[ Documentation · 03 / 08 ]
R1 Sovereign to R5 Frontier
Every vault carries exactly one tier, and it is stated on every surface the vault appears on. What each tier means in practice, what a member is accepting at each level, and how a tier is assigned.
01
Five tiers. Every vault carries exactly one, and it is stated on every surface the vault appears on — the index row, the deep-dive hero, the card in the featured band, the previous and next links at the foot of a vault page. There is no view of a vault that does not show its tier.
R1
Sovereign
Reserve-grade. Cash-equivalent, never gated, first to settle.
Reserve
R2
Senior
Hedged or late-stage. Predictable mark.
Anchor · Shield
R3
Balanced
Mixed stage, moderate concentration.
Perpetual · Structure · Compound
R4
Directional
Concentrated or momentum-led. Drawdown expected.
Keystone · Momentum
R5
Frontier
Earliest stage, longest hold. Uncapped upside, uncapped loss.
Frontier · Genesis
The scale describes what a member is exposed to. It is not a prediction and it is not a confidence rating. A low tier is not a promise of return — Reserve sits at R1 and is marked at 1.02×, which is as high as it will ever be marked. A high tier is not a warning that a vault is performing badly — Genesis sits at R5 and carries the strongest mark in the book at 2.74×.
02
What each tier commits a member to, in the terms that matter when it goes wrong.
R1
Sovereign
1 vault · $132.7M · 6.8% of the book
You are accepting the credit of the issuer of a tokenised bill and nothing else. There is no company risk in this tier because Reserve owns no company — it is 82% cash and 18% bills.
The upside is the bill yield. Reserve is marked at 1.02× and will never be marked higher. No gate has ever been applied to it and none can be applied without a governance vote published a full epoch in advance. In the March drawdown the House drew $41.0M out of Reserve in four days and every call settled at par.
R2
Senior
2 vaults · $554.4M · 28.6% of the book
You are accepting timing rather than survival. Anchor's six companies are past product risk and inside two years of a plausible listing; the return depends on the window they list into rather than on whether they make it. A closed market pushes the whole book out four to six quarters, and the 24-month lock-up exists so that a member cannot force a sale into that window.
Shield's residual exposure is basis. It sells a public technology index against all four of its holdings, which costs roughly 4.2% a year in carry and caps the upside on purpose. A private mark and a public index do not move together, so in a dispersion event the hedge can lose on both legs.
R3
Balanced
3 vaults · $637.2M · 32.8% of the book
You are accepting duration rather than direction. Capital in Perpetual is committed for as long as the company stays private, the weighted holding period is 4.1 years, and there is no scheduled liquidity of any kind. Members who need a date should be in Anchor or Reserve, and the vault says so on the way in.
Structure's exposure is the capital stack. A liquidation preference is only worth what the estate can pay, and two of its positions have been carried flat since entry because no round has repriced them. Compound is rated here on the positions it holds, not on the fact that it is funded only by realised gains.
R4
Directional
2 vaults · $453.6M · 23.4% of the book
You are accepting drawdown as an expected outcome rather than a failure. Keystone holds exactly ten positions, so a single write-down moves the whole vault: a total loss on its largest holding would take 11.4% off vault equity in one epoch. That figure is published and recomputed every time the book changes.
Momentum turns over, pays fees and takes the wrong side sometimes. It has taken two down rounds since Epoch 02, the worse of them 22% on a single position, and both are still held. Its maximum drawdown to date is 9.4% of vault equity, recovered inside two epochs. If primaries stop repricing upward the strategy has no entry at all.
R5
Frontier
2 vaults · $162.1M · 8.4% of the book
You are accepting that positions in this tier can return nothing, and that some are expected to. Genesis holds six positions written before a public price existed and expects two of them to be worth zero. That is the base case, not the failure case.
Frontier's companies fail on physics and on permits rather than on product-market fit. A reactor that does not reach net power is worth its scrap and its patents; a robotics programme that misses a customer milestone reprices 60% in a single round. The lock-up is 60 months and it is real — there is no secondary market of any depth for most of this book.
03
A tier is a property of the mandate, not of a quarter's performance. It is set when the vault opens and it moves only by governance vote published a full epoch in advance. A tier that could be reclassified quietly would be worth nothing, so it cannot be.
| Factor | Effect on the tier |
|---|---|
| Stage | Earlier is higher. Genesis writes before a price exists and is R5; Anchor requires audited revenue above $200M and is R2. |
| Hedge | A proved hedge lowers the tier. Shield holds the same class of asset as Anchor, sells the index against it and attests net exposure every block, so it is rated R2 rather than R3. |
| Concentration | Fewer positions is higher. Keystone's ten names put it in R4 despite holding companies that also sit in R2 vaults. |
| Duration | Longer commitment is higher. A 60- or 84-month lock-up with no secondary market behind it is a different instrument from an 18-month one. |
What does not decide it: the mark, the names in the book, or how the vault has done. Keystone and Anchor hold several of the same companies — Anduril, Groq, Perplexity, Ramp, Mistral, Vercel — and sit two tiers apart, because Keystone holds ten positions on a 48-month lock-up and Anchor holds six on a 24-month one. The tier reflects the structure, not the roster.
04
A loss inside a vault is contained by that vault. This is the single most important structural property of the book and it is worth stating in full.
Sizing is the second control, and it is applied before the risk is taken rather than after. Genesis is sized so that a total loss costs the House 2.2% of deployed capital. Frontier caps any single position at 24% of the vault and the whole vault is 6.1% of capital deployed. Keystone publishes its concentration number rather than modelling it away.
What isolation does not do
Isolation limits contagion between vaults. It does not limit loss inside the vault a member actually holds. If you allocate to Genesis and two of its six positions return nothing, that loss is yours in full — the fact that it could not reach Anchor is no comfort to somebody who is not in Anchor. Diversification across tiers is a decision each member makes at allocation, and nothing in the structure makes it for them.
05
The book is weighted to the middle. That is deliberate: a book where every tier holds the same weight is a book nobody underwrote.
| Tier | Name | Vaults | Deployed | Share |
|---|---|---|---|---|
R1 | Sovereign | 1 | $132.7M | 6.8% |
R2 | Senior | 2 | $554.4M | 28.6% |
R3 | Balanced | 3 | $637.2M | 32.8% |
R4 | Directional | 2 | $453.6M | 23.4% |
R5 | Frontier | 2 | $162.1M | 8.4% |
61.4% of deployed capital sits in R2 and R3. The two R5 vaults together hold 8.4%, which is what makes their base case survivable at the level of the whole book — and irrelevant at the level of a member who is only in them.
06
Each tier carries a colour, and the risk ramp is the only place this site uses colour beyond its single accent. R1 is the brand lime and the ramp runs warm through amber to red at R5.
R1
Sovereign
Reserve-grade. Cash-equivalent, never gated, first to settle.
#C6FF00
R2
Senior
Hedged or late-stage. Predictable mark.
#E8E64A
R3
Balanced
Mixed stage, moderate concentration.
#FFC53D
R4
Directional
Concentrated or momentum-led. Drawdown expected.
#FF8A3D
R5
Frontier
Earliest stage, longest hold. Uncapped upside, uncapped loss.
#FF5140
It is a hue ramp rather than five tints of one colour because a tier is read off a 3px bar or a small badge at a glance, and five tints separated only by brightness are not recoverable at that size. Running from lime to red also matches the direction of the reading: R1 is reserve-grade, R5 is uncapped loss.
The colour is an aid and never the statement. Every surface that shows a tier colour also shows the tier code and the tier name, so nothing on this site requires a reader to tell amber from orange. If you can only see one of the three, the code is the one that matters.
The tiers as they appear across the book are on the vault index, sorted by tier. The mandate behind each one is in Vaults.