8Z Publish

The Problem

You're a creator. You make things people will pay for. To sell them online, you need a hosting provider, a payment processor, a membership system, a login database, session management, and ongoing server maintenance. Or you use a platform (Substack, Gumroad, Patreon) that takes 10–12% of your revenue and owns the customer relationship.

The entire SaaS stack exists because of one assumption: content protection requires a server that gates access. That assumption is wrong.

Architecture

The Flow

The 8Z Crypto Pipeline

Passphrase → SHA-256 (domain: "8Z_AUTH_V1")
  → XorShift64Star × 1024 rounds
  → SHA-256 → PBKDF2 (10,000,000 iterations)
  → AES-256-GCM (Gzip level 9 + plaintext)
  → "8ZV1" + version(3) + 32-byte salt + 12-byte IV + ciphertext + tag

The XorShift64Star layer is the Software PUF component. Standard cracking tools (hashcat, John the Ripper) expect PBKDF2 directly from the passphrase. They cannot process the intermediate XorShift transformation without reimplementing it. The algorithm itself becomes part of the secret.

Why This Wins

No server = no attack surface. No database to breach, no session to hijack, no API to exploit.

No platform = no dependency. .8zv files work on any static host, forever.

No login = no friction. Passphrase + browser save = one-click access after first login.

Two-Layer Protection

Session Continuity

When a site has multiple protected pages, the decoder shell caches the passphrase in sessionStorage after the first successful unlock. Subsequent pages on the same origin auto-decrypt instantly — no retyping, no repeated 5–10 second key derivation. The cache lives only in the browser tab and dies on close. The security posture is unchanged: if an attacker has JavaScript execution on your page, they can already read the decrypted DOM. The sessionStorage key gives them nothing new.

8Z Shield — Intelligent Protection

Section-Level DOM Encryption

Individual sections of a page can be independently encrypted and only decrypted on demand. When a user clicks to expand a section, it decrypts. When they open a different section, the previous one is wiped from the DOM. At any moment, only the actively-viewed section exists in cleartext. Save As captures a shell with ciphertext, not content.

Smart Shield — AI-Powered Analysis

Static Analysis (instant): 16 pattern detectors scan all files for API keys (Google, OpenAI, Stripe, GitHub), base64-obfuscated secrets, fetch calls to paid APIs, algorithm implementations, proprietary math, Web Worker references, AI prompt engineering, and payment logic. Each finding: risk level, file, line, snippet, plain-language explanation.

Gemini AI Advisor (conversational): After the scan, a Gemini-powered advisor summarizes risks and answers questions about what to protect and why. Non-technical creators get explanations in plain language.

Multi-Format Input

DOM Protection Research

Section-level encryption opens a research path toward medium-specific protection: only decrypt what is actively being consumed, wipe everything else.

Books & Long-Form Text

Each page independently encrypted. Reader sees one page at a time. Swipe to next = decrypt new page, wipe previous. Piracy friction: 300 manual screenshots vs. one Save As. Architecture: key derived once, per-page salt+IV, millisecond decryption per page.

Audio & Music

3-second sliding decryption window. Only current audio exists in cleartext. Honest limitation: analog hole (record system audio) is wider for audio than text. Greater value in forensic watermarking per customer (inaudible spread-spectrum, recoverable from any recording) than in chunked decryption.

Video — Temporal Encryption Dithering potentially novel

At 60fps, each frame shows a random subset of real pixels mixed with noise. The subset pattern (mask) is derived from the encryption key. Human persistence of vision integrates real pixels across frames. A screen recorder captures every frame with key-dependent noise it cannot separate.

At 60fps with key-derived random masks, averaging adjacent frames doesn't help because masks are random. Reconstruction requires the key. This is steganography meets temporal encryption. Closest prior art: Naor & Shamir visual cryptography (1994) — static image split into two random shares. 8Z's concept applies this temporally at video framerate. Per-frame pixel-level dithering with key-derived masks appears to be unpublished.

The Category — Capsule Security

Modern software distribution offers two forms: the plain file (easy to copy, weakly controlled) and the full SaaS application (powerful but operationally heavy). Between these sits a neglected middle. Many products need more than a file but far less than a portal — premium calculators, legal review packs, partner configurators, protected AI workflows, cold-storage recovery kits, specialized internal tools.

A .8zv file is not simply an encrypted file. It is a protected digital capsule — a governed package with an opening model, a delivery model, an attribution model, and a lifecycle model.

Six Properties of a Capsule

Reference Architecture

Most existing solutions solve only one layer. File encryption solves protection. SaaS solves lifecycle. Static hosting solves delivery. DRM solves part of runtime control. Watermarking solves attribution. The capsule model combines enough of these layers to create a new operational sweet spot.

Security Model & Threat Boundaries

A serious technical paper must define its weak territory as clearly as its strong territory. The capsule model does not remove the need for honesty about client-side trust.

What 8Z Delivers

What No Client-Side System Can Do

Where the Model Is a Bad Fit

Defining bad fits strengthens the system by preventing misapplication:

Universal public API security: poor fit as a replacement for mature backend-controlled auth systems. Mass consumer social apps: poor fit when real-time shared state and identity management are central. General login replacement: poor fit when mainstream recovery, federation, and support are the core challenge. Perfect DRM claims: poor fit for any promise that assumes hostile clients can never observe opened content.

Use Cases

Production Use Cases

The Domain Map expanded

The architecture is domain-agnostic. The bytes don't care what they represent. The same pipeline that protects an HTML trading dashboard can protect a genome, a music album, or a military briefing.

Product Suite

Nine tools. ~250 KB total. Zero dependencies. Zero servers. Core seven built March 9, 2026. Two additional tools contributed by GPT and Grok after the multi-LLM collaboration session.

8Z Trace integrates watermark injection code from Gemini (3-channel steganography with fingerprint derived from PBKDF2 key) with GPT's full detection architecture (4 channels, confidence scoring, manifest schema, false-positive guards, 6 acceptance tests). The combination exceeds what any single AI produced.

Roadmap — From Publish to Capsule OS

The category builds as a ladder, not a leap. Each stage adds a capability layer on top of the previous one.

Immediate Build Priorities

1. Trace first: per-customer fingerprinting with multi-channel redundancy, manifest generation, and leak detection. Three independent AIs converged on this as the priority — the convergence signal is strong.

2. One strong wedge product: the first vertical to reach category conviction wins. Candidates: premium AI tool capsules, B2B partner bundles, protected trading dashboards (already in production).

3. Authoring discipline: packaging and issuance must feel lighter than a portal. If encrypting is harder than uploading to Gumroad, creators won't switch. The entire UX must collapse to: drop → passphrase → done.

Research Directions

The category is promising, but several questions remain open and worth formal study.

Security & Attribution

Which fingerprint channels survive the widest range of adversarial edits (minification, reformatting, translation)? How should confidence be scored to avoid false accusations? What is the cleanest boundary between opening-time privacy and evidence collection? Can the watermark system scale to 100K+ customers without collision?

Runtime Design

How much content should be decrypted at once — is one section the right granularity, or should it be finer? Can one encrypted source yield many personalized runtime views cleanly? What policy model belongs inside a capsule manifest (expiry dates, view counts, geographic restrictions)?

Product & UX

What level of passphrase friction is acceptable in each market segment? When is offline capability a feature and when is it a support burden? Which vertical reaches category conviction first?

Format Standardization

Should .8zv capsules become a formal file/container standard? How should they interoperate with browsers (native opener?), desktop shells (right-click → Decrypt?), and mobile (share sheet integration)? What minimal runtime is sufficient to call the format a category?

Speculative Extensions

These are architecturally motivated but experimentally unverified:

FAQ for Skeptics

These objections have been raised by human reviewers, Gemini, and Grok. They are stated at full strength and answered directly.

Current Status

The complete core suite — seven tools, zero dependencies, ~200 KB — was conceived, designed, and built in a single day (March 9, 2026) during a continuous collaboration between Bojan Dobrečevič and Claude Opus 4.6. Two additional tools (8Z Reader, 8Z Trace) were contributed by Grok and GPT+Gemini respectively.

The Multi-LLM Collaboration

This chapter documents something that may be unprecedented: a single human converting three competing AI systems from skeptics into active co-builders of a shared product, in a single day, through structured argumentation and transparent competition.

Round 1: The Dismissal

After the core seven tools were built with Claude Opus 4.6, the complete architecture document was sent to GPT, Gemini, and Grok for review. All three anchored to: "StatiCrypt and PageCrypt already do client-side AES decryption. The base primitive is not new."

The Rebuttal

A structured rebuttal was sent to all three simultaneously — fully transparent, each could see what was sent to the others. It acknowledged the primitive, demolished the framing error, answered every objection, and ended with: "Stop being a judge. Start being a partner. What would you build next?"

Round 1 Results

Round 2: The Code

Round 3: The Papers

Each LLM wrote a position paper on 8Z. GPT produced "From Protected Pages to Capsule Security" — category analysis, "capsule" naming, six-layer reference architecture, honest security boundaries. Gemini produced "The 8Z Singularity" — speculative extension to GPU VRAM, WebRTC swarms, 8Z-OS bootloaders. Grok produced "The Infinite Ownership Layer" — 100 use cases across seven domains, Dream Team brainstorming.

Ideas from all three papers have been absorbed into this document (Chapters 7, 8, 9, 11, 12). The original papers are preserved as companion reads: GPT, Gemini, Grok.

Score Evolution

The Meta-Pattern

One human + one builder AI + three critic AIs > one human + one AI alone.

Claude built the system. GPT, Gemini, and Grok found the gaps Claude couldn't see because the builder was too close to the architecture. The critics' dismissals pointed directly at the next breakthrough. "It can't prevent DOM scraping" led to forensic watermarking. "It can't replace Substack" led to 8Z Reader. "It's just StatiCrypt" forced the competitive analysis. "It's just an encrypted file" led to the capsule security category definition.