content.
• If Violation: The system triggers the “Break-Glass Protocol.” Only then is a redacted report generated for the Administrator.

The “Break-Glass” Protocol

We have a strict Community Code divided into two categories:

1. The Red Line (Zero Tolerance): Illegal acts, CSAM, real-world violence.
2. The Decency Standard: Harassment, trolling, nudity.

The AI is instructed to ignore political opinions, swearing, or “unpopular” ideas. It focuses purely on safety. If it flags a post, it generates a report where:

• Text is redacted to remove PII (Personally Identifiable Information).
• Images are heavily blurred.

This ensures that the Administrator is not snooping on conversations, but acts only when the “glass is broken” by a safety violation.

Why You Should Care

We are building a directory of these small, calm communities at **simplenet.directory**.

It’s an experiment. Can we build a place that feels like the old internet, but runs on the cutting edge of local AI? Can we have safety without surveillance?

I think we can.

I’m looking for early adopters — people who are tired of the noise — to come and help me test the limits of the Silent Auditor. Break it (gently). Try the Markdown support. Start a discussion that lasts for days, not minutes.

Come see what the web feels like when it’s simple again.

**Visit Simple.Net**

About the Author: I am a software engineer and a tech founder. SimpleNet is my passion project to reclaim the digital commons.