Open-source research knowledge base for delta-neutral crypto yield farming using Jupiter's JLP on Solana.
This is not a trading bot, a vault, or a dApp. It's a structured, sourced, and opinionated research project that asks: is delta-hedged JLP yield farming actually worth doing after you account for all the risks?
Long JLP (Jupiter Liquidity Provider token on Solana)
+ Short BTC/ETH/SOL perpetual futures at matching basket weights (2x leverage)
= Near-zero directional exposure
+ Two yield streams: JLP LP fees + perp funding rate income
JLP is a basket of SOL (~40%), BTC (~12%), ETH (~8%), and stablecoins (~35%). You hold JLP to earn trading fees from Jupiter's perp exchange, then short the volatile assets to neutralize price exposure. What remains is yield from two sources: traders losing money (flowing to JLP holders) and leveraged longs paying funding to shorts.
The claim: ~50% APY "when conditions were good" (from the community member who introduced the strategy).
The reality after research: 10-15% net APY in normal conditions, 20-40% in bull markets, 0-8% in bear markets. The 50% figure is a bull-market upper tail, not a baseline.
Every claim has a source tag. Every number has provenance. Every page has a "Known Unknowns" section that surfaces what we don't know.
| Section | Pages | What You'll Learn |
|---|---|---|
| Strategy | 4 | How the hedge works, P&L by scenario, yield sources, rebalancing mechanics |
| Concepts | 5 | Delta hedging, perps, funding rates, basis trades — the vocabulary |
| Platforms | 6 | JLP deep-dive + 5 venue evaluations (Drift, dYdX, Hyperliquid, Binance, Jupiter Perps) |
| Risks | 7 | 13-risk register, platform risk case studies, Solana outages, funding flips, oracle risk, correlation breakdown, failure mode taxonomy |
| Research | 5 | Historical funding rates, JLP drawdown history, yield model (3 scenarios), alternatives benchmark, MCP data capabilities |
- Yield Model (
wiki/research/yield-model.md): Conservative / Base / Optimistic scenarios with component breakdown - Risk Register (
wiki/risks/risk-register.md): 13 risks cataloged by severity, with mitigation status and gate assessment - Venue Comparison: 5 venues evaluated on fees, funding, security, and cross-chain friction
- Alternatives Benchmark (
wiki/research/alternatives.md): JLP hedge vs Ethena sUSDe, GMX GLP, pure funding arb, stablecoin lending - Failure Modes (
wiki/risks/failure-modes.md): How delta-neutral strategies actually fail (FTX collapse, Ethena depegs, May 2021 crash)
| Scenario | Net APY | When |
|---|---|---|
| Conservative | 0-8% | Bear market, negative funding |
| Base case | 10-15% | Normal ranging market (2026 conditions) |
| Optimistic | 20-40% | Bull market, high leverage, positive funding |
| NoLimit's historical | ~50% | Rare upper tail, peak conditions only |
| Venue | Verdict | Why |
|---|---|---|
| dYdX v4 | Best for security | Longest clean track record, regulatory-conscious |
| Hyperliquid | Best for yield | Higher funding volatility, good liquidity |
| Jupiter Perps | Simplest operations | Same chain as JLP, but maximum concentration risk |
| Binance | Cheapest fees | Deepest liquidity, but custodial + regulatory risk |
| Drift | ELIMINATED | $285-286M exploit April 2026 (DPRK-linked) |
The single most important structural finding: SOL is the critical risk component across every dimension.
- Largest JLP basket weight (34-45%)
- Highest volatility (68% vs ETH's 45%)
- Highest beta (2.13x relative to ETH)
- Most volatile funding rates (hit -8.2% APR during stress)
- Largest JLP drawdowns are SOL-driven (30% in March 2025 when SOL fell from $295 to <$100)
If you get the SOL hedge wrong, nothing else matters.
"Delta neutral" does NOT mean "risk free." It means directional price exposure is approximately zero. The strategy can still lose money from:
- Platform exploits — Drift lost $286M in April 2026. If your short venue gets drained, delta-neutral math is irrelevant.
- Funding regime shifts — When longs aren't crowded, shorts pay funding instead of collecting it.
- Cross-venue failure — If one leg freezes while the other is active, the structure breaks.
- Oracle manipulation — Though BTC/ETH/SOL are too liquid to manipulate easily.
- Hedge drift — JLP weights change daily; if you rebalance weekly, you accumulate residual delta.
The Perplexity research calls this strategy "high-touch, hedge-fund-like infrastructure, not passive DeFi income." That framing is correct.
DeltaYield/
├── raw/ # Source material (immutable after save)
│ ├── perplexity/ # Deep research exports
│ ├── articles/ # Published articles
│ ├── assembly-origin/ # Origin docs from community
│ └── videos/ # Video transcript notes
├── wiki/ # Structured knowledge base (start here)
│ ├── strategy/ # How the strategy works
│ ├── concepts/ # Foundational vocabulary
│ ├── platforms/ # Venue evaluations
│ ├── risks/ # Risk analysis (7 pages, 13 risks)
│ ├── research/ # Deep-dive research pages
│ ├── index.md # Master catalog (start here)
│ └── log.md # Audit trail of all research
├── data/ # Quantitative data (empty in RESEARCH phase)
│ ├── funding-rates/ # Historical funding rate snapshots
│ ├── jlp-composition/ # JLP basket weight snapshots
│ └── positions/ # Position state (gitignored)
├── output/ # Generated reports
├── scripts/ # Utility scripts (built in PAPER phase)
├── .claude/agents/ # AI agent definition for analysis
├── CLAUDE.md # Project operating instructions
└── README.md # This file
The wiki is plain Markdown. You can read everything directly on GitHub or clone and browse locally:
git clone https://github.com/PizzadatCode/DeltaYield.git
cd DeltaYieldStart with wiki/index.md — it's the master catalog that links to all 25 research pages.
The project includes a CLAUDE.md system prompt and a specialized yield-analyst agent that can answer questions about the strategy, run hedge calculations, and help you extend the research.
# Clone the repo
git clone https://github.com/PizzadatCode/DeltaYield.git
cd DeltaYield
# Open with Claude Code
claude
# The CLAUDE.md auto-loads — Claude now has full context about the strategy.
# Try asking:
# "What's the current risk picture?"
# "Explain the yield model scenarios"
# "Compare dYdX vs Hyperliquid for the short leg"
# "What are the remaining gaps before PAPER phase?"The yield-analyst agent activates on keywords like "check my hedge", "funding rates", "JLP", "delta neutral", "rebalance", "yield", or "venue comparison".
The project works with any AI coding assistant that reads project-level instructions:
- Claude Code: Reads
CLAUDE.mdautomatically - Cursor / Windsurf: Point it at
CLAUDE.mdas the system prompt - Other AI CLIs: Feed
CLAUDE.mdas context, then ask questions about the wiki
For live crypto prices and funding rate data, you can configure MCP servers. Copy the example config and add your API keys:
# From the parent workspace (if using the full assembly-agent setup)
cp .mcp.json.example .mcp.json
# Edit .mcp.json and add your CoinGecko API key (free at coingecko.com)Available MCP integrations:
| Server | What It Provides | API Key? |
|---|---|---|
| Yahoo Finance | JLP, BTC, ETH, SOL spot prices (daily OHLCV) | No |
| CoinGecko | Funding rates across venues, derivatives data, historical charts | Free Demo key |
| Alpha Vantage | Economic indicators (Fed Funds, CPI, yields) | Free key (crypto requires paid) |
See wiki/research/mcp-data-capabilities.md for the full capability map.
- Start with
wiki/index.md— the master catalog - Read
wiki/strategy/overview.mdfor the structure - Read
wiki/research/yield-model.mdfor realistic yield expectations - Read
wiki/risks/risk-register.mdfor the full risk picture - Read
wiki/research/alternatives.mdto see if simpler strategies are better for you
- Read all 7 risk pages before writing any code
- Read the venue evaluations to pick your short venue
- Read
wiki/strategy/rebalancing.mdfor operational requirements - Read
wiki/research/mcp-data-capabilities.mdfor live data sources
The wiki follows strict provenance rules:
- FACT: Verifiable from primary source — requires
[src: ...]tag - CLAIM: Stated by a person — requires
[src: raw/...]tag - INFERENCE: Agent reasoning — must be tagged and never promoted to FACT
Every number must have a source or be tagged [derived: formula] or [number not stated in source]. This discipline is what makes the research trustworthy.
| Phase | Status | Purpose | Gate |
|---|---|---|---|
| RESEARCH | ACTIVE | Build knowledge base, complete risk due diligence | Sections D+E of investigation TODO |
| PAPER | Not started | Paper-trade 30-60 days, validate yield model | All research gates clear + explicit approval |
| LIVE | Not started | Real capital with tight monitoring | Paper period validates assumptions |
Current gate status: Investigation TODO Sections A-E are complete (25 wiki pages). Gate is conditionally clear pending venue selection and operational readiness.
This project was born from an educational post by a community member describing a delta-hedged JLP strategy they used to run, claiming ~50% annualized returns "when conditions were good." Rather than taking that claim at face value, this project was commissioned to do proper due diligence — researching the mechanics, evaluating the risks, benchmarking alternatives, and building a knowledge base that could inform a real deployment decision.
The conclusion after 75+ sources of research: the strategy is legitimate and can generate attractive yields, but the 50% figure is a bull-market upper tail, the operational requirements are significant, and the tail risks (platform exploits, funding regime shifts) are real. Whether it's worth pursuing depends on your risk tolerance, operational capacity, and the current funding rate regime.
- Research: Perplexity deep research, web sources, official protocol documentation
- Knowledge base: Markdown wiki with provenance-tracked claims
- AI agent: Claude Code with a specialized yield-analyst agent
- Data: Yahoo Finance MCP, CoinGecko API (for live funding rates)
This is research material, not financial advice. The authors are not financial advisors. This project documents a learning and due-diligence process for a specific crypto yield strategy. Nothing here constitutes a recommendation to enter any position. Crypto yield farming involves significant risks including total loss of capital. Do your own research.
The strategy discussed involves perpetual futures, leveraged positions, and DeFi protocols — all of which carry smart contract risk, counterparty risk, regulatory risk, and market risk. The April 2026 Drift Protocol exploit ($285-286M stolen) is a live example of how delta-neutral positions can suffer catastrophic losses from platform failure.
MIT — use the research however you'd like. Attribution appreciated.