Executive Summary Seed Round 2026Confidential

Verifiable Authorship
Infrastructure for the AI Era

An open-standard platform that embeds verifiable authenticity directly into the creative process — making human effort visible and AI use transparent through cryptographic evidence. Built on the open WritersProof Protocol.

David Condrey

Founder & CEO, WritersLogic

Schedule a Call View Pitch Deck

The Platform

WritersProof is not an AI-detector; it is a cryptographic provenance engine. It captures the full behavioral record of writing -- keystrokes, paste attestations, revision depth -- making human effort visible and AI use transparent.

How it Works: The Hash Chain

At its core, WritersProof relies on a Continuous Audit loop. As a user writes, the system captures the exact state of the document at discrete intervals. Instead of saving raw text, the system compiles the data into a dense, deterministic binary payload (Bincode) and hashes it using SHA-256.

Crucially, each new hash mathematically includes the previous hash. If a bad actor tries to paste an AI-generated essay into the middle of the document, the entire cryptographic chain shatters because the historical hashes will no longer align.

The 5-Way Entanglement

To make a state-hash truly immutable and legally binding, it cannot just hash the text. WritersProof entangles five distinct data vectors into a single cryptographic block:

1. The Identity: The author's public key or verified identity assertion.
2. The Chronology: A secure timestamp proving when the work occurred.
3. The Payload: The actual text, code, or differential state (the edits made since the last block).
4. The Ancestry: The hash of the immediate prior state, locking the chronological sequence in place.
5. The Nonce: The Proof of Effort integer.

The Nonce & Proof of Effort

A "nonce" is a concept borrowed from blockchain architectures. In WritersProof, the nonce is the mathematical representation of "Proof of Effort," driven by Biological Friction.

If an AI generates 5,000 words in 3 seconds, the "effort" delta is near zero. For a human, those words represent hours of typing, pausing, and restructuring. The WritersProof engine calculates this biological friction and requires the machine to compute a nonce that corresponds to that effort — undeniable proof that human cognitive friction was expended.If an AI generates 5,000 words in 3 seconds, the "effort" delta is near zero. For a human, those 5,000 words represent hours of typing, backspacing, pausing, and restructuring. The WritersProof engine calculates this biological friction (keystroke velocity, edit distance, time elapsed) and requires the local machine to compute a mathematical nonce that corresponds to that effort. The nonce acts as undeniable, cryptographically verified proof that human time and cognitive friction were expended between State A and State B.

The Empirical Foundation

This is not a lab study. This is the field.

We tested this across 195,821 keystroke checkpoints from 13 independent datasets spanning every major typing context, from argumentative essays to scholarly manuscripts to population-scale transcription.We tested this across 195,821 keystroke checkpoints from 13 independent datasets spanning every major typing context — argumentative essays (KLiCKe, 4,992 writers), free-text composition (Mendeley, 644 users), scholarly LaTeX manuscripts (ScholaWrite, 8 authors), neurological patients (Tappy Parkinson's, 249 users), emotion-conditioned typing (EmoSurv, 81 participants), password authentication (CMU, 20,400 repetitions), and population-scale transcription (Aalto 136M, 168,513 volunteers).

1. Measuring the Signals

Five consciousness-correlate modules — thermodynamic entropy, integrated information (IIT/phi), temporal binding, free energy trajectory, and adaptive recovery — each grounded in neuroscience literature and independently validated. Against naive and statistical adversaries, these signals work: AUC 0.969-0.997. Detectable. Real.

2. The Attack Vectors

Then we attacked them. We built four adversary tiers with increasing knowledge of the system. The expert adversary — given full access to the signal extraction pipeline — reduced composite AUC to 0.31. Worse than random. Every signal fell.

This isn't a theoretical concern. We ran the attack. It worked.

3. The Bounds of Forgery

The gradient forger showed that the signal landscape is hard to optimize blindly (0% convergence), but an adversary with distributional knowledge finds solutions sequentially (20% convergence). Parallel forgery is hard, but sequential forgery is achievable.Then we asked why. The gradient forger told us: the signal landscape is hard to optimize blindly (0% convergence with gradient descent), but an adversary with distributional knowledge finds solutions sequentially (20% convergence, evolutionary strategy). Meaning: parallel forgery is hard, but sequential forgery — with enough information about the target — is achievable.

Using only publicly available keystroke datasets, we reproduced human writing statistics (KS=0.158, p=0.44). An adversary has access to the same data. The features are reachable without being observed.Then we confirmed the forgery is achievable. Using only publicly available keystroke datasets, we calibrated our simulation to match real human writing distributions. After calibration, the simulation's IKI distribution was statistically indistinguishable from Mendeley's 644 free-text composition writers (KS=0.158, p=0.44). WPM matched at the same significance level. Entropy effect sizes dropped from d=-12 to d=-5 across composition datasets. In other words: with nothing but published data, we reproduced human writing statistics. An adversary has access to the same data. The features are reachable without being observed.

The Causal Shield

But one thing held. Across all 13 datasets, across all adversary tiers, across gradient and evolutionary forgery — the hash-chained sequential attestation detected fabricated traces.

Not because the statistics were wrong. The statistics were right. But because sequential causal structure cannot be retroactively fabricated. A forger who matches the IKI distribution still cannot produce a valid h_prev → h_content chain without executing every event in order. That computation costs the same as actually writing the document.

The conclusion is empirical. Statistical process features are reachable by a calibrated adversary. Sequential process attestation is not. The difference is observational privilege — recording the causal chain as it happens, before the adversary knows what to forge.The conclusion is empirical, not theoretical. Statistical process features are reachable by a calibrated adversary. Sequential process attestation is not. The difference is observational privilege — recording the causal chain as it happens, under hardware attestation, before the adversary knows what they need to forge. That is what WritersLogic does. That is what 195,821 real human keystrokes, four adversary tiers, and a gradient forger told us cannot be faked.

The Industry Direction

The global standards landscape is rushing to solve the "Deepfake and AI" problem, but they are all missing the chronological execution layer. Here is the reality of the market and exactly where WritersLogic fits in:

1. NISO CRediT (The Vocabulary)

What they do: The National Information Standards Organization finalized the CRediT taxonomy (ANSI/NISO Z39.104-2022), which defines 14 distinct roles for contributors (e.g., "Writing – original draft").

The limitation: CRediT is just a dictionary; it provides a structured way to describe how research is produced. It relies entirely on the honor system.

2. C2PA & CAWG (The File Signers)

What they do: C2PA creates a standard for embedding provenance metadata into media files. CAWG extends this by allowing creators to bind verified identity (social profiles, government IDs) to the content they produce.

The limitation: C2PA and CAWG are overwhelmingly focused on media and signing the final asset. They prove who exported the PDF, but they cannot prove how the text inside the PDF was written.

3. W3C Verifiable Credentials (The Transport)

What they do: The W3C defines a standard for "Verifiable Credentials" containing a claim. A triangle of trust exists between Issuer, Holder, and Verifier for checking authenticity.

The WritersLogic Gap

The industry has a vocabulary for authorship (NISO), a way to sign final files (C2PA/CAWG), and a way to transport claims (W3C VCs). But no one is securing the actual chronological effort.

Existing standards (the claim)WritersLogic (the effort)

WritersLogic bridges this gap. We provide the necessary execution layer for NISO CRediT compliance. We capture the human "Proof of Effort" via the 5-way entanglement, map that effort directly to the 14 NISO CRediT roles, and output a W3C Verifiable Credential. We turn the "honor system" of authorship into a mathematical certainty.

Go-To-Market Strategy

We do not sell an "AI Detector." We sell Responsible AI Use Infrastructure. WritersLogic makes human effort visible and AI use transparent. Writers attest to their process; behavioral evidence corroborates it. The result is cryptographic proof that distinguishes AI-assisted from AI-authored work.

01 / Publishers & Scientific Journals

The Pain Point: Academic integrity is collapsing under retractions.

The Pitch: "Instead of guessing whether AI was involved, let authors attest to their process and back it with evidence. WritersLogic provides a fully C2PA and NISO CRediT-compliant Verifiable Credential that shows exactly how the work was created. You don't just get the manuscript; you get the cryptographic chain of custody."

02 / Universities & Education

The Pain Point: False-positive AI detection destroys trust.

The Pitch: "WritersLogic acts as a student's digital defense attorney. By writing in our environment, students generate an immutable Proof of Effort. If a professor accuses them of using AI, the student simply hands over their WritersProof hash chain, instantly proving their cognitive work."

03 / Authors & Creators

The Pain Point: Scraping bots and AI spam steal premium value.

The Pitch: "Secure your provenance. With WritersProof, your CAWG identity assertions are permanently entangled with your keystrokes. You can mathematically prove your copyright, outrank AI spam, and assert ownership over your intellectual property."

Business Model & Architecture

Frictionless Adoption: The Menubar App

By making WritersProof an editor-agnostic, background menubar application, we eliminate the highest hurdle in SaaS adoption: friction. We are no longer asking writers to abandon MS Word, Google Docs, or VS Code. We are simply running a silent cryptographic engine underneath their existing workflow. The business scales through a Freemium dual-revenue model: Premium Storage + Micro-transaction Verification.

Strategic Distinction

WritersLogic

The Corporate Entity, B2B Enterprise Hub, and API provider. This is the company building the verification infrastructure and managing institutional licensing.

WritersProof

The open standard Cryptographic Provenance Engine. The actual protocol, certificate authority, and underlying hash chain logic.

Get in Touch