# One sentence, fourteen steps, three chains > The health factor it has to hold, the fees at every hop, the places it can fail, and the three things that make this plan harder than it looks. - Source: https://defiloops.com/blog/a-loop-that-buys-and-borrows - Published: 2026-05-19 - Category: Strategy - Tags: worked-example, lending, bridging, rwa - Author: DeFiLoops --- Most strategy write-ups show you the idea and skip the execution. The execution is where the money goes, so this one does the opposite: the idea takes two paragraphs and the rest is what actually has to happen. ## The idea You hold $1,000 in USDC on Base. You want long cbBTC exposure, you do not want to sell it to raise cash, and you want a yield-bearing real-world asset paying down the borrow so the position services itself. In one sentence: Take $1,000, swap half to cbBTC, supply it, borrow $500, bridge it, buy the RWA, and let its yield repay the loan every month. Four protocols, three chains, and a monthly schedule. By hand this is roughly twenty minutes and eight signatures a month, every month, forever — which is the actual reason people stop running strategies like this, not the strategy being wrong. ## The compiled plan ## Reading it Three things in that board are worth stopping on. ### The dotted lines are transaction boundaries Steps 2→3 and 3→4 are dotted. That is the engine saying a runtime amount cannot be carried across that point, so a new transaction starts. It is not a warning. It is the shape of the plan, stated before you approve it rather than discovered when something reverts. ### "← linked step's output" is not a placeholder Step 3 supplies *what step 2 actually received* — the real fill after slippage, not the 500 USDC the plan hoped for. If the swap fills at a worse price, the supply is smaller and the borrow is sized against the smaller collateral. The health factor holds. This is a sharp edge. On a liquidity-provision step, `amount0` can be fed by a previous step while `amount1` is a literal. A plan that swaps into token1 and expects both legs to size themselves will come up short on one side. Check which parameters are runtime-fed before you rely on it. ### The bridges are the expensive part Steps 5, 7 and 9 cross chains. Each is off-chain — value leaves, time passes, value arrives. They are the slowest steps, the ones with real fees, and the ones a runtime amount cannot cross. The bridge in the catalogue is CCTP, and CCTP is USDC-only. Sizing a bridge step from a swap that produced WETH is refused at validation — which is the correct outcome, but it surprises people, because the sentence did not say anything about USDC. ## The numbers Note what is *not* a cost: you never bridge gas to Arbitrum or Ethereum. The relayer sponsors those calls and is reimbursed in the asset the plan already holds. For a three-chain plan that is the difference between one signature and about six. ## The same plan, at three sizes The step count does not change with size, but what is worth doing does. Bridge fees are flat, so they dominate a small plan and vanish in a large one. | Plan size | CCTP fees | Fees as % | Steps | Worth automating | | --- | --- | --- | --- | --- | | $1,000 | $1.50 | 0.15% | 14 | Marginal — fees eat a month of yield | | $10,000 | $1.50 | 0.015% | 14 | Yes | | $100,000 | $1.50 | 0.0015% | 14 | Yes, and slippage now matters more than fees | Above roughly $10,000 the flat costs stop being the thing to optimise and swap execution becomes the thing to optimise — which is a different plan shape, with the trade split across venues rather than routed through one. ## What can go wrong The supply and borrow size down with it. The health factor is preserved because the borrow is computed from the actual collateral, not the intended collateral. The health factor drops. At 1.56 there is meaningful room, but this is a leveraged position and it can be liquidated. The plan does not defend it — you would need a separate rule for that, and you should ask what protection costs before assuming it is worth buying. Step 13 repays what step 12 redeemed. A short month leaves debt outstanding, which is fine — it accrues, and the next run repays more. Nothing breaks. The plan halts there and reports the state it stopped in. Steps 11–14 simply do not run. The next scheduled run reads your balances fresh rather than assuming. Before running this, the read-only tools will price it: what the range would have collected over real history, what the borrow has cost at every rate since 2021, what liquidations have looked like at this health factor. None of that signs anything, and it is the half of the product most people underuse. ## Repaying by shares, not by amount One implementation detail that generalises well beyond this plan. Step 13 repays 500 USDC. If you instead repay *the position*, repay by shares rather than by a fixed amount — a fixed amount leaves dust debt behind as interest accrues between calculation and execution, and dust debt blocks the withdrawal you were trying to make. Interest accrues between the read and the transaction. A few wei of debt survive. The subsequent withdraw reverts and the reason is not obvious. The position closes cleanly regardless of how much interest accrued in between. The same shape shows up with aTokens: withdrawing exactly what you supplied can revert by one wei, because the token rounds down. Whole-position operations should be expressed as whole-position operations. ## Was it worth automating? Twenty minutes and eight signatures a month is not a lot of work. The reason to automate it is not the twenty minutes — it is that a strategy you run by hand is a strategy you run when you remember, at a price you did not choose, and stop running the month you are busy. The plan above runs on the first of the month whether or not you are paying attention, at a shape you approved once, and stops cleanly if anything is wrong.