# When the Stop-Loss Can’t Exit: What My AI Vault Is Teaching Me About Real DeFi

> Krystal AI Agent Auto Vault Update Episode 4

**Published by:** [Dadsdefispace.base.eth](https://paragraph.com/@daddefispace/)
**Published on:** 2026-03-05
**Categories:** krystal, vault, a.i., agents, defi, crypto
**URL:** https://paragraph.com/@daddefispace/when-the-stop-loss-cant-exit-what-my-ai-vault-is-teaching-me-about-real-defi

## Content

DADS DeFi Space · AI Vault Experiment UpdateCrypto automation sounds simple in theory. Write a strategy. Set your risk rules. Let the AI execute. But once those rules collide with real liquidity, real slippage, and real on-chain execution, things get a lot more interesting. This latest update from the DADS DeFi Space AI Vault experiment shows exactly that. The strategy is working. The AI agent is doing its job. But the biggest lesson right now isn’t about APR. It’s about execution.The Current SnapshotLet’s start with where the vault stands right now. At the moment of this update: Vault OverviewTotal Value Locked: ~$873Net PnL: –$162 (–15.67%)7-Day APR: ~113%Closed Strategies: 69Despite the negative vault PnL, the current active positions are actually profitable. Open Positions WETH / USDC~$397 TVL+$10.77 PnL~68% APRWETH / VVV~$147 TVL+$17.61 PnL~593% APRUSDC / cbBTC~$3 position (dust)+$0.03 PnLAll three positions were in-range and generating yield at the time of the snapshot. So the obvious question becomes: If the current positions are working… why is the vault still down?The Real Story: Churn and Execution FailuresThe answer isn’t the current strategy. It’s the legacy churn that happened earlier. The vault has already closed 69 strategies. That’s the signal. Every open, close, rebalance, and forced exit introduces:swap feesmint/burn LP costsslippageimperfect exit timingEven when the current positions look healthy, the vault PnL reflects the history of those decisions. In DeFi, the damage often comes from turnover, not just bad positions.Where Automation Meets RealityOne of the biggest lessons in this experiment is something many DeFi dashboards hide. Rules do not guarantee execution. The AI agent has repeatedly tried to follow the strategy rules correctly. But in several cases the execution failed. Examples include:No Quote AvailableMinimumAmountInsufficientSlippage failuresPool data sync errorsIn other words: The AI decided to exit… but the transaction could not complete. That gap between decision and execution is where real-world automation gets messy.The Stop-Loss ProblemOne of the vault rules was simple: Exit positions when PnL drops below –15%. In theory, this protects capital. But in practice, several exits failed because liquidity in certain pools was too thin. One example involved a withdrawal attempt reverting with MinimumAmountInsufficient. Another involved an attempted withdrawal failing due to slippage constraints. When that happens, the stop-loss rule still exists… But it becomes a suggestion instead of a guarantee.The Hidden Risk: Illiquid PoolsThis problem showed up most clearly with some of the degen liquidity pools. Pools like:CASHKELLYCLAUDEThese pools initially looked attractive because of high APR. But they introduced a new risk that many strategies underestimate: exit feasibility. If a position cannot exit cleanly due to routing failures or slippage, the strategy loses flexibility. That’s why some of the vault’s early churn came from positions that were technically profitable on paper but difficult to unwind.When Manual Intervention Becomes NecessaryEventually, I had to manually step in and unwind one of the problematic positions. The CASH pool was creating repeated execution failures. The AI was attempting to exit according to the rules, but the transaction path kept failing. So I manually withdrew the position. That likely realized some losses. But it removed the execution trap, which is sometimes the better long-term decision. This moment highlights something important about automation in DeFi. Automation reduces labor. It doesn’t remove responsibility.What Changed in the StrategyThis update marks a shift toward a more execution-aware vault strategy. Several improvements have already been implemented. Key adjustmentsMinimum TVL requirement raised to $100kMinimum range width increased to 10%Partial exit fallback logic if slippage prevents full withdrawalNo-quote fallback redeploys capital into CORE pools instead of retryingMinimum position size added to prevent tiny dust strategiesManual removal of execution-trap poolsThese adjustments are designed to reduce churn and increase the probability that exits can actually clear.What’s Working NowDespite the earlier turbulence, several things are encouraging. Positive signsThe AI agent is consistently maintaining positions.Current strategies are profitable and in range.Wider ranges and higher TVL filters are improving stability.The vault is rotating capital into stronger pools.In other words: The strategy rules are working. The next step is improving execution reliability.Lessons So FarIf you take one lesson from this experiment, it should be this: Automation doesn’t eliminate risk. It changes its shape. The AI can follow rules perfectly, but DeFi doesn’t guarantee execution. When liquidity is thin or routing breaks, your stop-loss becomes a suggestion. The real edge in automated DeFi strategies isn’t just picking pools. It’s designing rules that survive:slippagemissing quotesliquidity droughtsfailed exitsWhat Comes NextThe next iteration of this vault experiment will focus on reducing unnecessary complexity. Planned adjustments include:Increasing the minimum position size to ~$50Monitoring weekly churn as a primary KPIBlacklisting pools that repeatedly cause execution failuresTesting whether two positions (CORE + RISK) outperform three at this vault sizeAt smaller capital levels, simplicity often wins.Final ThoughtOne thing this experiment continues to prove is that DeFi automation is not magic. It’s infrastructure. The AI can follow the plan. But the market still decides what actually executes. And sometimes the most valuable lesson in DeFi is not the APR you see on the dashboard… It’s the trade that couldn’t exit. — Kevin DADS DeFi SpaceEducational content only. Not financial advice. DeFi strategies involve risk including impermanent loss, smart contract risk, and liquidity constraints. Want to Learn More?If you’re interested in learning more about DeFi strategies and following these experiments in real time, you can explore the resources below. 💬 Free Telegram Community https://t.me/DADSDefiSpace 🎓 Free DeFi Course https://www.dadsdefispace.org/challenges 👨‍🏫 Follow on X https://x.com/cryptozone1013 💬 Farcaster https://farcaster.xyz/thecaptain1013 📲 Base App Profile https://base.app/profile/dadsdefispace 🚀 Join the community on Base https://base.app/invite/dadsdefispace/62YVZ0B3 📰 Web3 Newsletter https://paragraph.com/@daddefispaceOn-Chain Experiments🪙 Community Coin: $DADSDEFISPACE Contract: 0x11c77e7a39c80e00f1c15bfb5f394e7b7e9a50c6 🌐 Creator Coin https://zora.co/dadsdefispace 🧱 Launch your own token on Zora https://zora.co/invite/dadsdefispaceFinal ReminderEverything shared here is educational only. DeFi carries real risks including smart contract risk, volatility, and liquidity loss. Always do your own research and manage risk accordingly. Until next time. Go outside. Touch some grass. — Kevin DADS DeFi Space

## Publication Information

- [Dadsdefispace.base.eth](https://paragraph.com/@daddefispace/): Publication homepage
- [All Posts](https://paragraph.com/@daddefispace/): More posts from this publication
- [RSS Feed](https://api.paragraph.com/blogs/rss/@daddefispace): Subscribe to updates
- [Twitter](https://twitter.com/Cryptozone1013): Follow on Twitter
- [Farcaster](https://farcaster.xyz/dadsdefispace.base.eth): Follow on Farcaster