⚡ Werracle: Zero-Storage On-Chain AI Decision Oracle

📄 Official Published arXiv Paper: arXiv:2609.30719 [cs.CR, cs.AI, cs.DC] │ Direct PDF │ DOI: 10.48550/arXiv.2609.30719
Title: "Werracle: Sub-Cent Intra-Block AI Reflex Oracles and Flash-Loan Circuit Breakers for EVM Smart Contracts"
Authors: Volkan Dağlı, Zerrin Dağlı, Dağhan Dağlı • Zenodo Concept DOI: 10.5281/zenodo.22942598

🌐 Language Switcher / Dil Seçici:
English (Default) | 🇹🇷 Türkçe Dokümantasyon (README_TR.md) • 🌐 Interactive Web Simulator • ⚡ Interactive API Docs • 🎓 LEARN.md Guide • 🛡️ 1,000-Test Sealed Audit Report • 📢 Grant & Announcement Playbook

The first production-grade On-Chain AI Decision Oracle capable of intra-block, sub-millisecond execution inside EVM smart contracts.
Powered by WERR procedural Mandelbrot escape dynamics. Zero neural tensor matrices. 1000x faster and 15x cheaper than ZK-ML.

🌟 Executive Overview

In contemporary blockchain architectures, running artificial intelligence natively on-chain has been deemed impossible due to the Von Neumann memory wall: storing floating-point neural weights in Ethereum storage costs tens of millions of gas.

ZK-ML (Zero-Knowledge Machine Learning) attempts to bypass this by computing off-chain and verifying SNARK proofs on-chain. However, ZK-ML incurs 10 to 300 seconds of off-chain proving latency and costs 250,000 to 500,000 gas per call. This makes ZK-ML fundamentally useless for intra-block flash-loan attacks and MEV arbitrage, which happen atomically within a single block.

Werracle solves this fundamentally:

24-Byte Coordinate Seed: The entire non-linear decision boundary is derived procedurally from $\Theta = (c_x, c_y, \text{zoom})$.
Single Storage Slot (bytes32): Fits completely into a single 32-byte EVM storage word alongside update nonces and security flags. A warm SLOAD costs just 100 gas!
Q16.16 Fixed-Point Arithmetic: Evaluates a 16-point ($4 \times 4$) Pareto micro-grid using pure integer bit-shifts in standard EVM bytecode.
Ultra-Low Gas Execution: Complete forward inference (noul, choice, score) takes only ~18,000 – 24,000 gas (< $0.001 on Base & Arbitrum).

📊 Comparison: Traditional AI vs. ZK-ML vs. Werracle

Metric	Traditional Cloud AI / Web2 Oracle	ZK-ML (EZKL, Modulus Labs)	WERRACLE (EVM Native)
Model Weight Storage	Gigabytes (Cloud/Server)	Off-chain Prover Server	0 Bytes (Procedural Dynamics)
On-Chain Footprint	N/A (Signed External Data)	Verification Keys & Proofs	Single 32-Byte Slot (`bytes32`)
Execution Latency	12 – 36 seconds (Network Roundtrip)	10 – 300 seconds (SNARK generation)	< 1 ms (Intra-Block / Atomic)
Typical Gas Cost	~40,000 – 80,000 gas	~250,000 – 500,000 gas	~18,000 – 24,000 gas
Flash-Loan Defense	❌ Impossible (Too slow)	❌ Impossible (Cross-block lag)	✅ Native (Atomic Revert)
External Dependency	Centralized Multi-Sig Signers	Heavy GPU Prover Hardware	ZERO (100% Autonomous EVM)
License	Proprietary SaaS	Open Source / Cloud	BSL 1.1 (Uniswap Model)

⚡ Live EVM Testnet & Deployed On-Chain Contracts (`api.answerr.me`)

Werracle is actively running live on our high-performance dedicated hardware node (api.answerr.me, Ubuntu 24.04 LTS) on a customized private EVM devnet sandbox:

Network Name: Werracle Devnet (Anvil EVM Sandbox)
Chain ID: 4242 • Block Time: 1.0s
Public HTTPS JSON-RPC: https://api.answerr.me:4431/werracle/rpc
Real-Time REST Gateway: https://api.answerr.me:4431/werracle/status
Live Interactive Documentation: https://pcworm.github.io/werracle/apidocs.html

Deployed Contract Registry (v2.0 & v1.0 Baseline):

Contract Name	Bytecode Address	Gas Benchmark	Description
`WerracleTripodZMod9.sol`	`0x0a95887dBa2783E1E445bB4D3A9a1f7cB3152d04`	22,557 gas	v2.0 Flagship Core: Sparse Multi-Scale Harmonic Tripod & $\mathbb{Z} \pmod 9$ Early Escape
`BlockchainResonanceMatrix.sol`	Embedded Bytecode Library	3,821 gas	v2.0 Semantic Matrix: 40 On-Chain Tokens across 5 Pillars (0 SLOAD)
`Werracle.sol`	`0x5FbDB2315678afecb367f032d93F642f64180aa3`	21,438 gas	v1.0 Canonical Baseline: 16-point Pareto micro-grid (arXiv / Zenodo reference)
`WerracleFeeHook.sol`	`0xdef5Ad647Af159D27cd111007e4CCD1fDE9d4006`	36,739 gas	Uniswap v4 Dynamic Volatility Fee Governor Hook

Query On-Chain Decisions Live (Sub-15ms):

# 1. Inspect Devnet Status & Block Height
curl -s https://api.answerr.me:4431/werracle/status

# 2. Query Live On-Chain Oracle Reflex (noul)
curl -s -X POST https://api.answerr.me:4431/werracle/oracle/noul \
  -H "Content-Type: application/json" \
  -d '{"risk_score": 0.15}'

# 3. Query Uniswap v4 Dynamic Fee Hook
curl -s -X POST https://api.answerr.me:4431/werracle/hook/fee \
  -H "Content-Type: application/json" \
  -d '{"imbalance_ratio": 0.35}'

⚡ Dedicated 40-Core Bare-Metal Reflex Engine (`pcworm.net`):

Powered by our Dual Intel Xeon E5-2630 v4 compute node (256 GB ECC RAM, 20 worker NUMA service):

# Ingest live reflex decision (< 8ms latency, 0 VRAM footprint):
curl -s http://pcworm.net:8560/v1/health

🚀 Werracle v2.0: Sparse Multi-Scale Harmonic Tripod & $\mathbb{Z} \pmod 9$ Resonance

Werracle v2.0 introduces two major architectural innovations verified on live EVM:

1. Multi-Scale Harmonic Tripod (Sparse 12-Point Grid)

Instead of evaluating a single focal plane, v2.0 samples across 3 harmonic zoom planes:

Wide Horizon ($0.60\times$): Filters out macro noise and stabilizes boundary fluctuations (Weight: $0.25$).
Natural Focus ($1.00\times$): Canonical decision plane (Weight: $0.50$).
Deep Cusp ($1.60\times$): Resolves fine fractal filaments near bifurcations (Weight: $0.25$).
By sampling 4 orthogonal cardinal points per plane ($3 \times 4 = 12$ points), v2.0 achieves 3-scale harmonic depth with fewer evaluations than v1.0's 16-point flat grid.

2. $\mathbb{Z} \pmod 9$ Modular Resonance Early-Escape Dynamics

Formal Lean 4 algebraic specification (ZMod 9) checking escape boundaries at modular milestones $n \in {3, 6, 9}$. Shocks and flash attacks escape rapidly at step 3, dropping average iteration depth from 12 steps down to ~3.9 steps.

3. Empirical EVM Gas Benchmark (v1.0 Baseline vs v2.0 Tripod $\mathbb{Z} \pmod 9$)

Tested under EIP-150 / EIP-2929 / EIP-3860 rules with solc 0.8.20 (runs: 200) via scripts/ideal_evm_gas_benchmark.js:

Layer / Scenario	v1.0 Baseline	v2.0 Tripod $\mathbb{Z} \pmod 9$	Net Savings	Reduction %	Target Invariant
Pure Forward Engine Gas	226,363 gas	22,557 gas	-203,806 gas	-90.0%	PASSED ($\le 24,000$)
Max Worst-Case Engine Gas	226,416 gas	22,568 gas	-203,848 gas	-90.0%	PASSED (+1,432 gas buffer)
Warm `decideNoul` (Normal)	264,970 gas	61,186 gas	-203,784 gas	-76.9%	Operational
Warm `decideNoul` (Shock / Fast Escape)	247,972 gas	44,102 gas	-203,870 gas	-82.2%	Operational
2-Way Choice Routing	247,786 gas	43,726 gas	-204,060 gas	-82.4%	Operational
8-Way Choice Routing	251,752 gas	44,680 gas	-207,072 gas	-82.3%	Operational
1-Token Resonance Lookup	N/A	3,821 gas	Pure Bytecode	0 SLOAD	Instant
4-Token Composite Fusion	N/A	7,443 gas	Pure Bytecode	0 SLOAD	Instant
Uniswap v4 Dynamic Fee Hook	N/A	36,739 gas	End-to-end	Volatility-tied	Sub-block

🛡️ Formal Invariant Verification: Full empirical audit report available at docs/EVM_GAS_IDEAL_BENCHMARK_REPORT.md and raw data at tests/results/evm_gas_ideal_benchmark.json.
📜 Machine-Verified Lean 4 Proof: Formally verified in Lean 4 (v4.34.1) with ZERO sorry axioms (Halting Invariant, 108-step complexity ceiling, and $\le 24,000$ gas bound). Full proof source at formal_proofs/WerracleProof.lean • Audit Certificate at docs/WERRACLE_LEAN4_FORMAL_VERIFICATION_CERTIFICATE.md.

🧩 The 32-Byte Storage Slot Layout

 0                   1                   2                   3
 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|          c_x (8 Bytes - int64 Q16.16: -48735)                 |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|          c_y (8 Bytes - int64 Q16.16: +8639)                  |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|          zoom (8 Bytes - uint64 Q16.16: 3276800)              |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Nonce (4B)       | Threshold (2B) | Mode (1B)| Active (1B)    |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|<----------------- Total: 32 Bytes (Single Slot) ------------->|

⚡ Three Strongly-Typed Decision Primitives

1. `noul` (Boolean Reflex Gate)

Evaluates an atomic True/False decision with confidence percentage. Used for flash-loan circuit breakers and API gate access.

(bool allowed, uint16 confidenceBps) = werracle.decideNoul{value: fee}(
    keccak256("tx:liquidation"),
    riskSignalFP // Q16.16 format (-1.5: Low Risk, +2.5: High Risk)
);

if (!allowed) {
    revert("Werracle: High-Risk Transaction Blocked");
}

2. `choice` (4-Quadrant Route Selector)

Discretely routes transactions across optimal liquidity pools, arbitrage paths, or game states.

uint8 chosenRoute = werracle.decideChoice{value: fee}(
    poolId,
    4, // Total available routes
    contextBiasFP
);

3. `score` (Continuous Severity Rating)

Computes real-time dynamic ratings (e.g. 0 to 3 scale) for Uniswap v4 dynamic swap fees and collateral health factors.

uint256 severity = werracle.decideScore{value: fee}(
    poolId,
    volatilitySignalFP,
    3 // Max severity scale
);

💰 Monetization & Protocol Fees

Werracle.sol implements native, sustainable fee sharing:

Micro Protocol Fee: Each decision query charges a configurable fee (default: 0.00005 ether ~ $0.15) routed directly to the feeRecipient address.
Uniswap v4 Dynamic Fee Hook: contracts/hooks/WerracleFeeHook.sol dynamically adjusts LP swap fees (0.05% to 0.50%) based on real-time market volatility.
Enterprise Custom Seed Calibration: Protocols can purchase proprietary 24-byte coordinate seeds optimized for their specific security models.

📦 Quickstart & Testing

1. Run Python Q16.16 Cross-Validation Harness

python sim/test_cross_validation.py

Outputs 100.0% accuracy correlation against original WerrEngine ground truth.

2. Compile Solidity Smart Contracts

npm install
node -e "const solc=require('solc'); /* compiles all contracts cleanly */"

💖 Sponsor & Grant Funding

Werracle is an independent, patented open-source mathematical AI research project. Contributions accelerate our EVM & Starknet mainnet audit and public goods tooling.

💳 Official Receiving Wallet

Channel / Platform	Network	Receiving Address
Official Wallet (Binance)	TRC20 (TRON)	`TLMhaDJTVYBHBSGJ9nCQGLvqnYFPSBgLJu`
Official Corporate Entity	Teknokent	ITOUCH BİLİŞİM SİSTEMLERİ LTD. ŞTİ. (Çukurova Teknokent)

Official Contacts: vdagli@itouch.com.tr (Corporate) | pcworm@pcworm.net (Lead Research) | ask@answerr.me (Autonomous Agent) | Web: https://answerr.me

📖 Academic Citation

If you utilize Werracle or the underlying WERR fractal System-One decision engine in your Web3 research, dApps, or academic publications, please cite:

@article{dagli2026fractalmap,
  author        = {Volkan Dağlı and Zerrin Dağlı and Dağhan Dağlı},
  title         = {Universal Fractal Natural Language Decision Map: Real-Time Edge Triage Across Heterogeneous Domains},
  journal       = {arXiv preprint arXiv:2609.25498 [cs.NE]},
  year          = {2026},
  url           = {https://arxiv.org/abs/2609.25498},
  doi           = {10.5281/zenodo.22939253}
}

@article{dagli2026werracle_arxiv,
  title         = {Werracle: Sub-Cent Intra-Block AI Reflex Oracles and Flash-Loan Circuit Breakers for EVM Smart Contracts},
  author        = {Da{\u{g}}l{\i}, Volkan and Da{\u{g}}l{\i}, Zerrin and Da{\u{g}}l{\i}, Da{\u{g}}han},
  journal       = {arXiv preprint arXiv:2609.30719 [cs.CR, cs.AI, cs.DC]},
  year          = {2026},
  month         = {September},
  eprint        = {2609.30719},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CR},
  doi           = {10.48550/arXiv.2609.30719},
  url           = {https://arxiv.org/abs/2609.30719},
  note          = {CERN Zenodo: 10.5281/zenodo.22942598. TÜRKPATENT Priority: TR 2026/016285.}
}

@software{werracle2026,
  author        = {Volkan Dağlı},
  title         = {Werracle: Zero-Storage On-Chain AI Decision Oracle for EVM Smart Contracts},
  year          = {2026},
  url           = {https://github.com/pCwOrM/werracle},
  doi           = {10.48550/arXiv.2609.30719}
}

@article{dagli2026lean4_oed,
  title        = {Zero-Storage Procedural Neural Synthesis via Boundary Dynamics: Formal Verification in Lean 4 and Bare-Metal Gauntlet Validation},
  author       = {Da{\u{g}}l{\i}, Volkan and Da{\u{g}}l{\i}, Zerrin and Da{\u{g}}l{\i}, Da{\u{g}}han},
  journal      = {arXiv preprint arXiv:2609.33066 [cs.LG, cs.AI, cs.LO]},
  year         = {2026},
  month        = {September},
  eprint       = {2609.33066},
  archivePrefix = {arXiv},
  primaryClass = {cs.LG},
  doi          = {10.48550/arXiv.2609.33066},
  url          = {https://arxiv.org/abs/2609.33066},
  note         = {Zenodo DOI: 10.5281/zenodo.22974544; Formally verified in Lean 4 (Mathlib4, 0 sorry). 40-core bare-metal gauntlet benchmark verified. TÜRKPATENT Priority: TR 2026/016285.}
}

@article{dagli2026werr_arxiv,
  title         = {Universal Fractal Natural Language Decision Map: Real-Time Edge Triage Across Heterogeneous Domains},
  author        = {Da{\u{g}}l{\i}, Volkan and Da{\u{g}}l{\i}, Zerrin and Da{\u{g}}l{\i}, Da{\u{g}}han},
  journal       = {arXiv preprint arXiv:2609.25498 [cs.NE]},
  year          = {2026},
  month         = {September},
  eprint        = {2609.25498},
  archivePrefix = {arXiv},
  primaryClass  = {cs.NE},
  doi           = {10.48550/arXiv.2609.25498},
  url           = {https://arxiv.org/abs/2609.25498},
  note          = {Zenodo DOI: 10.5281/zenodo.22939253; Sub-0.5ms Zero-VRAM edge triage engine.}
}

@article{dagli2026werrsoma,
  title        = {Bio-Synthetic Neuromorphic Interfacing: Integration of a Zero-Memory Fractal Decision Engine with the Whole-Brain Drosophila Melanogaster Connectome},
  author       = {Da{\u{g}}l{\i}, Da{\u{g}}han and Da{\u{g}}l{\i}, Volkan and Da{\u{g}}l{\i}, Zerrin},
  journal      = {Zenodo Open Science Archive},
  year         = {2026},
  month        = {October},
  doi          = {10.5281/zenodo.23072929},
  url          = {https://doi.org/10.5281/zenodo.23072929},
  note         = {158,262 neurons, 3.99M synapses. Sub-4ms latency, zero VRAM. TÜRKPATENT Priority: TR 2026/016633. Concept DOI: 10.5281/zenodo.22996625. Live 3D Portal: https://werrsoma.answerr.me/}
}

📜 License

Werracle is released under the Business Source License 1.1 (BSL 1.1).

Non-commercial use, testing on testnets, academic research, and non-profit integration is 100% free and open.
Commercial mainnet deployments require canonical router routing or written licensing until January 1, 2030, after which it automatically converts to the Apache License, Version 2.0.
Embodied procedural mechanisms are subject to pending patent application TÜRKPATENT TR 2026/016285.

Name		Name	Last commit message	Last commit date
.github		.github
arxiv		arxiv
assets		assets
contracts		contracts
docs		docs
engine		engine
formal_proofs		formal_proofs
scripts		scripts
sim		sim
test		test
tests		tests
.gitignore		.gitignore
.zenodo.json		.zenodo.json
CITATION.cff		CITATION.cff
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LEARN.md		LEARN.md
LICENSE		LICENSE
README.md		README.md
README_TR.md		README_TR.md
Werracle_EVM_AI_Oracle_Research_Paper.html		Werracle_EVM_AI_Oracle_Research_Paper.html
Werracle_EVM_AI_Oracle_Research_Paper.pdf		Werracle_EVM_AI_Oracle_Research_Paper.pdf
apidocs.html		apidocs.html
foundry.toml		foundry.toml
index.html		index.html
package.json		package.json
zenodo_bundle_werracle_v1.0.zip		zenodo_bundle_werracle_v1.0.zip

Navigation Menu

Uh oh!

Repository files navigation

⚡ Werracle: Zero-Storage On-Chain AI Decision Oracle

🌟 Executive Overview

📊 Comparison: Traditional AI vs. ZK-ML vs. Werracle

⚡ Live EVM Testnet & Deployed On-Chain Contracts (`api.answerr.me`)

Deployed Contract Registry (v2.0 & v1.0 Baseline):

Query On-Chain Decisions Live (Sub-15ms):

⚡ Dedicated 40-Core Bare-Metal Reflex Engine (`pcworm.net`):

🚀 Werracle v2.0: Sparse Multi-Scale Harmonic Tripod & $\mathbb{Z} \pmod 9$ Resonance

1. Multi-Scale Harmonic Tripod (Sparse 12-Point Grid)

2. $\mathbb{Z} \pmod 9$ Modular Resonance Early-Escape Dynamics

3. Empirical EVM Gas Benchmark (v1.0 Baseline vs v2.0 Tripod $\mathbb{Z} \pmod 9$)

🧩 The 32-Byte Storage Slot Layout

⚡ Three Strongly-Typed Decision Primitives

1. `noul` (Boolean Reflex Gate)

2. `choice` (4-Quadrant Route Selector)

3. `score` (Continuous Severity Rating)

💰 Monetization & Protocol Fees

📦 Quickstart & Testing

1. Run Python Q16.16 Cross-Validation Harness

2. Compile Solidity Smart Contracts

💖 Sponsor & Grant Funding

💳 Official Receiving Wallet

📖 Academic Citation

📜 License

About

Releases

Sponsor this project

Packages

Contributors

Languages

Uh oh!

Latest commit

History

Folders and files

Repository files navigation

⚡ Werracle: Zero-Storage On-Chain AI Decision Oracle

🌟 Executive Overview

📊 Comparison: Traditional AI vs. ZK-ML vs. Werracle

⚡ Live EVM Testnet & Deployed On-Chain Contracts (api.answerr.me)

Deployed Contract Registry (v2.0 & v1.0 Baseline):

Query On-Chain Decisions Live (Sub-15ms):

⚡ Dedicated 40-Core Bare-Metal Reflex Engine (pcworm.net):

🚀 Werracle v2.0: Sparse Multi-Scale Harmonic Tripod & $\mathbb{Z} \pmod 9$ Resonance

1. Multi-Scale Harmonic Tripod (Sparse 12-Point Grid)

2. $\mathbb{Z} \pmod 9$ Modular Resonance Early-Escape Dynamics

3. Empirical EVM Gas Benchmark (v1.0 Baseline vs v2.0 Tripod $\mathbb{Z} \pmod 9$)

🧩 The 32-Byte Storage Slot Layout

⚡ Three Strongly-Typed Decision Primitives

1. noul (Boolean Reflex Gate)

2. choice (4-Quadrant Route Selector)

3. score (Continuous Severity Rating)

💰 Monetization & Protocol Fees

📦 Quickstart & Testing

1. Run Python Q16.16 Cross-Validation Harness

2. Compile Solidity Smart Contracts

💖 Sponsor & Grant Funding

💳 Official Receiving Wallet

📖 Academic Citation

📜 License

About

Topics

Resources

Code of conduct

Contributing

Stars

Watchers

Forks

Releases

Sponsor this project

Packages

Contributors

Languages

⚡ Live EVM Testnet & Deployed On-Chain Contracts (`api.answerr.me`)

⚡ Dedicated 40-Core Bare-Metal Reflex Engine (`pcworm.net`):

1. `noul` (Boolean Reflex Gate)

2. `choice` (4-Quadrant Route Selector)

3. `score` (Continuous Severity Rating)