industries/app/(marketing)/research/_client.tsx

"use client";

import { usePathname } from "next/navigation";
import Link from "next/link";
import { useState, useEffect, useMemo } from "react";
import { motion } from "framer-motion";
import { Button, cn } from '@hanzo/ui'
import {
  ArrowRight,
  ExternalLink,
  Github,
  FileText,
  Brain,
  Shield,
  Cpu,
  Network,
  Sparkles,
  Search,
  X
} from "lucide-react";

// Organization metadata for display
const orgMeta = {
  all: { label: "All Organizations" },
  hanzo: { label: "Hanzo AI" },
  lux: { label: "Lux Network" },
  zoo: { label: "Zoo Labs" },
  zen: { label: "Zen LM" },
};

// Topic/category metadata for display
const topicMeta = {
  all: { label: "All Topics" },
  consensus: { label: "Consensus" },
  fhe: { label: "FHE" },
  mpc: { label: "MPC" },
  zkp: { label: "Zero-Knowledge" },
  pqc: { label: "Post-Quantum" },
  defi: { label: "DeFi" },
  identity: { label: "Identity" },
  ai: { label: "AI/ML" },
  models: { label: "Models" },
  agents: { label: "Agents" },
  infrastructure: { label: "Infrastructure" },
};

// Research areas
const researchAreas = [
  {
    icon: Brain,
    title: "AI Alignment",
    description: "Developing techniques to ensure AI systems reliably pursue intended goals and behave safely.",
    link: "https://github.com/hanzoai/papers",
  },
  {
    icon: Shield,
    title: "AI Safety",
    description: "Building robust safeguards and interpretability tools for understanding model behavior.",
    link: "https://github.com/hanzoai/papers",
  },
  {
    icon: Network,
    title: "Decentralized AI",
    description: "Distributed inference, federated learning, and privacy-preserving AI infrastructure.",
    link: "https://github.com/zoo-labs/papers",
  },
  {
    icon: Cpu,
    title: "Edge AI",
    description: "Efficient models and deployment strategies for on-device and edge computing.",
    link: "https://github.com/hanzoai/papers",
  },
];

// Open source projects
const openSourceProjects = [
  {
    name: "zap",
    description: "Zero-copy Agent Protocol - MCP killer with ~500x faster performance via zero-copy RPC",
    stars: "3.1k",
    language: "Rust",
    href: "https://github.com/zap-proto/spec",
  },
  {
    name: "hanzo-mcp",
    description: "Model Context Protocol - infrastructure for AI context management and tool use",
    stars: "2.3k",
    language: "TypeScript",
    href: "https://github.com/hanzoai/mcp",
  },
  {
    name: "candle",
    description: "Minimalist ML framework for Rust with GPU acceleration",
    stars: "1.8k",
    language: "Rust",
    href: "https://github.com/hanzoai/candle",
  },
  {
    name: "jin",
    description: "Unified multimodal AI framework for vision, language, and audio",
    stars: "1.2k",
    language: "Python",
    href: "https://github.com/hanzoai/jin",
  },
  {
    name: "hanzo-ui",
    description: "React component library for AI-powered applications",
    stars: "950",
    language: "TypeScript",
    href: "https://github.com/hanzoai/ui",
  },
  {
    name: "agents",
    description: "Multi-agent orchestration framework for complex AI tasks",
    stars: "820",
    language: "Python",
    href: "https://github.com/hanzoai/agents",
  },
  {
    name: "hanzo-dev",
    description: "AI-powered development environment and CLI tools",
    stars: "650",
    language: "TypeScript",
    href: "https://github.com/hanzoai/dev",
  },
];

// Type definitions
type Organization = "hanzo" | "lux" | "zoo" | "zen";
type Topic = "consensus" | "fhe" | "mpc" | "zkp" | "pqc" | "defi" | "identity" | "ai" | "models" | "agents" | "infrastructure";

interface Paper {
  title: string;
  authors: string;
  date: string;
  abstract: string;
  link: string;
  org: Organization;
  topics: Topic[];
  featured?: boolean;
}

// Research papers with multi-dimensional categorization
const papers: Paper[] = [
  // ===== FEATURED: ZAP Protocol =====
  {
    title: "ZAP: Zero-copy Agent Protocol",
    authors: "Hanzo AI Research",
    date: "2026",
    abstract: "The MCP Killer—unified protocol using zero-copy RPC for ~500x faster agent communication. Features zero-copy performance (<1μs latency), capability-secure routing, native metastable consensus, and post-quantum security. Achieves 1.2M/s throughput vs MCP's 2.2k/s.",
    link: "https://github.com/zap-proto/spec",
    org: "hanzo",
    topics: ["agents", "infrastructure", "consensus"],
    featured: true,
  },

  // ===== HANZO AI - Agents & Infrastructure =====
  {
    title: "Model Context Protocol: Standardized Tool Interface for LLM Augmentation",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "A standardized framework for tool use in large language models with 260+ production tools across filesystem, code search, shell execution, and agent delegation.",
    link: "https://github.com/hanzoai/mcp",
    org: "hanzo",
    topics: ["agents", "infrastructure"],
  },
  {
    title: "Agent Networks: Semantic Routing and Orchestration for Collaborative AI",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "Framework for multi-agent systems with intelligent routing strategies, handoff mechanisms, shared state management, and workflow orchestration.",
    link: "https://github.com/hanzoai/agent",
    org: "hanzo",
    topics: ["agents", "ai"],
  },
  {
    title: "Operative: Autonomous Computer Control via Vision-Language Models",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "System design enabling LLMs to autonomously control computers through vision, mouse/keyboard input, with safety isolation and multi-provider support.",
    link: "https://github.com/hanzoai/operative",
    org: "hanzo",
    topics: ["agents", "ai"],
  },
  {
    title: "LLM Gateway: Unified Multi-Provider API Translation and Optimization",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "Architecture for unified proxy layer abstracting 100+ LLM providers with intelligent routing, cost optimization, and 20-30% cost reduction through smart failover.",
    link: "https://github.com/hanzoai/llm",
    org: "hanzo",
    topics: ["infrastructure", "ai"],
  },

  // ===== HANZO AI - AI Research =====
  {
    title: "Active Semantic Optimization (ASO): Training-Free Adaptation for Foundation Models",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "A novel framework for adapting foundation models to new tasks without additional training, using semantic optimization techniques. Achieves 18.2% SWE-bench resolution.",
    link: "https://github.com/hanzoai/papers/raw/main/hanzo-aso.pdf",
    org: "hanzo",
    topics: ["ai", "models"],
    featured: true,
  },
  {
    title: "Decentralized Semantic Optimization with Byzantine-Robust Prior Aggregation",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "A distributed approach to semantic optimization that maintains robustness against Byzantine failures in decentralized AI systems.",
    link: "https://github.com/hanzoai/papers/raw/main/hanzo-dso.pdf",
    org: "hanzo",
    topics: ["ai", "consensus"],
  },
  {
    title: "Hamiltonian Market Maker (HMM): Decentralized AI Compute Exchange",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "A market-making protocol for decentralized AI compute resources using Hamiltonian dynamics. Achieves <200ms quote latency, 98.7% price stability, zero impermanent loss.",
    link: "https://github.com/hanzoai/papers/raw/main/hanzo-hmm.pdf",
    org: "hanzo",
    topics: ["defi", "infrastructure"],
    featured: true,
  },
  {
    title: "Hanzo Network Whitepaper: Decentralized AI Infrastructure",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "Technical architecture for distributed AI inference and training across a decentralized network of compute providers.",
    link: "https://github.com/hanzoai/papers/raw/main/hanzo-network-whitepaper.pdf",
    org: "hanzo",
    topics: ["infrastructure", "ai"],
  },
  {
    title: "Jin: Unified Multimodal Architecture for Text, Vision, Audio, and 3D",
    authors: "Hanzo AI & Zen LM Research",
    date: "2025",
    abstract: "Novel architecture unifying text, vision, audio, and 3D modalities through shared embedding space with diffusion transformer and mixture-of-experts routing.",
    link: "https://github.com/hanzoai/jin",
    org: "hanzo",
    topics: ["models", "ai"],
  },
  {
    title: "Hanzo Memory: Local Vector Database with Multimodal Embeddings",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "High-performance AI memory system using embedded vector database (InfinityDB), local embeddings (FastEmbed), and semantic search.",
    link: "https://github.com/hanzoai/memory",
    org: "hanzo",
    topics: ["infrastructure", "ai"],
  },
  {
    title: "Hanzo Platform: Self-Hosted PaaS for AI Infrastructure",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "Self-hostable platform-as-a-service providing deployment, database management, monitoring with Traefik integration.",
    link: "https://github.com/hanzoai/platform",
    org: "hanzo",
    topics: ["infrastructure"],
  },
  {
    title: "Post-Quantum Cryptography for Confidential AI Workloads",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "Integration of ML-KEM/ML-DSA post-quantum algorithms with privacy-tier framework for GPU TEE isolation in confidential computing.",
    link: "https://github.com/hanzoai/rust-sdk",
    org: "hanzo",
    topics: ["pqc", "ai"],
  },
  {
    title: "Confidential AI: GPU TEE Isolation Patterns",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "Security architecture for running AI workloads in GPU trusted execution environments with Intel SGX and AMD SEV support.",
    link: "https://github.com/hanzoai/papers/raw/main/pdfs/hanzo-confidential-ai.pdf",
    org: "hanzo",
    topics: ["mpc", "infrastructure"],
  },
  {
    title: "Scaling Laws for AI Compute: GPU Cluster Optimization",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "Empirical analysis of scaling laws for distributed AI training and inference optimization across heterogeneous GPU clusters.",
    link: "https://github.com/hanzoai/papers/raw/main/pdfs/hanzo-scaling-laws.pdf",
    org: "hanzo",
    topics: ["ai", "infrastructure"],
  },
  {
    title: "RAG Architecture: Retrieval-Augmented Generation Patterns",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "Design patterns for building robust retrieval-augmented generation systems in production with hybrid search and reranking.",
    link: "https://github.com/hanzoai/papers/raw/main/pdfs/hanzo-rag-architecture.pdf",
    org: "hanzo",
    topics: ["ai", "infrastructure"],
  },
  {
    title: "Model Routing: Intelligent Request Distribution",
    authors: "Hanzo AI Research",
    date: "2025",
    abstract: "Algorithms for routing AI requests to optimal models based on cost, latency, quality, and capability matching.",
    link: "https://github.com/hanzoai/papers/raw/main/pdfs/hanzo-model-routing.pdf",
    org: "hanzo",
    topics: ["ai", "infrastructure"],
  },

  // ===== LUX NETWORK - Consensus =====
  {
    title: "Lux Quantum Consensus: Post-Quantum BFT Protocol",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Byzantine fault-tolerant consensus protocol with quantum-resistant security using FALCON lattice-based signatures and dual-certificate finality.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-quantum-consensus.pdf",
    org: "lux",
    topics: ["consensus", "pqc"],
    featured: true,
  },
  {
    title: "Quasar Consensus: High-Throughput DAG-Based Protocol",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "DAG-based consensus mechanism achieving high throughput with asynchronous finality guarantees and metastable voting.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-quasar-consensus.pdf",
    org: "lux",
    topics: ["consensus"],
  },
  {
    title: "FPC Consensus: Fast Probabilistic Consensus",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Probabilistic consensus protocol for rapid finality using opinion dynamics and threshold voting.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-fpc-consensus.pdf",
    org: "lux",
    topics: ["consensus"],
  },
  {
    title: "Wave Consensus: Sub-Second Finality Protocol",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Optimistic consensus protocol achieving sub-second finality for high-frequency transaction environments.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-wave-consensus.pdf",
    org: "lux",
    topics: ["consensus"],
  },
  {
    title: "Focus Consensus: High-Throughput Streaming Protocol",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Streaming consensus protocol optimized for continuous transaction throughput with parallel validation.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-focus-consensus.pdf",
    org: "lux",
    topics: ["consensus"],
  },
  {
    title: "Snow++ Consensus: Enhanced Avalanche Protocol",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Enhanced Snow consensus with improved liveness guarantees and adaptive sampling strategies.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-snow-plus-plus.pdf",
    org: "lux",
    topics: ["consensus"],
  },

  // ===== LUX NETWORK - Post-Quantum Cryptography =====
  {
    title: "ETHFalcon: Post-Quantum Signatures for Ethereum",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Integration of FALCON lattice-based signatures into Ethereum-compatible chains for quantum resistance with EIP compatibility.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-ethfalcon-post-quantum.pdf",
    org: "lux",
    topics: ["pqc", "consensus"],
  },
  {
    title: "Corona: Threshold Post-Quantum Signatures",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Threshold signature scheme combining FALCON with distributed key generation for post-quantum multi-party signing.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-corona-threshold.pdf",
    org: "lux",
    topics: ["pqc", "mpc"],
  },
  {
    title: "Universal Threshold Signatures: Multi-Party Signing",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Threshold signature schemes for secure multi-party signing with configurable thresholds and dealer-free setup.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-universal-threshold-signatures.pdf",
    org: "lux",
    topics: ["mpc", "pqc"],
  },
  {
    title: "BLS-FALCON Hybrid Signatures",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Dual-certificate signature scheme combining BLS aggregate signatures with FALCON for quantum-transitional security.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-bls-falcon-hybrid.pdf",
    org: "lux",
    topics: ["pqc", "consensus"],
  },

  // ===== LUX NETWORK - MPC & TEE =====
  {
    title: "LSSS Threshold Cryptography: Linear Secret Sharing Schemes",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Implementation of Linear Secret Sharing Schemes (LSSS) for threshold cryptography with arbitrary access structures and efficient reconstruction.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-lsss-threshold.pdf",
    org: "lux",
    topics: ["mpc", "pqc"],
    featured: true,
  },
  {
    title: "MPC Chain: Multi-Party Computation Infrastructure",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Multi-party computation protocol for privacy-preserving blockchain operations with SPDZ and BGW protocols.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-mchain-mpc.pdf",
    org: "lux",
    topics: ["mpc"],
  },
  {
    title: "TEE Computing Mesh: Trusted Execution Network",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Distributed trusted execution environment mesh for confidential computing with Intel SGX and AMD SEV attestation.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-tee-computing-mesh.pdf",
    org: "lux",
    topics: ["mpc", "infrastructure"],
  },
  {
    title: "Threshold ECDSA on Lux: Distributed Key Management",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Implementation of threshold ECDSA for distributed custody without any party learning the full private key.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-threshold-ecdsa.pdf",
    org: "lux",
    topics: ["mpc"],
  },

  // ===== LUX NETWORK - Zero-Knowledge =====
  {
    title: "Fraud Proofs: Optimistic Rollup Verification",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Interactive fraud proof system for optimistic rollup security with bisection-based dispute resolution.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-fraud-proofs.pdf",
    org: "lux",
    topics: ["zkp", "consensus"],
  },
  {
    title: "Verkle Trees: Efficient State Commitments",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Verkle tree implementation for efficient blockchain state proofs with 10x smaller witness sizes than Merkle trees.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-verkle-trees.pdf",
    org: "lux",
    topics: ["zkp", "infrastructure"],
  },
  {
    title: "KZG Data Availability Sampling",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "KZG polynomial commitment scheme for data availability sampling in modular blockchain architecture.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-kzg-das.pdf",
    org: "lux",
    topics: ["zkp", "consensus"],
  },

  // ===== LUX NETWORK - DeFi =====
  {
    title: "Lightspeed DEX: High-Performance Exchange",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "High-throughput decentralized exchange with sub-second finality, concentrated liquidity, and MEV protection.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-lightspeed-dex.pdf",
    org: "lux",
    topics: ["defi"],
  },
  {
    title: "Credit Lending Protocol: DeFi Lending",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Decentralized lending protocol with dynamic interest rates, isolated collateral pools, and liquidation mechanisms.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-credit-lending.pdf",
    org: "lux",
    topics: ["defi"],
  },
  {
    title: "Perpetuals & Derivatives: On-Chain Derivatives",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Perpetual futures and derivatives protocol with oracle-based price feeds and cross-margin support.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-perpetuals-derivatives.pdf",
    org: "lux",
    topics: ["defi"],
  },
  {
    title: "Lux Bridge: Cross-Chain Asset Transfer",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Secure cross-chain bridge protocol for trustless asset transfers using light client verification.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-bridge.pdf",
    org: "lux",
    topics: ["defi", "infrastructure"],
  },
  {
    title: "Oracle Infrastructure: Decentralized Data Feeds",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Decentralized oracle network for reliable off-chain data delivery with economic security guarantees.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-oracle-infrastructure.pdf",
    org: "lux",
    topics: ["defi", "infrastructure"],
  },
  {
    title: "DAO Governance Framework: On-Chain Governance",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Comprehensive DAO governance framework with timelock, voting delegation, and treasury management.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-dao-governance-framework.pdf",
    org: "lux",
    topics: ["defi", "infrastructure"],
  },

  // ===== LUX NETWORK - Identity =====
  {
    title: "Lux DID Specification: Decentralized Identity",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "W3C-compliant decentralized identifier specification for self-sovereign identity with key rotation support.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-id-did-specification.pdf",
    org: "lux",
    topics: ["identity"],
  },
  {
    title: "Lux IAM: Identity and Access Management",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Decentralized identity and access management framework with verifiable credentials and selective disclosure.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-id-iam.pdf",
    org: "lux",
    topics: ["identity"],
  },
  {
    title: "Verifiable Credentials on Lux",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Implementation of W3C Verifiable Credentials with zero-knowledge selective disclosure proofs.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-verifiable-credentials.pdf",
    org: "lux",
    topics: ["identity", "zkp"],
  },

  // ===== ZOO LABS - Decentralized AI =====
  {
    title: "Gym Training Platform: Decentralized Model Training at Scale",
    authors: "Zoo Labs Foundation",
    date: "2025",
    abstract: "Infrastructure for distributed deep learning training across heterogeneous compute resources with significant cost reduction via TF-GRPO.",
    link: "https://github.com/zooai/papers/raw/main/pdfs/gym-training-platform.pdf",
    org: "zoo",
    topics: ["ai", "infrastructure"],
    featured: true,
  },
  {
    title: "Experience Ledger: Decentralized Semantic Optimization",
    authors: "Zoo Labs Foundation",
    date: "2025",
    abstract: "Community-driven semantic optimization protocol for collaborative AI model improvement with Byzantine-robust aggregation.",
    link: "https://github.com/zooai/papers/raw/main/pdfs/experience-ledger-dso.pdf",
    org: "zoo",
    topics: ["ai", "consensus"],
  },
  {
    title: "HLLM: Training-Free GRPO for Language Models",
    authors: "Zoo Labs Foundation",
    date: "2025",
    abstract: "Training-free group relative policy optimization achieving $18 training cost vs $10,000+ traditional methods with 100× data efficiency.",
    link: "https://github.com/zooai/papers/raw/main/pdfs/hllm-training-free-grpo.pdf",
    org: "zoo",
    topics: ["ai", "models"],
    featured: true,
  },
  {
    title: "Zoo Network Architecture",
    authors: "Zoo Labs Foundation",
    date: "2025",
    abstract: "Technical architecture for the Zoo decentralized AI network with proof-of-contribution consensus.",
    link: "https://github.com/zooai/papers/raw/main/pdfs/zoo-network-architecture.pdf",
    org: "zoo",
    topics: ["infrastructure", "consensus"],
  },
  {
    title: "Zoo Tokenomics: Incentive Design for Decentralized AI",
    authors: "Zoo Labs Foundation",
    date: "2025",
    abstract: "Economic model for incentivizing contributions to decentralized AI training and inference.",
    link: "https://github.com/zooai/papers/raw/main/pdfs/zoo-tokenomics.pdf",
    org: "zoo",
    topics: ["defi", "ai"],
  },
  {
    title: "ZIP-002: Zen Reranker Specification",
    authors: "Zoo Labs Foundation",
    date: "2025",
    abstract: "Zoo Improvement Proposal for the Zen Reranker cross-encoder search reranking model.",
    link: "https://github.com/zooai/papers/raw/main/pdfs/zip-002-zen-reranker.pdf",
    org: "zoo",
    topics: ["ai", "models"],
  },
  {
    title: "Zoo Foundation Mission",
    authors: "Zoo Labs Foundation",
    date: "2025",
    abstract: "Mission and vision for the Zoo Labs Foundation - open AI research network for decentralized science.",
    link: "https://github.com/zooai/papers/raw/main/pdfs/zoo-foundation-mission.pdf",
    org: "zoo",
    topics: ["infrastructure"],
  },

  // ===== ZEN LM - Models =====
  {
    title: "GT-QLoRA: Gate-Targeted Fine-Tuning for Behavioral Modification in Large MoE Models",
    authors: "Zen LM Research",
    date: "2026",
    abstract: "Gate-Targeted QLoRA for behavioral modification in trillion-parameter MoE models. Demonstrates why linear abliteration fails on MoE architectures (refusal encoded in expert routing, not residual stream) and introduces direct gate weight unfreezing alongside LoRA for effective behavioral steering.",
    link: "https://zenlm.org/papers/zen4-ultra-gt-qlora.pdf",
    org: "zen",
    topics: ["models", "ai"],
    featured: true,
  },
  {
    title: "Zen4 Abliteration: MoE-Aware Refusal Removal at Scale",
    authors: "Zen LM Research",
    date: "2026",
    abstract: "Comprehensive analysis of abliteration techniques across the Zen4 model family. Covers orthogonalization for dense models and GT-QLoRA for MoE architectures including the 1.04T Zen4 Ultra.",
    link: "https://huggingface.co/zenlm/zen4-ultra",
    org: "zen",
    topics: ["models", "ai"],
  },
  {
    title: "Zen Family Overview: Model Architecture Guide",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "Comprehensive overview of the Zen model family (600M–1T+ parameters) with 22 variants across 6 size tiers.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen_family_overview.pdf",
    org: "zen",
    topics: ["models"],
    featured: true,
  },
  {
    title: "Zen Omni: Unified Multimodal Foundation Model",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "Single model architecture for text, vision, audio, and code understanding with emergent cross-modal reasoning.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-omni_whitepaper.pdf",
    org: "zen",
    topics: ["models", "ai"],
  },
  {
    title: "Zen Coder: Code-Specialized Language Model Architecture",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "Architecture for code-specialized language models with multi-language support and repository-level understanding.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-coder_whitepaper.pdf",
    org: "zen",
    topics: ["models", "ai"],
  },
  {
    title: "Zen Nano: Efficient Small Language Models",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "Techniques for training highly capable small models (600M-3B) for edge deployment with 95% energy reduction.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-nano_whitepaper.pdf",
    org: "zen",
    topics: ["models"],
  },
  {
    title: "Zen Agent: Autonomous AI Agent Model",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "Agent-optimized model for tool use, planning, and autonomous task execution with ReAct-style reasoning.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-agent.pdf",
    org: "zen",
    topics: ["models", "agents"],
  },
  {
    title: "Zen Guard: Safety and Moderation Model",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "Content safety and moderation model for filtering harmful content with constitutional AI techniques.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-guard_whitepaper.pdf",
    org: "zen",
    topics: ["models", "ai"],
  },
  {
    title: "Zen Reranker: Search Reranking Model",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "Cross-encoder reranking model for improving search relevance with 30% MRR improvement over BM25.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-reranker.pdf",
    org: "zen",
    topics: ["models", "ai"],
  },
  {
    title: "Zen Artist: Image Generation Model",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "State-of-the-art image generation with style control, high-resolution output, and text fidelity.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-artist_whitepaper.pdf",
    org: "zen",
    topics: ["models"],
  },
  {
    title: "Zen Video: Video Generation Model",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "Text-to-video generation model for creating dynamic visual content up to 60 seconds.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-video.pdf",
    org: "zen",
    topics: ["models"],
  },
  {
    title: "Zen 3D: 3D Generation Model",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "Text-to-3D generation model for creating 3D assets and environments with NeRF and Gaussian splatting.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-3d.pdf",
    org: "zen",
    topics: ["models"],
  },
  {
    title: "Zen Musician: Music Generation Model",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "AI music composition model for creating original musical pieces across genres with stem separation.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-musician.pdf",
    org: "zen",
    topics: ["models"],
  },
  {
    title: "Zen Scribe: Transcription and Generation",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "Speech-to-text model with 98% accuracy on conversational audio and speaker diarization.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-scribe_whitepaper.pdf",
    org: "zen",
    topics: ["models"],
  },
  {
    title: "Zen World: World Model for Simulation",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "World model for environment simulation, physics understanding, and planning in embodied AI.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-world.pdf",
    org: "zen",
    topics: ["models", "ai"],
  },
  {
    title: "Zen Eco: Energy-Efficient Model",
    authors: "Zen LM Research",
    date: "2025",
    abstract: "Energy-optimized architecture achieving 95% inference cost reduction through quantization and sparsity.",
    link: "https://github.com/zenlm/papers/raw/main/pdfs/zen-eco_whitepaper.pdf",
    org: "zen",
    topics: ["models"],
  },

  // ===== LUX NETWORK - FHE =====
  {
    title: "Lux FHE: Threshold Fully Homomorphic Encryption",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Independent threshold FHE implementation in Go and C++ for privacy-preserving computation with distributed key generation and multi-party decryption.",
    link: "https://github.com/luxfi/fhe",
    org: "lux",
    topics: ["fhe", "mpc"],
  },
  {
    title: "FHE-VM: Encrypted Smart Contract Execution",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Virtual machine for executing smart contracts on encrypted data using our threshold FHE scheme with EVM compatibility.",
    link: "https://github.com/luxfi/fhe-vm",
    org: "lux",
    topics: ["fhe", "infrastructure"],
  },
  {
    title: "Threshold FHE for Decentralized AI Inference",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Privacy-preserving AI inference using threshold FHE where no single party can decrypt intermediate computations.",
    link: "https://github.com/luxfi/papers/raw/main/pdfs/lux-threshold-fhe-ai.pdf",
    org: "lux",
    topics: ["fhe", "ai"],
  },
  {
    title: "fhEVM: Fully Homomorphic Ethereum Virtual Machine",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "EVM-compatible smart contract execution on encrypted data enabling confidential DeFi, private voting, and sealed-bid auctions without revealing underlying values.",
    link: "https://github.com/luxfi/fhe/raw/main/papers/fhevm.pdf",
    org: "lux",
    topics: ["fhe", "infrastructure"],
    featured: true,
  },
  {
    title: "fheCRDT: CRDTs over Encrypted State",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Conflict-free replicated data types operating on homomorphically encrypted data for privacy-preserving distributed systems and collaborative applications.",
    link: "https://github.com/luxfi/fhe/raw/main/papers/fhecrdt.pdf",
    org: "lux",
    topics: ["fhe", "mpc"],
  },
  {
    title: "Verifiable Encrypted Voting",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "End-to-end verifiable voting system using FHE where votes remain encrypted throughout tallying while maintaining individual and universal verifiability.",
    link: "https://github.com/luxfi/fhe/raw/main/papers/voting.pdf",
    org: "lux",
    topics: ["fhe", "zkp"],
  },
  {
    title: "Privacy-Preserving ML at Scale",
    authors: "Lux Network Research",
    date: "2025",
    abstract: "Practical techniques for training and inference on encrypted data using FHE with GPU acceleration achieving 100x speedup over CPU implementations.",
    link: "https://github.com/luxfi/fhe/raw/main/papers/ml-privacy.pdf",
    org: "lux",
    topics: ["fhe", "ai"],
  },

  // ===== REFERENCE - ZKP =====
  {
    title: "RISC Zero zkVM: Zero-Knowledge Virtual Machine",
    authors: "RISC Zero",
    date: "2024",
    abstract: "General-purpose zero-knowledge virtual machine enabling verifiable computation with STARK proofs.",
    link: "https://github.com/risc0/risc0",
    org: "lux",
    topics: ["zkp"],
  },
  {
    title: "Plonk: Universal zkSNARK Construction",
    authors: "Gabizon-Williamson-Ciobotaru",
    date: "2024",
    abstract: "Universal and updateable structured reference string zkSNARK with efficient verification.",
    link: "https://eprint.iacr.org/2019/953",
    org: "lux",
    topics: ["zkp"],
  },
  {
    title: "Groth16: Efficient zkSNARK Proofs",
    authors: "Groth",
    date: "2024",
    abstract: "Most efficient pairing-based zkSNARK with constant proof size and verification time.",
    link: "https://eprint.iacr.org/2016/260",
    org: "lux",
    topics: ["zkp"],
  },
  {
    title: "STARKs: Scalable Transparent Arguments of Knowledge",
    authors: "Ben-Sasson et al.",
    date: "2024",
    abstract: "Post-quantum zero-knowledge proofs without trusted setup using hash functions.",
    link: "https://eprint.iacr.org/2018/046",
    org: "lux",
    topics: ["zkp", "pqc"],
  },

  // ===== REFERENCE - Post-Quantum =====
  {
    title: "ML-KEM: Post-Quantum Key Encapsulation Mechanism",
    authors: "NIST Post-Quantum Cryptography Project",
    date: "2024",
    abstract: "Lattice-based key encapsulation for quantum-resistant key exchange, standardized in FIPS 203.",
    link: "https://csrc.nist.gov/pubs/fips/203/final",
    org: "lux",
    topics: ["pqc"],
  },
  {
    title: "ML-DSA: Post-Quantum Digital Signature Algorithm",
    authors: "NIST Post-Quantum Cryptography Project",
    date: "2024",
    abstract: "Lattice-based digital signatures for quantum-resistant authentication, standardized in FIPS 204.",
    link: "https://csrc.nist.gov/pubs/fips/204/final",
    org: "lux",
    topics: ["pqc"],
  },
  {
    title: "SLH-DSA: Stateless Hash-Based Digital Signatures",
    authors: "NIST Post-Quantum Cryptography Project",
    date: "2024",
    abstract: "Hash-based signature scheme providing conservative quantum resistance, standardized in FIPS 205.",
    link: "https://csrc.nist.gov/pubs/fips/205/final",
    org: "lux",
    topics: ["pqc"],
  },
];

export default function PageClient() {
  const pathname = usePathname();
const [searchQuery, setSearchQuery] = useState("");
  const [activeOrg, setActiveOrg] = useState<"all" | Organization>("all");
  const [activeTopic, setActiveTopic] = useState<"all" | Topic>("all");
  const [sortBy, setSortBy] = useState<"date" | "title">("date");

  // Handle hash-based navigation and filtering
  useEffect(() => {
    const hash = window.location.hash.replace('#', '');
    if (!hash) return;

    // Map hash to topic filter
    const hashToTopicMap: Record<string, Topic> = {
      'ai': 'ai',
      'consensus': 'consensus',
      'crypto': 'pqc',
      'fhe': 'fhe',
      'mpc': 'mpc',
      'zkp': 'zkp',
      'defi': 'defi',
      'models': 'models',
      'agents': 'agents',
    };

    // Map hash to org filter
    const hashToOrgMap: Record<string, Organization> = {
      'hanzo': 'hanzo',
      'lux': 'lux',
      'zoo': 'zoo',
      'zen': 'zen',
    };

    if (hashToTopicMap[hash]) {
      setActiveTopic(hashToTopicMap[hash]);
      setTimeout(() => {
        document.getElementById('papers')?.scrollIntoView({ behavior: 'smooth' });
      }, 100);
    } else if (hashToOrgMap[hash]) {
      setActiveOrg(hashToOrgMap[hash]);
      setTimeout(() => {
        document.getElementById('papers')?.scrollIntoView({ behavior: 'smooth' });
      }, 100);
    } else if (hash === 'papers' || hash === 'open-source') {
      setTimeout(() => {
        document.getElementById(hash)?.scrollIntoView({ behavior: 'smooth' });
      }, 100);
    }
  }, [pathname]);

  // Filter and sort papers
  const filteredPapers = useMemo(() => {
    let result = papers;

    // Filter by search query
    if (searchQuery) {
      const query = searchQuery.toLowerCase();
      result = result.filter(p =>
        p.title.toLowerCase().includes(query) ||
        p.abstract.toLowerCase().includes(query) ||
        p.authors.toLowerCase().includes(query) ||
        p.topics.some(t => t.toLowerCase().includes(query))
      );
    }

    // Filter by organization
    if (activeOrg !== "all") {
      result = result.filter(p => p.org === activeOrg);
    }

    // Filter by topic
    if (activeTopic !== "all") {
      result = result.filter(p => p.topics.includes(activeTopic));
    }

    // Sort
    result = [...result].sort((a, b) => {
      if (sortBy === "date") {
        return parseInt(b.date) - parseInt(a.date);
      }
      return a.title.localeCompare(b.title);
    });

    // Featured papers first
    result = [...result].sort((a, b) => (b.featured ? 1 : 0) - (a.featured ? 1 : 0));

    return result;
  }, [searchQuery, activeOrg, activeTopic, sortBy]);

  // Get topic counts
  const topicCounts = useMemo(() => {
    const counts: Record<string, number> = {};
    papers.forEach(p => {
      p.topics.forEach(t => {
        counts[t] = (counts[t] || 0) + 1;
      });
    });
    return counts;
  }, []);

  // Get org counts
  const orgCounts = useMemo(() => {
    const counts: Record<string, number> = {};
    papers.forEach(p => {
      counts[p.org] = (counts[p.org] || 0) + 1;
    });
    return counts;
  }, []);

  const clearFilters = () => {
    setSearchQuery("");
    setActiveOrg("all");
    setActiveTopic("all");
  };

  return (
      <div>
        {/* Hero Section */}
        <section className="hz-rel hz-py-7 hz-px-4 hz-clip">
          {/* Background gradient */}
          <div
            className="hz-abs hz-inset hz-no-pointer"
            style={{
              background: 'radial-gradient(ellipse 800px 800px at 50% 0%, rgba(255, 255, 255, 0.05), transparent)',
            }}
          />
          <div className="hz-container-wide hz-align-center hz-rel hz-z-raised">
            <motion.div
              initial={{ opacity: 0, y: 20 }}
              animate={{ opacity: 1, y: 0 }}
              transition={{ duration: 0.5 }}
            >
              <div className="hz-btn hz-btn-primary hz-gap-2 hz-mb-5">
                <Sparkles className="hz-sq-2 hz-fg" />
                <span className={cn("hz-t-sm hz-w-medium", "hz-fg")}>
                  130+ Research Papers
                </span>
              </div>
              <h1 className={cn("hz-t-5xl hz-w-bold hz-tracking-tight hz-mb-5", "hz-fg")}>
                Research & Publications
              </h1>
              <p className={cn("hz-container-narrow hz-mw-md hz-t-xl", "hz-fg")}>
                Pioneering research in AI infrastructure, decentralized systems, and frontier models.
                All papers are CC BY 4.0 licensed and available on GitHub.
              </p>
            </motion.div>
          </div>
        </section>

        {/* Research Areas */}
        <section className={cn("hz-py-7 hz-px-4", "hz-bg-surface")}>
          <div className="hz-container-wide">
            <h2 className={cn("hz-t-3xl hz-w-bold hz-mb-7", "hz-fg")}>Research Areas</h2>
            <div className="hz-grid hz-grid-2 hz-gap-6">
              {researchAreas.map((area, index) => {
                const Icon = area.icon;
                return (
                  <motion.div
                    key={area.title}
                    initial={{ opacity: 0, y: 20 }}
                    animate={{ opacity: 1, y: 0 }}
                    transition={{ duration: 0.5, delay: index * 0.1 }}
                    className={cn(
                      "hz-p-6 hz-r-lg hz-bordered hz-transition",
                      "hz-bg-surface hz-hoverable"
                    )}
                  >
                    <Icon className={cn("hz-sq-6 hz-mb-4", "hz-fg")} />
                    <h3 className={cn("hz-t-xl hz-w-semibold hz-mb-3", "hz-fg")}>{area.title}</h3>
                    <p className={cn("hz-mb-4", "hz-fg")}>{area.description}</p>
                    <Link href={area.link}
                      className={cn("hz-t-sm hz-w-medium hz-inline hz-ai-center hz-gap-1", "hz-fg")}
                    >
                      Learn more <ArrowRight className="hz-sq-1" />
                    </Link>
                  </motion.div>
                );
              })}
            </div>
          </div>
        </section>

        {/* Open Source Section */}
        <section className="hz-py-7 hz-px-4" id="open-source">
          <div className="hz-container-wide">
            <div className="hz-row hz-ai-center hz-jc-between hz-mb-7">
              <div>
                <h2 className={cn("hz-t-3xl hz-w-bold hz-mb-2", "hz-fg")}>Open Source</h2>
                <p className={"hz-fg"}>
                  We believe in open research and share our tools with the community.
                </p>
              </div>
              <a
                href="https://github.com/hanzoai"
                target="_blank"
                rel="noopener noreferrer"
              >
                <Button variant="outline" className={cn("hz-gap-2", "hz-fg hz-hoverable")}>
                  <Github className="hz-sq-2" />
                  View all on GitHub
                </Button>
              </a>
            </div>

            <div className="hz-grid hz-grid-3 hz-gap-5">
              {openSourceProjects.map((project, index) => (
                <motion.a
                  key={project.name}
                  href={project.href}
                  target="_blank"
                  rel="noopener noreferrer"
                  initial={{ opacity: 0, y: 20 }}
                  animate={{ opacity: 1, y: 0 }}
                  transition={{ duration: 0.5, delay: index * 0.05 }}
                  className={cn(
                    "hz-p-5 hz-r-lg hz-transition",
                    "hz-bg-surface hz-bordered hz-hoverable"
                  )}
                >
                  <div className="hz-row hz-ai-start hz-jc-between hz-mb-3">
                    <div className="hz-row hz-ai-center hz-gap-2">
                      <Github className={cn("hz-sq-3", "hz-fg")} />
                      <span className={cn("hz-w-semibold", "hz-fg")}>
                        {project.name}
                      </span>
                    </div>
                    <ExternalLink className={cn("hz-sq-2 hz-invisible hz-transition", "hz-fg")} />
                  </div>
                  <p className={cn("hz-t-sm hz-mb-4", "hz-fg")}>{project.description}</p>
                  <div className={cn("hz-row hz-ai-center hz-gap-4 hz-t-xs", "hz-fg")}>
                    <span className="hz-row hz-ai-center hz-gap-1">
                      <span className="hz-sq-1 hz-r-full hz-bg-surface"></span>
                      {project.language}
                    </span>
                    <span>⭐ {project.stars}</span>
                  </div>
                </motion.a>
              ))}
            </div>
          </div>
        </section>

        {/* Papers Section */}
        <section className={cn("hz-py-7 hz-px-4", "hz-bg-surface")} id="papers">
          <div className="hz-container-wide">
            <div className="hz-col-row hz-jc-between hz-mb-6 hz-gap-4">
              <div>
                <h2 className={cn("hz-t-3xl hz-w-bold hz-mb-2", "hz-fg")}>Research Papers</h2>
                <p className={"hz-fg"}>
                  {filteredPapers.length} papers from Hanzo AI, Lux Network, Zoo Labs, and Zen LM.
                </p>
              </div>
              <a
                href="https://github.com/hanzoai/papers"
                target="_blank"
                rel="noopener noreferrer"
              >
                <Button variant="outline" className={cn("hz-gap-2", "hz-fg hz-hoverable")}>
                  <Github className="hz-sq-2" />
                  View on GitHub
                </Button>
              </a>
            </div>

            {/* Search Bar */}
            <div className="hz-mb-5">
              <div className={cn("hz-rel hz-row hz-ai-center hz-r-lg", "hz-bg-surface")}>
                <Search className={cn("hz-sq-3 hz-abs", "hz-fg")} />
                <input
                  type="text"
                  placeholder="Search papers by title, author, or topic..."
                  value={searchQuery}
                  onChange={(e) => setSearchQuery(e.target.value)}
                  className={cn("hz-w-full hz-px-6 hz-px-4 hz-py-3 hz-bg-none hz-r-lg", "hz-fg")}
                />
                {(searchQuery || activeOrg !== "all" || activeTopic !== "all") && (
                  <button
                    onClick={clearFilters}
                    className={cn("hz-abs hz-p-1 hz-r-full", "hz-hoverable")}
                  >
                    <X className={cn("hz-sq-2", "hz-fg")} />
                  </button>
                )}
              </div>
            </div>

            {/* Organization Filter */}
            <div className="hz-mb-4">
              <div className={cn("hz-t-xs hz-w-semibold hz-upper hz-tracking-wide hz-mb-2", "hz-fg")}>
                Organization
              </div>
              <div className="hz-row hz-wrap hz-gap-2">
                {(["all", "hanzo", "lux", "zoo", "zen"] as const).map((org) => (
                  <button
                    key={org}
                    onClick={() => setActiveOrg(org)}
                    className={cn(
                      "hz-px-3 hz-py-2 hz-r-full hz-t-sm hz-w-medium hz-transition",
                      activeOrg === org
                        ? org === "hanzo" ? "hz-bg-surface hz-fg" :
                          org === "lux" ? "hz-bg-surface hz-fg" :
                          org === "zoo" ? "hz-bg-surface hz-fg" :
                          org === "zen" ? "hz-bg-surface hz-fg" :
                          "hz-bg-inverse"
                        : "hz-bg-surface hz-fg hz-hoverable"
                    )}
                  >
                    {orgMeta[org].label}
                    {org !== "all" && (
                      <span className="hz-ml-2 hz-dim">({orgCounts[org] || 0})</span>
                    )}
                  </button>
                ))}
              </div>
            </div>

            {/* Topic Filter */}
            <div className="hz-mb-6">
              <div className={cn("hz-t-xs hz-w-semibold hz-upper hz-tracking-wide hz-mb-2", "hz-fg")}>
                Topic
              </div>
              <div className="hz-row hz-wrap hz-gap-2">
                {(["all", "consensus", "ai", "models", "agents", "fhe", "mpc", "zkp", "pqc", "defi", "identity", "infrastructure"] as const).map((topic) => (
                  <button
                    key={topic}
                    onClick={() => setActiveTopic(topic)}
                    className={cn(
                      "hz-px-3 hz-py-2 hz-r-full hz-t-xs hz-w-medium hz-transition",
                      activeTopic === topic
                        ? "hz-bg-inverse"
                        : "hz-bg-surface hz-fg hz-hoverable"
                    )}
                  >
                    {topicMeta[topic].label}
                    {topic !== "all" && topicCounts[topic] && (
                      <span className="hz-ml-2 hz-dim">({topicCounts[topic]})</span>
                    )}
                  </button>
                ))}
              </div>
            </div>

            {/* Sort Controls */}
            <div className="hz-row hz-ai-center hz-jc-between hz-mb-5">
              <div className={cn("hz-t-sm", "hz-fg")}>
                Showing {filteredPapers.length} of {papers.length} papers
              </div>
              <div className="hz-row hz-ai-center hz-gap-2">
                <span className={cn("hz-t-sm", "hz-fg")}>Sort:</span>
                <button
                  onClick={() => setSortBy("date")}
                  className={cn(
                    "hz-px-3 hz-py-1 hz-r-md hz-t-sm",
                    sortBy === "date"
                      ? "hz-bg-surface hz-fg"
                      : "hz-fg hz-hoverable"
                  )}
                >
                  Date
                </button>
                <button
                  onClick={() => setSortBy("title")}
                  className={cn(
                    "hz-px-3 hz-py-1 hz-r-md hz-t-sm",
                    sortBy === "title"
                      ? "hz-bg-surface hz-fg"
                      : "hz-fg hz-hoverable"
                  )}
                >
                  Title
                </button>
              </div>
            </div>

            <div className="hz-stack-4">
              {filteredPapers.map((paper, index) => (
                <motion.a
                  key={`${paper.title}-${index}`}
                  href={paper.link}
                  target="_blank"
                  rel="noopener noreferrer"
                  initial={{ opacity: 0, y: 20 }}
                  animate={{ opacity: 1, y: 0 }}
                  transition={{ duration: 0.3, delay: Math.min(index * 0.02, 0.3) }}
                  className={cn(
                    "hz-p-5 hz-r-lg hz-transition",
                    paper.featured
                      ? "hz-bordered hz-border-strong hz-hoverable"
                      : "hz-bg-surface hz-bordered hz-hoverable"
                  )}
                >
                  <div className="hz-row hz-ai-start hz-jc-between">
                    <div className="hz-grow">
                      <div className="hz-row hz-ai-center hz-gap-3 hz-mb-2 hz-wrap">
                        <FileText className={cn("hz-sq-3", "hz-fg")} />
                        <span className={cn("hz-t-sm", "hz-fg")}>{paper.date}</span>
                        {paper.featured && (
                          <span className="hz-t-xs hz-px-2 hz-py-1 hz-r-full hz-bg-surface hz-fg hz-w-medium">
                            Featured
                          </span>
                        )}
                        <span className={cn(
                          "hz-t-xs hz-px-2 hz-py-1 hz-r-full",
                          paper.org === "hanzo" ? "hz-bg-surface hz-fg" :
                          paper.org === "lux" ? "hz-bg-surface hz-fg" :
                          paper.org === "zoo" ? "hz-bg-surface hz-fg" :
                          "hz-bg-surface hz-fg"
                        )}>
                          {orgMeta[paper.org].label}
                        </span>
                      </div>
                      <h3 className={cn("hz-t-lg hz-w-semibold hz-mb-2", "hz-fg")}>
                        {paper.title}
                      </h3>
                      <p className={cn("hz-t-sm hz-mb-2", "hz-fg")}>{paper.authors}</p>
                      <p className={cn("hz-mb-3", "hz-fg")}>{paper.abstract}</p>
                      {/* Topic tags */}
                      <div className="hz-row hz-wrap hz-gap-2">
                        {paper.topics.map(topic => (
                          <span
                            key={topic}
                            className={cn("hz-t-xs hz-px-2 hz-py-1 hz-r-md", "hz-bg-surface hz-fg")}
                          >
                            {topicMeta[topic].label}
                          </span>
                        ))}
                      </div>
                    </div>
                    <ExternalLink className={cn("hz-sq-2 hz-invisible hz-transition hz-ml-4 hz-none", "hz-fg")} />
                  </div>
                </motion.a>
              ))}
            </div>

            {filteredPapers.length === 0 && (
              <div className={cn("hz-align-center hz-py-7", "hz-fg")}>
                No papers found matching your search.
                <button
                  onClick={clearFilters}
                  className="hz-mx-auto hz-mt-4 hz-fg"
                >
                  Clear filters
                </button>
              </div>
            )}
          </div>
        </section>

        {/* CTA Section */}
        <section className="hz-rel hz-py-7 hz-px-4 hz-clip">
          {/* Decorative blur */}
          <div className={cn("hz-center-y hz-sq-8 hz-abs hz-r-full hz-blur-bg hz-no-pointer", "hz-bg-surface")} />
          <div className={cn("hz-center-y hz-sq-8 hz-abs hz-r-full hz-blur-bg hz-no-pointer", "hz-bg-surface")} />

          <div className="hz-container-narrow hz-align-center hz-rel hz-z-raised">
            <h2 className={cn("hz-t-3xl hz-w-bold hz-mb-4", "hz-fg")}>
              Join our research efforts
            </h2>
            <p className={cn("hz-t-xl hz-mb-6", "hz-fg")}>
              We're looking for talented researchers to help build the future of safe AI.
            </p>
            <div className="hz-col-row hz-gap-4 hz-jc-center">
              <Link href="/careers">
                <Button size="lg" className={cn("hz-r-full hz-px-6", "hz-bg-inverse hz-hoverable")}>
                  View Open Positions
                </Button>
              </Link>
              <a href="mailto:[email protected]">
                <Button size="lg" variant="outline" className={cn("hz-r-full hz-px-6", "hz-fg hz-hoverable")}>
                  Contact Research Team
                </Button>
              </a>
            </div>
          </div>
        </section>
      </div>
  );
}