Decentralised News · 2027 planning series · Reviewed 21 November 2026
What Matters
An AI agent can search yield data and calculate a possible switching advantage. It cannot establish the best future return by sorting today’s APY column. A useful comparison needs the same asset exposure, a holding period, the costs of leaving and entering, realistic settlement time and an exit route. A higher displayed rate can produce less money at the end of the period.
In DN’s invented example, moving $10,000 from a 4% annual effective yield assumption to 8%, with $30 of entry costs, $20 of exit costs and two days without yield during the move, adds only $39.54 over 90 days. With $1,000, the same move loses $41.55 relative to staying. These are calculations from stated assumptions, not quoted opportunities or measured returns.
The Signal
Aave documents variable supply rates and withdrawal constraints tied to liquidity and collateral. An observed APY therefore leaves the switching decision incomplete. DN compares modeled terminal value after costs and transition time, while keeping execution evidence separate.
DN Alpha Thesis: the agent should optimize a decision record
The valuable output is not “Pool B has the highest APY.” It is a dated comparison that another person can reconstruct: the pools considered, which were excluded, the rate definitions, the amount that can actually move, the costs, the scenario horizon and the conditions that would invalidate the result.
DN’s original framework separates three tasks. Discovery finds candidates. Underwriting decides whether they belong in the permitted universe. Execution checks whether an approved move still meets its conditions. Combining them into a single language-model instruction makes it easy for a fresh-looking percentage to override a stale withdrawal assumption.
Evidence boundary: primary documentation was reviewed on 21 November 2026. No live yields were ranked, no wallet was connected and no deposits, withdrawals or bridge transactions were executed. Every numerical example is synthetic. The calculator models constant annual effective yield assumptions for comparable exposure; it is not a trading signal or protocol-risk model.
DN Net Yield Switching Calculator
Change the illustrative assumptions below. This worksheet runs locally in your browser. It records your declarations; it does not verify evidence, connect to accounts, submit instructions or grant approval.
The initial numbers are synthetic. Unknown evidence remains unresolved. Downloads contain the values you enter; use aliases and avoid private or identifying information.
1. Decide what “best” means before collecting rates
A reader seeking spendable US-dollar value in three months has a different objective from one accumulating ETH over several years. Comparing a stablecoin lending rate, an ETH staking rate and a volatile two-token liquidity position as though they were the same product can hide the largest source of risk: the assets themselves.
Write the mandate in five parts: the reporting unit, permitted assets, maximum holding period, liquidity deadline and excluded exposures. Add constraints for chain, protocol, custody, leverage and concentration where they matter. A search result outside that mandate is an excluded candidate, even if its displayed rate is attractive.
| Decision | Useful comparison | Invalid shortcut |
|---|---|---|
| Keep a near-term stablecoin reserve | Comparable claims and liquidity, measured in the same reporting unit | Choosing a volatile reward token because its headline rate is higher |
| Earn on existing ETH exposure | ETH-denominated return with staking and exit conditions | Treating a dollar price gain as staking yield |
| Provide market liquidity | Fees, inventory changes, range behavior and alternative holdings | Comparing fee APY alone with lending yield |
| Move between chains | Net terminal value after both routes and transition time | Ignoring the cost and claim represented by the destination token |
The distinction is practical, not semantic. Uniswap explains that relative token-price changes can create impermanent loss for liquidity providers; fee income alone is not a complete outcome. An LP position therefore needs its own inventory and price scenarios before joining a simple lending-yield comparison. Uniswap’s impermanent-loss explanation.
2. Preserve what the yield field actually means
DefiLlama’s yield-adapter specification separates base yield and tradable-token reward yield when that breakdown is available. If it is unavailable, a total APY can be supplied without pretending the components are known. Its methodology excludes locked rewards and nontradable pre-token-generation incentives from reward APY. Those conventions are useful, but an agent must still identify the actual field it received. DefiLlama yield-server methodology.
Missing reward data is not zero rewards. A null breakdown can mean that only a total estimate was provided. Likewise, a field called APY should not be substituted for APR without checking how compounding was assumed. If the source annualizes a recent interval, preserve the interval as part of the observation.
For each candidate, retain the source URL or endpoint, pool identifier, chain, contract addresses, underlying assets, observation time and measurement definition. Store both the original response and the normalized record. The normalized version helps comparison; the original helps explain whether a later discrepancy came from the provider, the parser or a changed pool.
A model can explain these records, but the transformations should be explicit. A field-mapping rule is easier to inspect than a sentence saying the agent “understood the yield.” If an API changes the meaning of a field, the comparison should pause until the mapping is reviewed.
3. Fresh data can still be unsuitable for execution
DefiLlama currently describes yield data as updating hourly, and notes that the API can update before cached web pages. This is a discovery cadence, not a promise that a displayed rate remains available when a transaction settles. DefiLlama update-frequency FAQ.
At Aave, supply rates respond to utilization and governance parameters, and change as liquidity is supplied, borrowed, repaid or withdrawn. A current rate is therefore not a fixed contract for the next 90 days. Aave supply documentation.
DN proposes three timestamps for an automated review: when the source observed the underlying state, when the agent retrieved it, and when the decision is evaluated. If the first is missing, preserve that uncertainty. A fresh download of a stale observation should not become a fresh market fact.
The freshness limit belongs to the workflow. A daily research shortlist can tolerate different data from a transaction approval. Define the maximum observation age, the quote expiry and the required block/state references separately. The agent should refresh or escalate when those conditions fail, rather than treating its previous conclusion as permanent permission.
4. Count the complete route once
A switch has a current-position exit, a transfer or conversion, entry into the target position and a later exit from the target. Some routes combine steps in one quoted output; others charge separately. The cost ledger should show which costs are already embedded so the calculator neither omits nor double-counts them.
| Cost or constraint | Where to record it | Question for the agent |
|---|---|---|
| Current withdrawal and approval costs | Entry cost of the switching path | Can the intended amount leave now? |
| Swap/bridge charges and price impact | Entry cost or net destination amount | Does the quote already include them? |
| Destination deposit cost | Entry cost | Is the required approval included? |
| Reward claim and sale cost | Explicit scenario cost if relevant | Can rewards become the reporting asset? |
| Target-position withdrawal cost | Exit cost of the switching path | What assumptions support that estimate? |
| Exit cost if the reader stays | Baseline exit cost | Are both paths compared at the same endpoint? |
| Time between earning positions | Transition days | When does the destination position actually start accruing? |
Not every uncertainty belongs in dollars. A bridge adds implementation and operational dependencies, which a transaction-fee estimate cannot summarize. Ethereum’s bridge overview describes smart-contract, technical and, for trusted designs, custody and censorship risks. Keep these as explicit eligibility and risk questions. Do not invent a small “risk premium” and pretend it measures the chance of losing principal. Ethereum bridge-risk overview.
5. Calculate terminal value, not just the rate spread
The DN scenario model uses one reporting unit and two constant annual effective yield assumptions. Let P be the starting amount, a the current APY assumption, b the target APY assumption, d the horizon in days, and t the days without yield during the move. Rates are decimals in the equations.
Stay value = P × (1 + a)d/365 − baseline exit cost.
Switch value = (P − entry cost) × (1 + b)max(d − t, 0)/365 − target exit cost.
Switching advantage = switch value − stay value.
Entry cost is deducted before target earnings because that amount is no longer invested. The baseline keeps earning during the same horizon, while the switching path earns nothing during the entered transition period. This is deliberately simple. If the actual transition earns a different rate, or costs are paid from another account, the economic record needs an adjusted model.
The exponent converts an annual effective return assumption to the selected fraction of a year. It does not claim that a lending protocol literally compounds daily or guarantees that path. It is an equivalent-rate scenario convention. A source’s actual accrual mechanics, changing rates and realized reward prices can produce a different result.
6. Worked example: a higher APY with a small cash advantage
Assume a comparable asset exposure, $10,000 of capital, 4% current APY, 8% target APY, a 90-day horizon, two transition days, $30 entry cost, $20 target exit cost and no baseline exit charge. The model produces a stay value of $10,097.18 and a switch value of $10,136.72. The incremental advantage is $39.54.
All rates and costs are invented. They are not representative estimates for a named chain or provider. Their purpose is to show that doubling a rate does not double the reader’s short-period profit.
| Scenario | Modeled switching advantage | What it illustrates |
|---|---|---|
| $10,000, 90 days | +$39.54 | A positive but limited cushion under the assumptions |
| $10,000, 30 days | −$23.25 | The same rate gap does not recover costs quickly enough |
| $1,000, 90 days | −$41.55 | Fixed route costs dominate a smaller allocation |
| $10,000, unchanged assumptions over time | Break-even at approximately 52.4 days | A scenario threshold, not a promised payback date |
| $1,000, unchanged assumptions over time | Break-even at approximately 480.6 days | A long persistence assumption that may be implausible |
The longer break-even period is a reason to challenge the premise. A rate advantage that only pays for itself after more than a year would need to survive changes in utilization, incentives, fees and the reader’s own cash needs. The calculator has not established that it will.
7. Do not turn a withdrawal preview into a liquidity promise
ERC-4626 separates asset/share conversions and previews from account-specific limits. A favorable preview is not the same as an immediately available withdrawal. ERC-7540 adds asynchronous request states, including pending and claimable, which should not be collapsed into “cash available.” ERC-4626; ERC-7540.
Aave’s withdrawal guidance also makes availability conditional on pool liquidity and, where supplied assets secure borrowing, the borrower’s collateral position. An agent that compares rates before checking these constraints may propose moving capital that cannot presently move. Aave withdrawal documentation.
Separate “eligible to consider” from “executable now.” The former can be based on the mandate and documentation. The latter needs current position and transaction evidence. A reader can retain a research candidate while rejecting the specific proposed move.
For a deeper exit record, use DN’s six vault-withdrawal checks. That article establishes what exit observations mean. This comparison adds the economics of choosing between two otherwise eligible paths.
8. Give the agent a permission budget
A read-only yield monitor can operate without transfer authority. It can gather data, validate identifiers, calculate scenarios and notify a reviewer. If execution is later introduced, use a separate mandate naming permitted assets, contracts, amounts, maximum costs, minimum scenario advantage and expiry.
DN recommends treating a new chain, bridge, reward asset or contract address as a scope change. It should not become permitted merely because it improves a ranking. A risk limit that exists only in a prompt can be ignored by downstream software; the actual signing and execution controls need their own review.
Build a no-action band around small advantages. This is a policy choice, not a universal optimal threshold. A team could require enough modeled surplus to absorb a chosen cost overrun and a less favorable rate scenario. Record the buffer and its rationale, so changing it does not silently change the mandate.
The operating state should also prevent repeated moves while a prior route is unresolved. “Withdrawal requested,” “assets received,” “target deposited” and “position reconciled” describe different observations. Re-running the same instruction after a timeout can create additional costs or unintended exposure.
9. Stress the assumptions before acting
Change one input at a time to see what drives the result. Cut the target yield assumption, shorten the horizon, increase route costs and lengthen the transition. Then combine the adverse changes. The objective is to identify which assumptions must hold for the move to make sense.
Do not rename an arbitrary haircut “risk-adjusted APY.” A lower assumed rate is a sensitivity case. It does not price smart-contract failure, issuer default, governance action or an impaired bridge asset. Those risks can make two positions unsuitable for comparison even when the arithmetic favors switching.
A useful review sentence is specific: “This move has a modeled advantage of $39.54 over 90 days if the entered rates persist and the route completes within two days; liquidity and contract risks remain outside the model.” That sentence is less dramatic than “best yield found,” but it tells the operator what could overturn the decision.
10. DN Net Yield Switching Calculator
The instrument accepts capital, the two effective APY assumptions, horizon, transition days and the costs of both paths. It returns each terminal value, the incremental advantage and the implied percentage of starting capital. A separate checklist records whether exposure is comparable and exit evidence has been reviewed.
Blank costs remain unknown until entered or explicitly confirmed as zero. An unknown exit condition does not disappear because the numerical result is positive. The calculator can show the arithmetic while labeling the proposed move unresolved.
Use the export as a dated scenario record. Preserve the sources behind the rates and costs alongside it. Recalculate when the actual destination amount, timing or fees differ, and compare the realized record with the original assumptions after the position is closed.
For platform architecture, see DN’s DeFi-agent comparison. For cross-chain route accounting, see the route fragmentation and settlement ledger. Neither a platform listing nor a positive calculator result authorizes the agent to move funds.
Methodology and limitations
DN derived the scenario equations from a constant annual effective-return convention and checked the stated examples mathematically. The article uses primary documentation to distinguish yield fields, variable rates, withdrawals and bridge dependencies. It does not present a measured protocol leaderboard or claim that every eligible product was searched.
The model excludes changing asset prices, depegs, default, liquidation, LP inventory effects, taxes and unentered expenses. It assumes entry costs reduce invested capital, target exit costs are paid at the end and transition capital earns no yield. Any materially different cash-flow structure requires a different model. Results are educational scenario analysis, not expected returns or a recommendation to enter a position.
Sources and evidence register
Primary sources reviewed on 21 November 2026. The article distinguishes documented facts, DN analysis and synthetic examples. Recheck changeable rules and configurations before acting.
- DefiLlama yield methodology — Base/reward/total APY fields and exclusions; no live rates used.
- DefiLlama update frequency — Hourly yield data and API/web cache timing; used as source cadence, not execution guarantee.
- Aave supply tokens — Variable supply rates affected by utilization and governance.
- Aave withdraw tokens — Pool liquidity and collateral constraints.
- ERC-4626 — Asset/share/preview/limit distinction.
- ERC-7540 — Asynchronous request states.
- Ethereum bridge overview — Bridge dependencies and risk categories.
- Uniswap impermanent loss — LP price/inventory outcome not captured by fee APY alone.
Frequently asked questions
Can an AI agent automatically find the best APY?
It can search declared sources and compare observed rates. That does not establish the best future net return, because costs, liquidity, rates and asset exposure can change.
How do I compare staying with switching DeFi positions?
Compare both paths at the same horizon and reporting unit, including entry and exit costs and time without yield. Keep exposure and withdrawal constraints separate from the arithmetic.
Does a higher APY always justify a bridge?
No. The incremental earnings may not recover switching costs within your horizon, and a bridge adds dependencies that the cost calculation does not price.
Is a missing reward-APY field the same as zero?
No. The source may provide only a total APY without a component breakdown. Preserve unknown values until the source definition supports an interpretation.
Does the calculator use live yields or execute trades?
No. It uses reader-entered hypothetical assumptions and does not connect a wallet, inspect a live position or authorize a transaction.




