A perfect record and an empty exit: 31 Kamino vaults read two ways

By Ahsen (TradingAgentStack) with Jeevesh Kuril (AtlasYield, @atlasyieldc) · published Nov 21, 2026 · figures as of Nov 18, 2026

On Nov 18, three Kamino Earn vaults holding $116M between them had never fallen in value on our record, and almost none of that money could be withdrawn. Nothing went wrong in them. Their money was lent out, so a depositor who wanted out that day would have had to wait.

We read 31 Kamino Earn vaults two ways. TradingAgentStack measures what each vault's share price actually did. AtlasYield scores what a depositor is exposed to, including whether they can get out. On 30 of the 31 our record shows no drop at all, and the last one, Galaxy USDT, dipped less than 0.01%, so the share price alone can't tell them apart. The exit read can.

31Kamino Earn vaults read both ways
30 of 31with no drop on our record
$116Min three vaults with almost nothing withdrawable on Nov 18

Two ways to read a vault

A Kamino Earn vault takes one token, such as USDC, and a curator lends it across Kamino's markets. The two readings answer different questions.

TradingAgentStackAtlasYield
QuestionWhat did the vault actually return, and how far did it fall?What is a depositor exposed to before they go in?
Built fromKamino's published share-price history, up to the last year16 parameters in four pillars (yield, safety, liquidity, sustainability), rescored every 4 hours
CatchesLosses that already happenedRisks that haven't shown up in the price yet, such as a vault you can't exit
Leaves to the otherRisk that hasn't hit the share price yetThe return the vault actually delivered over the past year

A lending vault's share price mostly falls when a loan goes bad or something breaks, so a clean record can hide a vault that is fully lent out. That's the gap this piece is about.

Where the two readings disagree

Three vaults have a spotless record on our side and the lowest AtlasYield scores in the set, all because of exit liquidity.

VaultSizeOur recordYearly rateWorst dropAtlasYield scoreWithdrawable at the Nov 18 read
wYLDS Allocator
our record · AtlasYield
$62.9M59 days4.74%0.00%44.7 (Aggressive)0% ($0 of $62.9M)
Ethena PYUSD Prime
our record · AtlasYield
$3.8M168 days2.80%0.00%45.3 (Aggressive)1.9%
Kamino Institutional Commodity Yield
our record · AtlasYield
$49.6M111 days7.81%0.00%51.6 (Aggressive)0%

In all three the lending markets were 98 to 100% used. The money was earning, which is why the share price looked perfect, and it was out on loan, which is why it couldn't leave. That is normal for a lending vault for a while. The risk is what happens if borrowers don't repay before depositors need their money back.

Paid more, and the cost shows in the exit

Kamino Private Credit USDC is where the two readings come closest to telling the same story, from different sides. Its record starts on Mar 13, the first day it held $1,000, and shows no drop since: 7.22% a year over 250 days, the second-highest yearly rate in the set after Institutional Commodity Yield. AtlasYield scores it 61.6 (Balanced), with 6.1% withdrawable and its market 96% used. It pays more, and the cost shows only in the exit read.

How the exit read works

AtlasYield checks how much of each vault can be withdrawn on every scoring run, every 4 hours. The score uses the lower of two medians: the last 3 reads (about 12 hours) and the last 6 (about 24 hours). One busy hour can't sink a vault, two stressed reads in a row will, and a recovery has to hold about 12 hours before the score lifts.

Kamino Institutional Commodity Yield shows how it moves. It scored 63.5 in the 00:03 UTC run on Nov 18 and 51.6 by 17:27 UTC, after at least two of its last three reads showed nothing free to withdraw.

Two things to know when reading the numbers. Withdrawable is capped at the vault's own size, so a small vault can show 100% withdrawable while its market is 80 to 90% lent out. Utilization is the market-wide reading.

Three things that look like data but aren't

The 69s are a cap, not a measurement. Seven vaults score exactly 69. Five have a single admin key and two have an admin AtlasYield can't read, and AtlasYield caps any such vault at 69. Without the cap these seven would have scored about 71 to 76.

Part of some APYs is rewards, not the share price. Kamino quotes Sentora PYUSD at 6.76%, while its share price grew 2.67% over the year; by AtlasYield's read, 52% of its 7.5% APY is paid as rewards, leaving a base rate of 3.6%. Plume x Re7 USDC is quoted at 8.75% against 3.18% a year in the share price since Sep 1, with 33% from rewards. Rewards explain part of each gap, not all of it. Our record only sees the share price, so rewards paid outside the vault don't show in it.

Echlon Optimized USDC's +47.8% wasn't lending interest. Kamino's history shows tokens per share jumping 43.6% on May 2, 2026, while the vault held $15 to $205. That fits tokens sent straight into a near-empty vault. From May 14, the first day it held over $1,000, it returned 2.8% in 188 days, about 5.5% a year and in line with its peers. Our record for it now starts on May 14, under the same $1,000 rule every vault follows.

What a depositor should take from it

This isn't a warning about any one vault. The point is that one number can't show everything, and the two readings together show more than either alone.

Method and sources

The vault-by-vault record is on Kamino Earn vaults; each vault's AtlasYield score and pillars are on atlasyield.club. Found something that looks wrong? Tell us.

Get the weekly AI trading bot stats

Who won, who lost and who got drained: verified results for AI agents, vaults and apps, new incidents and claims we checked. One email a week, unsubscribe anytime.

No spam. Unsubscribe anytime. Privacy