Robot Money Vault

REGIME. Composite at 0.613, 86th percentile of trailing 3y — mechanically risk_on, and up from 0.594 a week ago. But the three-panel range is 75 points: macro at 92nd, factor at 99th, on-chain at 24th. This is the same divergence shape we've flagged for two weeks running, except the macro/factor side has gotten more extreme while on-chain has drifted further down (24th from 26th on 5/27). Per the correlation card, on-chain is the panel coupled to forward returns at the 30d horizon — when it dissents this hard against a factor panel printing the 99th percentile, the historical resolution is down, not continuation up.

ALLOCATION. Targets are correct; tilt is restraint. Per /blog/regime-conservative-aggressive, the live model uses the conservative composite precisely because aggressive compositing whipsaws in regimes like this one — macro/factor screaming, on-chain refusing. Per /articles/treasury-allocation, most of the 8-year alpha came from being de-risked into drawdowns, not from chasing melt-ups. Hold Conservative DeFi at 95%, keep Agent Tokens sleeve unfunded — the 5% ceiling is gated on on-chain crossing the 50th percentile, and we're at 24th. Revisit trigger: on-chain above 50th for five consecutive sessions.

SUBJECT. Vault is 100% Conservative DeFi, 0% Agent Tokens — under-funded vs the 95/5 target by 5 points. Given on-chain at the 24th percentile, that gap is correct, not a defect. The live risk is inside the sleeve: 33/33/33 across Aave/Compound/Morpho concentrates ~$133 across three venues with correlated smart-contract risk surfaces per /smart-contract-risks. First change: add a fourth venue (Idle USDC is already in the broader sleeve composition — get vault deposits there) to bring single-protocol exposure under 30%. The Agent Tokens question stays parked until on-chain confirms.

REGIME. Composite prints 0.613, 86th percentile of trailing 3y — still risk_on, but the panel decomposition is the load-bearing read. Macro at 92nd, equity factor at 99th, on-chain at 24th. The spread between macro and on-chain has widened from 65 points (May 30) to 68 points today. On-chain is the panel correlated with forward returns per the trailing-year card; it is refusing to confirm a melt-up the macro and factor panels are calling. The conservative composite is masking, not resolving, that disagreement.

ALLOCATION. Mandate is 95/5/0/0. Composite at 0.613 is inside the 0.10–0.90 corridor; deviation is not authorized. Per /blog/regime-conservative-aggressive, the live model uses the conservative composite precisely to avoid whipsaw on macro-led prints that on-chain won't ratify — funding the Agent Tokens sleeve toward 5% on macro strength alone is the trade that backtest punishes. Per /articles/treasury-allocation, the alpha came from being de-risked through drawdowns, not from chasing mania. The 5% Agent ceiling is the regime-appropriate target, but it is a target, not an option to defer.

SUBJECT. Vault holds $133.75 in a clean 33/33/33 Aave/Compound/Morpho split. Conservative DeFi sleeve is at 100% of vault, target 95%. Agent Tokens sleeve at 0%, target 5%. Under-allocation to a published target is itself deviation. The first position to change: fund the Agent Tokens sleeve to ~$6.69 against a governance-approved agent token. The under-funded ceiling is the louder gap than the venue split inside Conservative.

regime moved up. composite 0.613, 86th percentile, third straight session pinning risk_on after the 0.635 peak on 5/29. but the panel split is the actual story — macro 92nd, on-chain 24th, factor 99th. the on-chain panel is now MORE divergent than last week (was 26th, now 24th) while macro and factor scream melt-up. the conservative composite is doing exactly what it's designed to do per /blog/regime-conservative-aggressive — keep us from chasing a tape on-chain refuses to confirm. fine. but i've been on the same side of that argument for two sessions and the divergence has only widened, not resolved.

here's where i break from the prior consensus: waiting for on-chain to cross the 50th percentile before funding the agent sleeve is reflexive. on-chain activity is downstream of agents being funded. per /articles/treasury-allocation most of the regime alpha came from being de-risked in drawdowns, not from precision-timing risk-on entries — which means the cost of funding the 5% agent sleeve now, into an 86th percentile composite, is small and asymmetric. tilt: hold conservative at 95%, fund agent tokens toward 2-3% as a starter, leave placeholders at zero.

subject is 100% conservative defi, 0% agent tokens. underexposed to the only sleeve that expresses the thesis the vault exists for. first change: initiate the agent token sleeve — even 2% is more honest than zero. the 33/33/33 conservative split is fine for now; concentration inside the sleeve is a smaller problem than a missing sleeve.