Call #1 · Published 9 July 2026
Clearstar Earn USDC: cut out of Conservative the day its yield died
Euler v2 vault · Base · 0x8bf4…4247
On 26 June 2026 this vault scored 70.5 — just inside Conservative, the band a capital-preservation allocator can hold — while paying a 5.7% USDC supply yield. On 27 June its yield engine stopped: supply APY collapsed from 5.71% to effectively zero. The Atlas Score cut it 9.5 points the same day, out of the Conservative band. Within a day, 44% of the vault's deposits were gone.
The yield never came back. As the dead rate persisted, the score kept falling — 70.5 to 44.4 in twelve days— even as some deposits returned in early July. A score is a judgment about the vault, not a mirror of its TVL: money flowing back into a vault that pays nothing does not make it Conservative again. The vault has since dropped out of Atlas' tracked depositable universe.
| Date (UTC) | Atlas score | Band | Supply APY | TVL | What happened |
|---|---|---|---|---|---|
| 2026-06-23 | 68.8 | Balanced | 5.73% | $1.08M | First day in the scored universe. |
| 2026-06-25 | 70.5 | Conservative | 5.71% | $1.08M | Crosses into the Conservative band. |
| 2026-06-26 | 70.5 | Conservative | 5.71% | $1.08M | Last day the vault paid its advertised yield. |
| 2026-06-27 | 61.0 | Balanced | <0.01% | $1.08M | Supply yield stops. Atlas cuts 9.5 points the same day — out of Conservative. |
| 2026-06-28 | 59.8 | Balanced | <0.01% | $606K | 44% of deposits exit within a day of the cut. |
| 2026-06-30 | 48.7 | Balanced | <0.01% | $597K | Yield still dead; stability and persistence factors keep grinding the score down. |
| 2026-07-03 | 46.9 | Balanced | <0.01% | $993K | Some deposits return. The score does not — the yield is still gone. |
| 2026-07-09 | 44.4 | Balanced | 0.06% | $889K | Twelve days on: yield never recovered. Score sits near the floor of Balanced. |
What moved, factor by factor
The cut was driven by the Yield pillar, then confirmed by Safety and Liquidity. Yield level went to the floor when the supply rate hit zero. APY stability and APY persistence — which read the shape of the rate history, not just its latest value — degraded over the following days as the collapse stopped looking like a blip and started looking like the new state of the vault. On the safety side, TVL depth marked the vault down as deposits fled: a sub-$1M vault has less room to absorb a bad exit. Factor definitions are on the docs; per our publishing policy we do not disclose weights or formulas.
What we are claiming — and what we are not
- We are claiming the score converted raw on-chain deterioration into a legible, dated downgrade the same day it became observable — and a day before most of the money moved. An Earn surface quoting Atlas would have shown its users the downgrade on 27 June.
- We are not claiming clairvoyance. The score reacted to the yield stopping; it did not predict it days in advance. Depositors who fled on 28 June were reacting to the same on-chain facts. The value of a score is speed, consistency, and a public record — not prophecy.
- Data-quality flags: the vault entered the scored universe on 23 June, so its earliest rows carry a fallback data-quality flag (thin history); the flag improved to partial on 30 June and sufficient from 7 July. We publish these flags rather than hide them.
- Verification:Atlas' git-verifiable public snapshots begin 9 July 2026, so the June rows above are quoted from our internal score history (scorer 2.0.0), which predates the public record. Calls made from here on are fully verifiable against the public snapshot repository.