This edition of the newsletter dives into part 2 of TEE and its role in improving transaction ordering in Unichain's rollups to ensuring private AI model execution in Phala Network. This technology creates trustless systems that maintain both security and performance, marking a significant advancement in how sensitive computations are handled across decentralized platforms. We'll also share some interesting articles, portfolio updates and market highlights.
1. Research Articles
a) AI Agents: Research & Applications
• This comprehensive piece explores the emerging landscape of Large Language Model (LLM)-based autonomous agents and their applications across various domains, including gaming, finance, robotics, governance, and sciences.
• The article details how these agents combine perception, memory, reasoning, and planning modules to perform complex tasks independently, while also discussing their potential future implications through frameworks like multi-agent systems and specific implementations in areas such as blockchain technology and social simulations.
b) Exploring Crypto Data I: Data Flow Architectures
• In cryptocurrency systems, data flows through a complex pipeline starting with client nodes retrieving blockchain data, which then undergoes decoding (using contract ABIs), transformation through specialized engines, and integration with off-chain data sources like price feeds to produce meaningful metrics.
• The process requires careful management of request pipelines, handling of both raw and decoded on-chain data, aggregation across multiple protocols and chains, and special consideration for view functions and state changes, ultimately enabling valuable blockchain insights and analytics.
c) Bitcoin L2s
• This comprehensive report analyzes the emerging landscape of Bitcoin Layer 2 (L2) solutions, including Rollups and Sidechains, which aim to enable DeFi applications and yield-generating opportunities for Bitcoin while addressing its base layer limitations such as lack of Turing-completeness.
• The analysis reveals that Bitcoin L2s have attracted $447m in VC funding since 2018, with a projected total addressable market of up to $47bn by 2030, though success will largely depend on their ability to compete with existing wrapped Bitcoin solutions on Ethereum by offering higher yields and more robust DeFi ecosystems.
2. Portfolio Highlights
a) Blendy
• Blendy is a Solana-based money market protocol that enables users to borrow stablecoins against their memecoin holdings, addressing the $55 billion memecoin market through two innovative risk modules: the Moon Module (which uses put options to protect against downside volatility) and the Throttle Module (which dynamically adjusts lending and borrowing rates).
• The platform aims to unlock new DeFi primitives by making volatile memecoins usable as collateral while protecting users from full liquidation through its safety fee system and hedging mechanisms.
b) SeaSeed Network
• Tokenized assets are digital representations of real-world assets (like real estate, commodities, and art) on blockchain technology, allowing fractional ownership and making previously inaccessible investments more available to a broader range of investors.
• SeaSeed Network is positioned as a platform for tokenized asset investment, offering institutional-grade security and partnerships with licensed financial institutions across Asia-Pacific, while utilizing a permissioned public blockchain model to ensure regulatory compliance and seamless integration with Web3 ecosystems.
c) Fluence Network
• GlobalStake, a carbon-negative infrastructure provider operating in tier 4 & 5 data centers globally, has joined Fluence Network as a compute provider to offer DePIN (Decentralized Physical Infrastructure Network) compute resources at prices up to 80% lower than traditional cloud providers.
• The company operates its own bare metal infrastructure across multiple locations including Washington DC, Phoenix, Amsterdam, and plans to expand to every habitable continent by 2025, while maintaining SOC-2 compliance and managing over 150 Petabytes of storage across various blockchain networks.
3. Part 2: Applications and Projects that utilize TEEs
Rollup Boost - Unichain powered by Flashbots x Uniswap - Rollup Boost introduces the concept of rollup extensions, enhancing programmability, performance, and decentralization.
Key Features
• 250ms Block Times: Delivers faster confirmation, native revert protection, and higher gas throughput
• Verifiable Priority Ordering: Ensures fair MEV internalization and execution guarantees for users (powered by TEEs)
Status Quo
The rollup-centric roadmap faces a fundamental tradeoff: optimizing for best execution often conflicts with preserving decentralization. Fast confirmation times (e.g., 400ms block times on Solana) enhance user experience by prioritizing execution speed but tend to sacrifice decentralization. Striking a balance between speed and decentralization remains a complex challenge.
MEV extraction remains an inherent challenge, manifesting in several forms: explicit auctions where transactions compete for block inclusion, spam auctions that create congestion and bidding wars (seen in the Solana ecosystem), and latency-related timing games that further incentivize centralization factors such as geographical colocation and vertical integration.
Among various architectural advancements, Trusted Execution Environments (TEEs) stand out as a pivotal solution for reducing trust requirements without compromising scalability.
TEE Implementation
TEEs facilitate scalable solutions across trust boundaries by securely delegating computation and ensuring integrity through remote attestation. Their strong security guarantees enable users to make informed decisions and optimize execution, effectively internalizing MEV.
Pre-confirmations offer users early assurance that their transactions or data will be executed with integrity. By leveraging the secure environment of a TEE, transactions are processed and verified in real-time before being finalized on-chain. This process effectively locks in the transaction outcome, providing guarantees that mitigate risks such as reverts or reordering, even before final block confirmation. This early-stage validation creates a bridge between trustless execution and real-time responsiveness, which is especially valuable for latency-sensitive use cases like DeFi, gaming, and high-frequency trading.
Verifiable Priority Ordering
Verifiable priority ordering enables users to validate the execution sequence of their transactions by utilizing the secure information guarantees provided by the TEE. This capability to commit to any predefined ordering rule unlocks a broad design space for innovative ordering algorithms.
A priority-based ordering rule allows applications to capture and redistribute MEV in a controlled and transparent manner, ensuring fair value allocation. In the event of TEE degradation, the system gracefully reverts to the traditional priority fee-based ordering mechanism, preserving functionality and reliability while minimizing disruption.
Unichain leverages verifiable ordering to:
• Decentralize Block Building: Ensures that transactions will not only be executed but also guarantees their specific execution order, enhancing transparency and trust
• Internalize MEV for Liquidity Providers (LPs): Reduces adverse selection costs by allowing LPs to capture and manage MEV more effectively
• Attract and Retain Third-Party Applications: Provides a clear and predictable environment for apps that previously struggled to internalize MEV, fostering innovation and long-term engagement
Azuki x TEEs
Beyond MEV, TEEs unlock a new realm of social experiences by offering trust guarantees that enable novel interactions and secure engagements.
Traditional social platforms are governed by centralized structures, which often result in issues such as opaque data practices, limited user control over personal information, and challenges in transferring profiles and reputations between platforms. Moreover, the validation of social metrics—such as user engagement, follower counts, and behavioral data—tends to be unclear and susceptible to manipulation or inaccuracies. This lack of transparency undermines trust and limits users' ability to manage their online identities and interactions.
The Azuki team collaborated with Flashbots to explore the potential of TEEs in a social context on Twitter, introducing the community to the concept and capabilities of TEEs.
TL;DR - Bobu Experiment
Bobu, a fractionalized Azuki NFT, has become highly popular within the Azuki community. As part of an innovative experiment, the team enabled a prominent community member to tweet through Bobu's Twitter account. The crucial aspect of this experiment was that TEEs guaranteed the community member could only tweet once and under specific conditions—an additional restriction not defined by the original OAuth scope.
Security Assumptions
The primary security assumption is that when a one-time use token is issued to allow tweeting on behalf of an account:
• The account owner is shown a clear authorization window outlining how the token will be used
• The account owner explicitly approves the action
To ensure this process is secure, we need to guarantee:
• Restricted Access to the X Access Token: The OAuth process for generating the Twitter access token is routed through a TEE via a callback URL, ensuring that only the TEE can access the token
• Execution Integrity: The TEE generates remote attestation quotes, signed by Intel, containing hashes of the program binary and its dependencies. These hashes can be independently verified to ensure the executable code matches the expected version
The attestation quotes include a hashed public key derived from a private key known only to the TEE. Users can verify that the domain certificate their browser connects to is associated with the public key of the TEE's keypair, ensuring the authenticity and integrity of the execution.
Phala Network
Phala Network is an advanced cloud platform designed to offer a low-cost, user-friendly trustless environment, making zero-trust computing accessible to developers across various industries. By utilizing a hybrid infrastructure that combines TEEs, Multi-Party Computation (MPC), and Zero-Knowledge Proofs (ZKPs), Phala provides flexible, open-source, and cost-effective verification solutions suitable for any developer, regardless of the application type.
As AI rapidly advances, ensuring scalable privacy solutions for LLMs and inference processes is becoming increasingly vital. With the growing complexity and widespread use of AI models, protecting sensitive data during hosting and inference while maintaining performance and scalability is essential. Privacy concerns are especially pressing as AI systems handle vast amounts of personal and proprietary information.
To ensure the integrity of AI systems, we require several key components:
• Tamper-proof data to prevent unauthorized alterations of user request/response data through secure communication and encryption
• A secure execution environment utilizing TEEs to isolate and protect both hardware and software
• Verifiable execution results ensuring that AI outputs are trustworthy and have not been manipulated
Phala Confidential AI Inference
Phala Network utilizes NVIDIA's TEE GPU technology to deliver private inference services by harnessing the Confidential Computing capabilities of NVIDIA Blackwell. This ensures that both AI model execution and data processing are fully protected within a secure environment. This solution allows organizations to run LLM workloads with robust privacy and security guarantees, safeguarding against unauthorized access to both the model and user data during inference operations.
Example: Phala x 0G
Operators on 0G's platform can choose to run their AI nodes within a TEE environment powered by Phala's SDK. When users interact with 0G's AI service, their requests are securely routed through a proxy to an LLM instance protected by a TEE. This setup guarantees that sensitive data remains confidential throughout the entire process, from the initial request to the final response.
Once the LLM generates the output, it is accompanied by a Remote Attestation (RA) report, which provides cryptographic evidence of the execution environment's integrity. This report allows users to independently verify the authenticity and accuracy of the response using standard RA verification libraries, ensuring that the results are both trustworthy and tamper-proof. By combining secure data handling with verifiable execution, 0G ensures a high level of privacy and reliability for AI interactions.
By isolating sensitive data and execution logic within secure enclaves, TEEs provide a robust shield against unauthorized access and tampering, even in potentially untrusted environments. This combination of data confidentiality, secure computation, and verifiable results makes TEEs an ideal solution for deploying LLMs in a privacy-preserving manner, offering both organizations and users a secure, transparent, and reliable AI service.
