Hard Wiki — A Cryptographic Knowledge Ledger for AI and Humanity

Version 1.0 • 2 August 2025

Hello@hardwiki.org | X @HardWikiOrg | https://www.hardwiki.org/

Table of Contents

1. Abstract

2. Introduction & Problem Statement

3. Solution Overview

4. System Architecture

   • 4.1 Base-First Layer Stack

   • 4.2 On-Chain Modules on Base

   • 4.3 Submission & Challenge Flow

   • 4.4 Illustrative User Story

5. Claim Types & Data Schema

6. Storage & Data-Integrity Model

7. Incentive, Token & Governance Layer

8. Premise Packs & Ontology

9. AI Grounding Interface

10. Economic Analysis & Cost Deterrents

11. Security, Compliance & Open-Source Policy

12. Competitive Landscape

13. Roadmap

14. Open Questions & Future Work

15. Appendix

1. Abstract

Hard Wiki is a public, cryptographically‑secured ledger of machine‑verifiable knowledge (theorems, computations, and experimental results). Deployed on Base, Hard Wiki couples proof‑of‑personhood, financial staking, and zero‑knowledge verification so that claims are immutable, challengeable, and easily retrievable by LLMs and researchers alike.

By aligning cryptographic proofs with economic incentives, Hard Wiki aims to become the canonical backbone for fact‑grounded AI, open science, and institutional due‑diligence.

2. Introduction & Problem Statement

Large‑language models (LLMs) excel at generating fluent text but often hallucinate facts, citations, and even entire proofs. In high‑stakes domains (e.g. healthcare, law, finance, aviation) this unreliability demands costly manual verification or carries unacceptable risk. Simultaneously, AI‑generated content is set to saturate the internet, collapsing the signal‑to‑noise ratio and burying vetted human knowledge under probabilistic output.

Key Challenges

1. Hallucinations & false citations
   LLM outputs must be re‑fact‑checked line‑by‑line, imposing heavy cognitive and financial overhead.

2. Sybil‑vulnerable publication channels
   Existing wikis and journals rely on social trust; adversaries can cheaply create sock‑puppet identities and push misinformation.

3. Unstructured knowledge
   Most scientific artifacts are posted as PDFs or blogs, lacking machine‑readable metadata, proofs, or reproducibility hooks.

4. Economic misalignment
   Today’s platforms rarely punish inaccurate claims; the cost of error is externalized onto downstream users.

Hard Wiki proposes a deterministic, incentivized, on‑chain backbone where every fact carries (i) cryptographic provenance, (ii) human uniqueness proofs, and (iii) meaningful financial skin‑in‑the‑game.

3. Solution Overview

Hard Wiki rests on four mutually reinforcing pillars:

1. Typed, machine‑verifiable claims
   Submissions are labelled Theorem, Computation, or Experiment and, when feasible, accompanied by a zero‑knowledge (ZK‑SNARK) proof or container hash attesting to deterministic correctness.

2. Proof‑of‑personhood gating
   Authors verify unique‑human status via World ID, mitigating Sybil attacks and enabling per‑human rate‑limits.

3. Financial staking & slashing
   A 0.05 ETH (or HARDW‑equivalent) bond must accompany every claim; successful challenges trigger automatic or DAO‑mediated slashing, rewarding auditors and funding the public treasury.

4. Content‑addressed, redundant storage
   Concise proofs are kept on‑chain, while bulky artifacts live on Arweave, Filecoin, or IPFS and are referenced by immutable CIDs.

Together these pillars yield a trustworthy, auditable knowledge graph that any AI system or researcher can query without re‑deriving every intermediate result.

4. System Architecture

4.1 Base‑First Layer Stack

1. Base L2 (OP-Stack)

   • Role: Primary execution: claim submission, World ID checks, SNARK verification, staking, dispute logic.

   • Rationale: Low, predictable fees; Coinbase fiat on-ramp; broad user base.

2. Ethereum L1

   • Role: Finality for Base state commitments; emergency escape hatch for submissions.

   • Rationale: Immutable history; highest security guarantees.

3. Arweave / Filecoin / IPFS

   • Role: Storage for artifacts (datasets, scripts, PDFs).

   • Rationale: Cheap, content-addressed permanence; multi-network redundancy mitigates data rot.

4.2 On‑Chain Modules on Base

• ClaimRegistry: submitClaim, challengeClaim, resolveChallenge.

• StakeVault: Holds publisher bonds; auto‑distributes slashed funds.

• PremisePackRegistry: ERC‑721 packs of axioms, constants, or domain ontologies.

• GovTimelock: 7‑day timelock for parameter upgrades executed via HARDW‑token DAO.

• WorldIDRouter: Verifies World ID roots bridged from Ethereum mainnet.

All contracts conform to ERC‑7641 upgradeable‑proxy guidelines and will undergo formal verification before mainnet.

4.3 Submission & Challenge Flow

1. Bundle & Upload
   Author bundles artifacts (CLI or web UI). Large files are compressed and pinned to Arweave → returns CID.

2. Optional ZK Proof
   For deterministic claims (Theorem / Computation), a Groth16 or Plonk proof is generated locally.

3. Proof‑of‑Personhood
   User scans World ID QR, obtaining {root, nullifierHash, proof}.

4. submitClaim(tx)
   Transaction includes metadata, proof (or analysis_hash), premise packs, 0.05 ETH bond, and World ID fields.

5. Verification & Indexing
   • Automatic path for Theorem/Computation: on‑chain verifier executes; success emits ClaimSubmitted.
   • Flag path for Experiment (Phase 2): claim enters a 30‑day observation window where replication challenges can be filed.

6. Challenge Window
   Anyone may stake 0.01 ETH to challengeClaim. If the counter‑proof succeeds or DAO vote passes, the original bond is slashed and distributed (50 % challenger, 25 % treasury, 25 % Replication Fund).

7. Commitment & Finality
   Every ~30 min a Base state root is posted to Ethereum L1 via the canonical bridge, giving ~30–45 min economic finality.

4.4 Illustrative User Story

Dr Ada publishes a Lean proof of a new combinatorial lemma. Bundle size: 80 KB. She locks a 0.05 ETH bond (~$175) and pays ~$4 gas on Base. The claim imports math‑foundations‑zfc‑v1 and passes the on‑chain verifier within 12 s.

Dr Bob spots an edge‑case error. He stakes 0.01 ETH to challenge, submits a counter‑example script, and the verifier detects a contradiction. Bob receives 0.025 ETH, 0.0125 ETH goes to the treasury, and 0.0125 ETH tops up the Replication Fund. Ada’s claim is marked Rejected and archived.

Ms Chloe later submits an improved proof under math‑foundations‑zfc‑v2; authors depending on the old pack receive automated migration alerts via webhook and email.

5. Claim Types & Data Schema

Claim:

  id: bytes32                # = keccak256(bundle CID | publisher)

  type: enum {Theorem, Computation, Experiment}

  cid: bytes32               # Arweave root of artifact bundle

  snark_proof?: bytes        # present for Theorem/Computation

  analysis_hash?: bytes32    # OCI digest for containerised runs

  premisePacks: bytes32[]    # list of imported ERC-721 packs

  inlinePremRoot: bytes32    # Merkle root for ad-hoc premises

  dataset_root?: bytes32     # Merkle root for raw data (Experiments)

  reviewers: bytes32[]       # optional attestations (EAS / VC)

  stake: uint256             # bond locked in StakeVault

  publisher: address         # EOA or smart wallet

  timestamp: uint64          # block.timestamp at submission

Field Highlights
snark_proof — Groth16/Plonk proof verifying correctness without revealing proprietary code.
analysis_hash — Reproducible Docker/Nix image hash ensuring environmental determinism.
reviewers — Off‑chain attestations anchored via Ethereum Attestation Service (EAS) or Verifiable Credentials.

6. Storage & Data‑Integrity Model

1. Proof & metadata (≈256 B)

   • Location: Base L2 storage

   • Permanence: Immutable post-finality

   • Rot Mitigation: Checkpoints to Ethereum L1 every 30 min.
     
     

2. Artifact bundles

   • Location: Arweave / Filecoin / IPFS

   • Permanence: Pay-once permanence (Arweave) + mutable pinning (IPFS)

   • Rot Mitigation: DAO-funded pinning bots; quarterly CID health-check bounties.
     
     

3. L2 state batches

   • Location: Base EIP-4844 blobs

   • Permanence: 18-day rolling availability

   • Rot Mitigation: Snapshot mirrors into EthStorage for cold-storage recovery.

7. Incentive, Token & Governance Layer

HARDW Token (100 M fixed supply): Powers governance, fee discounts, bounty funding, and staking alternatives.

• Publishing Bond: 0.05 ETH or HARDW equivalent locked in StakeVault per claim.

• Protocol Fee: 1 % on all bonds and replication bounties; funds foundation grants and audits.

• Slashing Distribution: 50 % → successful challenger, 25 % → protocol treasury, 25 % → Replication Fund.

• Rate Limit: Max one claim per World ID nullifier per 24 h epoch.

• Governance: HARDW‑token DAO with a 7‑day timelock on key parameters (bond size, fee %, pack limits).

8. Premise Packs & Ontology

Hard Wiki uses a hybrid model for shared definitions:

• Premise Packs (ERC‑721) containing versioned JSON‑LD bundles of axioms, constants, and units.

• Imports: Packs specify an imports[] DAG, enabling reuse without duplication.

• Inline Premises: Claims can include an inlinePremRoot with a Merkle root of ad‑hoc premises.

Governance & Limits

• Creation Bond: 0.02 ETH + 0.00001 ETH × pack size (bytes).

• Byte Ceiling: 16 KB per pack.

• Metadata Required: title, abstract, imports[], cid.

• Soft Deprecation: Packs can be flagged or burned by a 2⁄3 majority in DAOs holding ≥ 0.1 % of circulating HARDW.

Audit Tools

• Pack Explorer: Visual diff and dependency graph.

• LLM Summarizer: Automated 200‑word English digest.

• Flag Dashboard: Shows packs under dispute and bounty sizes.

• Alert Subscriptions: Notifications for changes or deprecations.

9. AI Grounding Interface

• GraphQL / REST API: getClaim(claimID) → {cid, proof, metadata}.

• Fact Token URI: hard://<claimID>#<hash> embeddable in AI outputs.

• Experimental zk‑Retrieval: Proofs that LLM callers only accessed whitelisted claims.

10. Economic Analysis & Cost Deterrents

1. 80 KB calldata

   • Cost on Base: ≈ $0.005

   • Notes: Data fee after EIP-4844.

2. Publishing Bond

   • Cost on Base: $175 (0.05 ETH)

   • Notes: Ensures only confident claims are published.

3. Total Upfront

   • Cost on Base: ≈ $180

   • Notes: Combined cost of calldata and publishing bond.

Protocol fees alone (1 % on 200 k annual claims): ~$350 k / yr — covers audits and grants but necessitates additional commercial revenue streams for team salaries.
While protocol fees already cover ongoing audits and community grants, additional revenue from optional commercial services ensures we can scale R &D, 24 / 7 support, and regulatory compliance without taxing the DAO

11. Security, Compliance & Open‑Source Policy

11.1 Security & Audits

• Sybil Resistance: World ID nullifier + rate limits.

• Formal Audits: Trail of Bits review of core contracts.

• Fuzzing & Static Analysis: CLI and smart contracts tested pre‑release.

11.2 Compliance

• GDPR: Personal data prohibited; only hashed or consented information allowed.

• Entity Structure: > _

  • Hard Wiki Verein (CH) — non-profit steward of open-source contracts, tokens, and governance keys.

• Hard Wiki Inc. (US Delaware C-corp, Public-Benefit) — optional commercial arm that offers dashboards, SLAs, and compliance tooling. All core protocol IP is dual-licensed to the Verein under a perpetual, no-fee licence; upgrade authority rests with DAO-controlled multisig.

11.3 Open‑Source Policy

1. Protocol contracts & ZK code

   • License: Apache 2.0

   • Repository: github.com/HardWikiOrg/protocol

   • Maintainer: Hard Wiki Foundation

2. CLI & Reference frontend

   • License: MIT

   • Repository: github.com/HardWikiOrg/frontend

   • Maintainer: Community + Foundation

3. Enterprise SaaS & dashboards

   • License: Proprietary

   • Repository: Private repos under Hard Wiki Inc.

   • Maintainer: Hard Wiki Inc.

   • Notes: Public repositories welcome PRs under a CLA. Proprietary code funds team salaries and commercial support.

12. Competitive Landscape

1. Everipedia

   • Focus: General knowledge wiki

   • Hard Wiki Differentiator: No machine-proof, no staking, unstructured content

2. Kleros Curate

   • Focus: Token-curated lists

   • Hard Wiki Differentiator: Manual list curation, no ZK or human proof gating

3. OriginTrail

   • Focus: Supply-chain knowledge

   • Hard Wiki Differentiator: No on-chain staking, no proof-of-personhood gating

4. OpenAlex

   • Focus: Scholarly metadata graph

   • Hard Wiki Differentiator: Not cryptographically anchored; no economic incentives

13. Roadmap

1. Phase 0

   • Milestone: Base Testnet: CLI & Web α, automatic Challenge ⚙︎

   • Objective: Validate end-to-end flow with fiat on-ramp.
     
     

2. Phase 1

   • Milestone: Base Mainnet: HARDW token launch

   • Objective: Enterprise launch + public auditability.

3. Phase 2

   • Milestone: Replication DAO & Flag ⚑ path for Experiment claims

   • Objective: Enable subjective dispute resolution.

4. Phase 3

   • Milestone: Multi-Chain Mirror: optional bridge to other roll-ups (community approved)

   • Objective: Extend presence across ecosystems.

5. Phase 4

   • Milestone: Commercial Services

   • Objective: Achieve sustainable commercial revenue.
     
     

14. Open Questions & Future Work

• Private data support: Investigate FHE and access‑controlled IPFS for sensitive datasets.

• zk‑Retrieval proofs: Efficient proofs that AI retrievers only accessed approved claims.

• Cross‑chain interoperability: Bridge claims and verifiers across StarkNet, Scroll, and others.

• Sequencer decentralization: Monitor Based Sequencer rollout and evaluate multi‑sequencer security.

15. Appendix

Glossary

SNARK — Succinct Non‑interactive Argument of Knowledge.
CID — Content Identifier for IPFS/Arweave.
Premise Pack — Versioned bundle of definitions and axioms.
EIP‑4844 Blob — Low‑cost data‑availability chunk on Ethereum L1.
World ID Nullifier — Unique per‑person hash binding proof‑of‑personhood.
OCI Digest — Immutable hash of a container image.
SecureCore — Proton VPN’s multi‑hop network.

References

1. Vitalik Buterin et al., “EIP‑4844: Proto‑Danksharding”, 2023.

2. Worldcoin, World ID Protocol Docs, 2025.

3. Trail of Bits, Smart Contract Audit Report, 2025.

Hard Wiki v1.0 delivers a Base‑centric, enterprise‑ready deployment backed by open‑source protocol rails and a sustainable commercial ecosystem—anchoring humanity’s most rigorous knowledge in a ledger future AIs will trust.

Hello@hardwiki.org | X @HardWikiOrg | https://www.hardwiki.org/