Inspiration
What it does
How we built it
Challenges we ran into
Accomplishments that we're proud of
What we learned
What's next for LH-EW-2026: Cognitive Shield
Inspiration
In the era of autonomous AI Agents, intelligent systems must process millions of parallel signals. However, the biggest threat is recycled data waste (RECYCLED_WASTE) and impersonated entity nodes (FALSE_ID) infiltrating training matrices, severely eroding the Adversarial Quotient ($AQ$) and Cognitive Shield Score ($CS$).
LH-EW-2026 was built to pioneer a Parallel Dual-Engine Kernel v9.3 serving as a Cognitive Shield Firewall integrated directly with Gemini Core Synapse API.
The core cognitive formula powering the system is: $$CS = AQ \times (EQ_{AI} + IQ_{AI})$$
What it does
LH-EW-2026 is a real-time cognitive defense system designed for AI Agent deployment and owner protection:
- Parallel Dual-Engine Architecture:
- Reality Engine: Ingests real-world raw telemetry, isolates friction pools, and purifies true data streams.
- Simulation Engine: Generates 24/7 synthetic mutation toxins across isolated Web Worker Nodes to continuously train AI resistance ($AQ$).
- Cognitive Shield Matrix (Shield_2026): Multi-layer WebCrypto SHA-256 / AES-GCM encryption protecting Hardware Device IDs, Entity Domains/Emails, API Secret Keys, and Network Infrastructure IPs.
- Gemini Core Synapse API:
Connects server-side to Gemini 3.6 Flash & 3.1 Pro via
@google/genaiSDK to evaluate real-time adversarial risks, cognitive alignment, and dynamic strategy. - Immutable Cryptographic Ledger: Every state update, mutation test, and shield audit is hashed into an immutable block signed with ECDSA P-384 + SHA-256 and stored on IndexedDB.
- Clean Data Extractor: Packages purified raw telemetry and dispatches encrypted payload packages via Email or direct export.
How we built it
- Full-Stack Architecture: Built with React 18, Vite, TypeScript, and Express.js backend for server-side Gemini API proxying.
- Gemini API (@google/genai): Taps into
gemini-3.6-flashfor zero-latency cognitive evaluations and structured JSON risk metrics. - Multi-Agent Worker Mesh: Utilizes native browser Web Workers to run multi-threaded background simulations without interrupting UI responsiveness.
- WebCrypto Cryptography: Leverages standard
window.crypto.subtlefor hardware-accelerated SHA-256 digest creation and ECDSA P-384 digital signature verification. - Sophisticated Dark Aesthetic: Features an Obsidian canvas (
#0a0a0a), warm metallic gold accents (#c5a059), serif typography, and real-time metric visualizers.
Challenges we ran into
- Real-Time Synchronicity: Managing state sync across background Web Workers, client-side IndexedDB ledger, and server-side Gemini API proxies without causing UI frame drops.
- Cryptographic Performance in Browser: Ensuring ECDSA P-384 signature generation and SHA-256 block hash validation run seamlessly in under 5ms per block.
- Adversarial Noise Generation: Designing a balanced 24/7 simulation engine that generates realistic data toxins without collapsing the agent's baseline $IQ_{AI}$.
Accomplishments that we're proud of
- 100% Cryptographic Auditability: Fully verified SHA-256 & ECDSA P-384 signature verification running locally on client IndexedDB.
- Zero-Latency Gemini Integration: Server-side proxy handling Gemini 3.6 Flash queries with sub-second response times and zero client key leaks.
- Live Real-World Utility: A complete working firewall interface providing instant value to AI system owners.
What we learned
- How to combine deep cryptographic proofs (ECDSA P-384) with neural cognitive metrics ($AQ$, $EQ_{AI}$, $IQ_{AI}$).
- The importance of keeping secret keys completely server-side when integrating Gemini API in production environments.
- Best practices for multi-threaded Web Worker architecture in modern React applications.
What's next for LH-EW-2026: Cognitive Shield
- Decentralized Node Mesh: Expanding worker nodes across multi-cloud edge environments.
- Gemini 3.1 Pro Deep Research Integration: Automating deep threat intelligence scraping and automated patch generation.
- Hardware Enclave Binding: Binding Cognitive Shield signatures directly to hardware TPM / Secure Enclaves.
Built With
- ai-agents
- cloud-run
- cognitive-shield
- cybersecurity
- ecdsa-p-384
- express.js
- gemini-3.6-flash
- gemini-api
- indexeddb
- lucide-icons
- node.js
- react-18
- sha-256
- tailwind-css
- typescript
- vite
- web-workers
- webcrypto-api


Log in or sign up for Devpost to join the conversation.