Disclosure first: I’m part of the team behind Sentralis, which published an independent risk teardown of the Endowment here last month, built on the June book. Nobody at ENS or kpk asked or paid for it, and none of it is advice. That analysis ended by proposing three quantities worth tracking on a standing basis. The H1 review above and the July community update moved all three, so here are the updated readings, re-run on the July book (positions reconstructed from the July update’s own transaction list, prices as of August 17; all figures are model estimates under the same stated assumptions as the teardown, none of them facts or advice).
1. Stablecoins sellable within a crisis week: $12.4M → $23.4M. This is the largest single risk improvement in the book this year, and it did not come from prices. July’s redeployment of the exited $16M DAI position into USDC (and a little USDT and GHO) replaced the book’s slowest stablecoin with its fastest: under the same crisis assumptions (5% of daily volume, order books only), the DAI needed ~11 days and missed the week, while the USDC that replaced it clears in about a day. Nearly all of the $26.9M stablecoin side is now week-scale money. The redemption-path caveat from the teardown still applies in the treasury’s favor.
2. Largest single protocol: 25.3% → about 22.7%, with a nuance. After the July moves (Aave exited, ether.fi and the new Lido position added, StakeWise and Stader trimmed), no protocol holds more than ~22.7% against the 30% cap, and the modeled single-protocol incident (temporary freeze, 80% recovery, 20% haircut) now comes to $5.4M–$6.3M (7.0–8.2%) for any of the top four, down from ~9% for the top two in June. The nuance: those top four (Fluid, StakeWise, ether.fi, Stader) are now nearly equal in size and together hold ~85% of the book, so the effective number of protocols, ~5.2 by concentration index, is essentially unchanged from June’s ~5.5. The book got flatter at the top, not more spread out.
3. Floor-coverage probability: the model spread widened to 49% - 90%. Across the same two Monte Carlo models (25,000 paths, one year, seeded), the probability of ending the year above the $49.34M floor figure is now 90% under the zero-drift model and 49% under the trailing-year bootstrap (it was 88% and 61% on the June book). Two things moved it: at August prices the book is back to ~65/35 ETH/stables, and the bootstrap’s trailing year now contains the full August-2025→2026 ETH decline, which makes its resampled year more negative (its median simulated year is −36%). The gap between the two models remains the point: anything sized off the floor inherits a spread that model choice alone moves by a factor of five.
Two smaller updates. A repeat of the current bear’s core leg on the July book still costs on the order of 40% (the replay window has lengthened as the event ran on, so this is not comparable to the June figure). And crisis-regime correlation stress still does nothing (−1% VaR change): the diversification that exists continues to live entirely on the stablecoin side.
None of this changes the teardown’s overall reading: an ETH-ecosystem endowment accepting ETH risk deliberately, with the defensive side now faster than it was. The H2 items in the review (the RWA sleeve phase-in, the ETH/stable split proposal, PUR #11) will each change these numbers when they land; happy to re-run any of them against the actual parameters at no cost, the same offer as before.
Analysis produced with Sentralis, a cryptocurrency portfolio risk- and scenario-analysis solution, suitable for beginner HODLers and professional investors alike.