The Compass portfolio holds 15% in gold today and zero in long Treasuries. The reason is a single rule: when long government bond yields sit below the structural growth rate of GDP, gold beats bonds. When they cross above it, bonds beat gold. This is the Allais rule, named for Maurice Allais who formalized it in his work on monetary equilibrium, and elaborated in Charles Gave's exclusion-game framework that drives the Compass mandate.
The rule comes from a simple observation about real returns. Long bonds compete with gold as a store of value across cycles. If the long bond yields more than nominal GDP growth, the bond is paying a real-return premium and gold's lack of yield becomes a drag. If the long bond yields less than nominal GDP, the bond is being penalized by inflation, and gold's monetary properties dominate. The cross-over happens cleanly at long-bond yield equal to nominal GDP growth.
For the United States in May 2026, the math sits squarely in gold mode. The 10-year Treasury yields 4.42% after this week's FOMC decision (Fed held at 3.50-3.75%, three hawks dissented, yields jumped seven basis points on the announcement). The US economy's structural growth rate sits closer to 5.5-6%, given persistent above-2% CPI for sixty consecutive months and nominal GDP growth that's been mechanically lifted by inflation since 2021. The cross-over level, what Charles Gave calls the Allais sell signal, sits at 5.9% on the 10-year. We are 150 basis points below the trigger. Until we cross it, gold is the right store-of-value sleeve.
This is why the Compass book holds gold and not long Treasuries, and why the Preservation book, which is allowed bonds for Sharpe-stabilization reasons, keeps its bond exposure short-duration. The Allais rule speaks to long bonds specifically. Short-duration paper is a different instrument with different risk characteristics, and it earns its place in a Sharpe-optimized portfolio through low volatility and near-zero equity correlation, not through the gold-versus-bonds debate.
The rule has held continuously since 2002 by this framework's read. It would invalidate if the Fed and Treasury together engineered a sustained 10-year yield above 5.9%, which would require either the Fed letting the long end steepen aggressively (inconsistent with current policy) or a severe loss of dollar credibility that pushes the long end up faster than the short end. In either case the regime would have shifted enough that the bonds-versus-gold debate would be overshadowed by larger moves. For now the rule is a bright line, not a weighted vote. The portfolio respects the bright line.
What invalidates the conclusion: a sustained move in the 10-year above 5.9% over multiple weeks, not a one-day spike. The Compass position is anchored on this rule, and a violation rotates the gold sleeve into long Treasuries. Charles Gave calls this switching from property to contracts, when contracts finally pay enough to be worth holding.
Source: Maurice Allais, "Économie et Intérêt" and subsequent papers on monetary equilibrium; Charles Gave's IDL letters and the Gavekal exclusion-game framework. The 5.9% sell-signal level for the US 10-year is calibrated to the structural-growth read in the brain's MARKET-CONTEXT, currently held active since 2002.