AIxFI Whitepaper
Last updated
Last updated
AIxFI: AI-Powered Stablecoin Yield Optimization
This whitepaper introduces AIxFI, an advanced Artificial Intelligence (AI) agent developed by Interstable, designed to autonomously analyze, forecast, and optimize yield strategies for stablecoin portfolios within the Decentralized Finance (DeFi) ecosystem. Leveraging real-time and historical data from prominent DeFi protocols and data aggregators (e.g. DefiLlama, Exponential.fi, Artemis), AIxFI aims to maximize risk-adjusted returns while navigating the complexities of liquidity, protocol risk, cyclical market dynamics, and actual yield distributions. AIxFI's intelligence will be the core engine powering Interstable's upcoming yield-bearing stablecoin, aiUSD, and will provide transparent market insights via the Interstable platform.
1.1. The Stablecoin Yield Challenge The stablecoin market, a cornerstone of the digital asset economy, represents a significant volume of capital, projected to exceed $2 trillion by 2028 (). While stablecoins offer stability and efficient value transfer, a vast majority of these assets sit idle, earning no intrinsic yield for their holders. The DeFi and DeFAI ecosystem presents hundreds of applications offering yields and in specific yields on stablecoins; however, these opportunities are characterized by:
Fragmentation: Yield sources are scattered across numerous blockchains and protocols.
Dynamic Complexity: APYs, risk profiles, and liquidity conditions shift rapidly. Advertised APYs often diverge from actual Yield Paid Out (YPO).
Information Asymmetry: Identifying and assessing the true risk-adjusted potential of these opportunities, including understanding seasonal and cyclical impacts, requires specialized knowledge, significant time investment, and continuous monitoring.
Operational Overhead: Manual tracking and allocation adjustments are inefficient, error-prone, and often too slow to capitalize on optimal conditions.
Moreover:
DeFAI fails to anticipate user needs: Every AI and DeFAI chatbot requires users to know exactly what to ask, leaving them unaware of critical information they need but don’t know to request.
This landscape results in most stablecoin holders either forgoing potential returns or unknowingly engaging with suboptimal or overly risky strategies.
1.2. Interstable's Vision: DeFi Smart Earnings Interstable is committed to solving these challenges by providing "DeFi Smart Earnings." Our mission is to abstract the complexity of stablecoin yield generation, making it accessible, efficient, and intelligent for all users. The foundational phase of this vision is AIxFI.
1.3. Introducing AIxFI: Interstable's AI Yield Engine AIxFI is Interstable's proprietary AI Yield Engine. It intuitively understands user intentions and proactively executes the best possible actions, delivering solutions without requiring constant user input or attention. It is an autonomous agent that continuously ingests, normalizes, and analyzes DeFi protocol data to:
Identify and rank optimal stablecoin yield opportunities.
Perform sophisticated risk-adjusted yield analysis, considering actual Yield Paid Out (YPO) as a key performance indicator.
Analyze market seasonality, cyclical trends, and historical patterns to enhance strategy formulation and forecasting.
Forecast yield trends using predictive modeling.
Inform the dynamic management of Interstable’s upcoming AI-native stablecoin, aiUSD.
AIxFI compares yields across stablecoin DeFi protocols and chains to select the best real-time returns, adjusted for multifaceted risk factors, liquidity constraints, and informed by predictive insights.
AIxFI operates as a modular, layered system designed for robust data processing and intelligent decision-making. It is primarily an off-chain application with future potential for on-chain oracle components for critical data verification.
2.1. Data Ingestion Layer Responsible for fetching and pre-processing raw data from diverse sources, with all data timestamped upon ingestion.
Primary APIs: DefiLlama & Artemis (Pools, Protocols, Stablecoins data including APY, Base APY, Reward APY, TVL, audit links, etc.), Exponential.fi (Risk Scores, Factor Breakdowns), Stablewatch.io (Peg Stability, historical YPO if available).
Direct Blockchain Node Access: For supported chains (e.g., Ethereum, Base, Arbitrum, Solana, Sui) for real-time verification, transaction volumes, and granular on-chain data.
Individual Protocol APIs/Subgraphs: For specific metrics like precise utilization rates, LP fee generation, or lock-up queue details not covered by aggregators.
Frequency of data pulls varies from near real-time for volatile metrics (APYs, prices) to daily/weekly for more static data (protocol metadata, audit links).
2.2. Data Processing & Enrichment Layer This layer cleans, transforms, normalizes, and enriches the ingested data.
Data Cleaning: Handling missing values, outliers, and inconsistencies.
Normalization: Standardizing APY/APR formats to unified time-series, harmonizing volatility metrics, converting all values to USD equivalents.
Time Series Aggregation: Calculating moving averages (7D, 30D, 90D for APY and YPO), TVL changes, etc.
Data Enrichment: Linking data from multiple sources (e.g., DefiLlama pool with Exponential.fi risk score), calculating derived metrics (yield volatility, Base APY vs. Reward APY ratio, historical YPO vs. advertised APY divergence).
2.3. Analytical Engine The core intelligence of AIxFI, comprising several specialized modules:
2.3.1. Yield Analysis Module:
Differentiates Base APY from Reward APY, prioritizing sustainable real yield.
Conducts APY stability analysis (historical averages, volatility).
Models the impact of TVL on APY to anticipate yield dilution from potential capital deployment.
Focuses on "Real Yield" paid in stablecoins or reputable blue-chip assets.
Tracks and Analyzes Yield Paid Out (YPO): Monitors the actual historical yield distributed by protocols to their users over various periods (e.g., 7D, 30D, 90D, All Time). This provides a concrete measure of realized past performance, offering a crucial counterpoint to potentially misleading advertised APYs. AIxFI analyzes discrepancies between APY and YPO to gauge yield reliability.
2.3.2. Market Cycle & Seasonality Analysis Module:
Objective: To identify recurring patterns, cyclical trends, and seasonal effects within the DeFi yield market, which significantly influence APY and YPO.
Data Inputs: Long-term historical APY and YPO data, TVL flows, trading volumes, crypto market sentiment indices, correlation with broader crypto market cycles (e.g., bull/bear phases, major narratives).
Methodology: Employs time-series analysis techniques (e.g., Fourier analysis, decomposition, autocorrelation) and pattern recognition models to detect:
Seasonal spikes or dips in liquidity, yield, and YPO (e.g., end-of-quarter effects, impacts of major crypto events or airdrop farming seasons).
Longer-term yield compression or expansion cycles tied to market sentiment or capital flows.
Correlation between specific protocol yields/YPO and broader market movements.
Output: Insights into expected future yield behavior based on historical cycles, which informs the Forecasting Module and strategy risk assessment. For example, an unusually high current APY might be flagged as unsustainable if it deviates significantly from established seasonal norms or cyclical trends for that protocol type, or if its YPO history is weak.
2.3.3. Risk Assessment Module:
Utilizes a multi-factor approach incorporating:
Quantitative data: Exponential.fi risk scores, number/quality of audits, TVL, protocol age ("Lindy" effect), on-chain metrics, stablecoin peg stability (from Stablewatch.io).
Qualitative inputs: Team anonymity/reputation, mechanism complexity, centralization vectors (admin keys, oracle reliance), chain security.
Insights from the Market Cycle & Seasonality module regarding cyclical risks.
Generates an Internal Risk Score for each protocol and opportunity using a weighted algorithm, categorized into risk tiers (e.g., Low, Medium, High) or a numerical score (e.g., 0-100, higher is safer).
Considers "Leading," "Emerging," "New" qualitative tiers (assigned by the Interstable team for whitelisting) as an additional risk dimension.
2.3.4. Liquidity & Lock-up Module:
Identifies and parses lock-up terms, unbonding periods, and withdrawal queue mechanisms.
Quantifies potential yield premiums offered for illiquidity, comparing them against comparable liquid strategies.
Informs portfolio construction to meet target liquidity profiles for aiUSD, balancing yield from locked assets with the need for overall portfolio liquidity.
2.4. Strategy Identification & Scoring Module (The Optimization Core) This module integrates outputs from the Analytical Engine to identify and rank optimal yield strategies.
2.4.1. Mathematical Framework: Risk-Adjusted Yield Index (RAYI)
AIxFI employs a formal optimization model based on the Risk-Adjusted Yield Index (RAYI). RAYI is designed to quantify the yield generated by a portfolio relative to its consolidated risk exposure, adapted for DeFi's unique characteristics.
Let the following variables be defined:
The variable
(or r(i,t) in plain text) represents the expected sustainable APY of protocol 'i' at a given time 't'. This value is derived by AIxFI from the current Base APY, critically adjusted by AIxFI's forecast which considers historical APY stability, Yield Paid Out (YPO) trends (as higher historical YPO can indicate more reliable and actually realized yield generation), and anticipated cyclical/seasonal impacts.
The variable
(or sigma(i) in plain text) represents the volatility (as a proxy for risk) of protocol 'i', potentially adjusted for cyclical risk factors and historical YPO consistency.
The variable
(or w(i) in plain text) represents the portfolio weight allocated to protocol 'i'.
The variable 'A' represents the total capital under management.
The variable
or L(i) in plain text) represents the observable liquidity or TVL cap for protocol 'i' relevant to the deployment size from 'A'.
The RAYI Objective Function to be maximized is described as follows: RAYI at time t equals the sum over all protocols 'i' of (the weight of protocol 'i' multiplied by the expected APY of protocol 'i' at time 't'), this entire sum then divided by the square root of the sum over all protocols 'i' of (the square of the weight of protocol 'i' multiplied by the square of the volatility of protocol 'i').
Mathematically, the Risk-Adjusted Yield Index (RAYI) at time is defined as:
This objective is subject to the following constraints: First, the sum over all protocols 'i' of the weight of protocol 'i' must equal 1, ensuring full allocation of the portfolio. Second, each weight of protocol 'i' must be greater than or equal to 0, meaning no short selling of positions. Third, the capital allocated to any protocol 'i' (which is the weight of protocol 'i' multiplied by the total capital 'A') must be less than or equal to L(i), the available liquidity or capacity constraint for that protocol.
These constraints can be expressed mathematically as:
AIxFI utilizes quadratic programming techniques to solve this optimization problem, finding the weights that maximize RAYI under the given constraints.
2.4.2. Filtering & Selection Logic:
Initial Filtering: Excludes opportunities not meeting baseline criteria (e.g., insufficient audit history, unacceptably high external risk scores, very low TVL or historical YPO, non-stablecoin focus).
"Safest Highest Yields" Prioritization: Within acceptable risk tiers, AIxFI prioritizes strategies offering the highest RAYI.
Diversification Rules: Enforces maximum allocation percentages per protocol, per chain, and per type of lock-up duration to maintain a balanced portfolio for aiUSD.
2.5. Forecasting Module (Leveraging Transformer Architecture) To enhance strategic allocation and anticipate market shifts, AIxFI incorporates a forecasting module for APY and potentially YPO trends.
Model: A multi-layer Transformer neural network, chosen for its efficacy in capturing complex temporal patterns in financial time series. The model processes historical windows of APY and YPO inputs (represented as x1, x2, through xt), complemented by features derived from market cycle analysis, seasonality indicators, and historical Yield Paid Out (YPO) data, to produce multi-step forecasts (represented as y-hat at t+1, through y-hat at t+k) for 30 and 90-day horizons.
Advantages: Self-attention mechanisms allow Transformers to weigh relevant past observations, adapting to cyclical yields, the impact of sustained high or low YPO on future APY sustainability, sudden rate shifts, and seasonal liquidity flows more effectively than traditional models like ARIMA or LSTMs in the dynamic DeFi environment.
Output: Forecasted APYs with confidence intervals, feeding into the risk assessment and strategy scoring modules.
2.6. Backtesting Module AIxFI includes a backtesting module to simulate the historical performance of identified strategies and the RAYI optimization model. This module replicates market conditions over defined historical periods (e.g., past 12–24 months) across selected stablecoin yield protocols, allowing evaluation against baseline strategies (e.g., static allocations).
Simulated Historical Results (Illustrative Example):
Protocols Tested: Aave, Spark, Curve, Pendle, Ethena (sUSDe).
Period: 2023-2024.
Avg. APY & Simulated Avg. YPO of AIxFI-Optimized Portfolio: For example, 9.5% – 11.2% APY (net of simulated transaction costs), with a corresponding strong YPO figure indicating a high percentage of this APY was realized as actual distributed yield.
RAYI Optimized Portfolio outperformed static 60/40 (Aave/Curve) benchmark by, for example, 2.5% annualized in terms of net realized returns. (These results are illustrative and will be derived from actual backtests.)
2.7. Output & API Layer Exposes AIxFI's findings through secure, internal APIs for:
The Interstable platform frontend (powering the "AIxFI" insights page – note: AIxFI is the public-facing name for the AIxFI engine's showcase).
The aiUSD Management System (providing rebalancing suggestions for MVP, evolving to automated execution triggers).
The internal Monitoring & Alerting Module.
Data Exposed: Processed opportunity lists, risk scores, illustrative optimal portfolios, forecasts, YPO analysis, backtest summaries.
2.8. Monitoring & Alerting Module Tracks critical metrics in real-time (sudden APY drops, significant TVL drains, YPO deviations from APY, risk score changes, de-peg events, security alerts from external feeds and social channels) and notifies the Interstable team for review or intervention.
aiUSD (details in a separate forthcoming whitepaper/section) will be Interstable’s decentralized, yield-bearing stablecoin, designed to be 1:1 pegged to the US dollar. Its backing collateral (composed of other reputable stablecoins like USDC, USDT) will be actively managed by AIxFI, routing these assets into the optimized, diversified portfolio of yield-bearing strategies identified by the AI.
Yield Accrual: Yield generated by AIxFI-managed strategies, reflected in strong Yield Paid Out (YPO) from underlying protocols, will accrue to the aiUSD backing collateral. This causes aiUSD's redemption value to increase over time in a non-rebasing manner.
Transparency: All backing assets and their deployment will be verifiable on-chain and displayed on Interstable's Transparency Dashboard.
4.1. Phase 1 (Current): AIxFI Engine Showcase (via "AIxFI" on Interstable website)
Virtual Genesis Launchpad and Initial Angel round for Development and Marketing Expenses.
Community building and education. Waitlist for aiUSD.
Launch of Interstable informational website with the "AIxFI" page showcasing AIxFI's insights.
4.2. Phase 2: aiUSD Stablecoin Launch
Full audit and mainnet deployment of aiUSD smart contracts.
AIxFI actively informs aiUSD portfolio management (initially AI-assisted, human-approved rebalancing).
4.3. Phase 3: Enhanced Automation & Ecosystem Integration
Gradual rollout of AI-automated rebalancing for aiUSD strategies, driven by AIxFI.
Development of "Advanced Mode" allowing users to leverage AIxFI's insights for personalized portfolio construction.
API integrations for wallets and platforms to create a "universal earnings switch."
AIxFI represents a significant step forward in applying Artificial Intelligence to optimize stablecoin yield generation in DeFi. By combining comprehensive data analysis, a robust risk framework that scrutinizes Yield Paid Out (YPO), advanced forecasting sensitive to market cycles and seasonality, and a principled optimization model (RAYI), AIxFI is engineered to navigate the complexities of the DeFi landscape effectively. It will serve as the intelligent foundation for Interstable's aiUSD stablecoin, aiming to deliver "DeFi Smart Earnings" to a broad user base with simplicity and transparency.
Appendix A: References
Vaswani, A., et al. (2017). Attention Is All You Need. Proceedings of the NeurIPS Conference.
Relevant whitepapers of protocols AIxFI analyzes (e.g.,Ethena, Sku).
Academic papers on financial time series forecasting, portfolio optimization, and DeFi risk analysis.
DeFiLlama Documentation & API:
Exponential.fi Risk Methodology:
