Every currency in history has been a melting ruler — a unit of measure that shrinks while you hold it. GLU is designed the way physicists design units: define the invariant first, then build the instrument that realizes it.
1 GLU := the cost of the reference world consumption basket W(t), a consumption-weighted composite of the five major currency areas' CPI baskets, fixed at epoch t₀ = 2026-01.
The metre was once a platinum bar that warped; it is now defined by the speed of light. The dollar is still the platinum bar. GLU is the redefinition: not a claim on any nation's money, but on a fixed quantity of what money is for.
Nothing is stable in itself — only stable relative to something. USDC has zero volatility against the dollar and lost roughly a fifth of its purchasing power since 2020. The first design decision of "the most stable asset on Earth" is therefore not collateral or consensus. It is the choice of numeraire: the thing you measure stability against.
GLU's answer: measure against consumption itself. Define the world price level as the consumption-weighted geometric mean of the five major currency areas, each term being the USD cost of that region's CPI basket:
W(t) = ∏ᵢ [ CPIᵢ(t) · Sᵢ(t) ]vᵢ i ∈ {USD, EUR, JPY, GBP, CHF}
CPIᵢ = region i's consumer price index (local currency) · Sᵢ = USD price of currency i · vᵢ = consumption weights (0.40 / 0.30 / 0.15 / 0.10 / 0.05). The GLU token's target price in dollars is W(t): as the world's cost of living rises, GLU's fiat price rises to match, so its purchasing power stays constant by construction.
To see why this target is the right one, I backtested it against 27 years of real data — monthly CPI from FRED, the OECD and the BIS, and daily FX fixes, January 1999 through December 2025. The chart below is the purchasing power of holding each currency in cash, measured in the world basket.
Formally, GLU minimizes the variance of purchasing-power returns: choose the asset X that minimizes Var(Δ log [Pₓ(t)/W(t)]), then remove the drift E[Δ log(Pₓ/W)]. The backtest shows these are two separate problems — and every prior "stable" asset solved at most one of them.
Solving the minimum-variance problem min wᵀΣw over monthly purchasing-power returns of the five currencies gives a basket with 0.86% annualized volatility — against 3.9% for the dollar, the calmest single fiat. And the optimizer's weights come out almost exactly equal to the consumption weights themselves. That is the deep result: the best hedge for world consumption is the world's own spending pattern. No fitting cleverness required — which is also why it survives out of sample.
| Asset | PP volatility /yr 1999–2025 | Out-of-sample 2013–2025* | PP kept 1999→2025 | Max real drawdown |
|---|---|---|---|---|
| Japanese yen | 6.61% | 6.12% | 44% | −61% |
| British pound | 4.99% | 4.99% | 49% | −52% |
| Swiss franc | 4.95% | 4.08% | 105% | −20% |
| Euro | 4.21% | 3.30% | 61% | −42% |
| US dollar | 3.86% | 3.23% | 61% | −45% |
| Min-variance basket | 0.86% | 0.94% | 58% | −42% |
| GLU (index-pegged) | ≈ 0† | ≈ 0† | 100% | 0% |
* Basket weights fitted on 1999–2012 only, then held fixed — the 0.94% is honest out-of-sample performance. † By construction at the index level; realized stability depends on the peg machinery of Spec 05. Min-variance weights: USD 40.0 / EUR 31.5 / JPY 12.7 / GBP 11.6 / CHF 4.2 — versus consumption priors 40 / 30 / 15 / 10 / 5.
Look at the basket's row: 0.86% volatility, and it still lost 42% of its purchasing power. An SDR-style currency basket is smooth but melting — it diversifies away the wiggles while keeping the average of everyone's inflation. This is the error every basket-currency proposal makes. Volatility and drift are different diseases:
Consumption-weighted diversification across currency areas: 6.6% → 0.86%/yr. Robust out of sample because the weights are structural, not fitted.
The peg target is the price level W(t), not the currencies. GLU's fiat price ratchets up with world inflation (≈1.9%/yr over the sample), so purchasing power never bleeds.
Which means GLU appreciates against every fiat, forever, in nominal terms — that is not a bug, it is the entire product:
The inflation oracle is the single biggest trust surface in this design — the prior art proves it. Truflation, the leading "real-time CPI" oracle, was hacked for ~$5M in 2024, and in 2026 a single internal model change moved its measured housing inflation by 1.56 percentage points. Nuon pegged to an index run by its own founders and died with zero credibility. GLU's oracle is designed against both failures.
The index uses official statistics — BLS CPI-U, Eurostat HICP, Japan/UK/Swiss national CPIs — published by agencies that are slow but canonical, and cannot be repriced by any crypto-native actor.
Each statistic enters onchain as the median of N independent oracle operators reading the same public release. Real-time feeds (Truflation-class, TIPS breakevens) may nowcast between releases but are capped at a few basis points of influence and always converge to the official print.
Monthly CPI releases are interpolated into a smooth daily accrual (USDi already does this with CPI day-weighting). The target moves single basis points per day, on a schedule known weeks in advance — arbitrageurs can never be ambushed by the peg itself.
Basket weights vᵢ are revised at most annually, inside hard caps (±2pp per revision), by a body that holds no GLU seigniorage. The issuer can no more move the index than a fund manager can edit the S&P 500.
A CPI-pegged token is a perpetually growing liability. This is what quietly killed Frax's FPI: it promised CPI and backed the promise with generic DeFi yield, betting it could out-earn inflation. When the yield fell short, the design bled real value by construction. The only honest fix is asset–liability matching: back an inflation-indexed liability with inflation-indexed assets.
Inflation-linked government bonds are the exact instrument: their principal is contractually indexed to the same CPIs that define W(t). A ladder of them, held in index weights, replicates the liability almost term by term — and pays a real yield (currently ~1–2%) on top, which funds operations and a stability buffer instead of leaking value.
| Sleeve | Weight | Instrument | Function |
|---|---|---|---|
| US real | 40% | TIPS ladder, 0–5y | Matches the vᵤₛ·CPI-U term of W(t) |
| Euro real | 30% | Bund-ei / OAT-ei ladder | Matches the HICP term |
| Japan real | 15% | JGBi ladder | Matches the Japan CPI term |
| UK real | 10% | Index-linked gilts, short end | Matches the UK CPI term |
| Swiss + hard buffer | 5% | CHF bills + gold (CH issues no linkers) | Matches the CHF term; gold hedges coordinated fiat debasement |
| Liquidity sleeve | ~5% overlay | T-bills / tokenized money funds | Same-day redemptions without touching the ladder |
Everything above already exists onchain in fragments — $15B of tokenized Treasuries, and exactly one tokenized TIPS ETF (via Ondo) — but nobody has assembled the matched ladder. That is the empty lane. A capped, overcollateralized crypto sleeve (ETH/BTC vaults, RAI-style) can be bolted on later for censorship-resistant minting, but v1's credibility comes from the boring ladder.
BLS · Eurostat · e-Stat · ONS · FSO, each as an N-operator onchain median
Chainlink-class feeds for EUR, JPY, GBP, CHF vs USD
Truflation-class + TIPS breakevens, capped at basis-point influence
The unit definition. Publishes today's + next 30 days' target path.
Primary market: 1 GLU ⇄ W(t) dollars of collateral, always. Hard arbitrage, small spread.
Global inflation-linked bond portfolio in index weights + liquidity sleeve.
GLU/USDC concentrated liquidity, recentred daily along the published drip path
Wallets display GLU with a live fiat conversion — users spend fiat amounts, hold GLU units
Real-yield surplus accrues to a junior buffer that absorbs tracking error before holders ever do
Ampleforth's SPOT "targets, but does not promise" the CPI-adjusted dollar. In August 2026 it trades at $0.21 against a ~$1.25 target — an 84% miss with eight dollars of daily volume. A target without a redemption right is a story, not a peg.
GLU's peg is a hard, always-on primary market: any holder mints new GLU by depositing W(t) dollars of eligible collateral, and redeems 1 GLU for W(t) dollars of collateral (in kind or via the liquidity sleeve), minus a small fee that widens only under stress. Every deviation on any venue becomes a riskless arbitrage against the NAV — the same mechanism that keeps a $2.5B tokenized money fund at par, pointed at a moving target that everyone can compute in advance. Because the target path is published 30 days forward (Spec 03), market makers can quote around the drip with no oracle ambush risk.
The peg is not defended. It is arbitraged into existence, continuously, by anyone.
Every prior attempt at a purchasing-power-stable coin either worked technically and died commercially, or never bound its peg at all. Each corpse contributes one rule to this spec.
| Project | Mechanism | State · Aug 2026 | Rule it wrote into GLU |
|---|---|---|---|
| Frax FPI | CPI-U peg, backed by protocol yield strategies | ZOMBIE · ~$7M | Never back an indexed liability with nominal yield — match it (Spec 04) |
| Ampleforth SPOT | Senior tranche on rebasing AMPL, soft target | 84% OFF TARGET | Hard mint/redeem or nothing (Spec 05) |
| RAI | ETH-backed, PI-controller floating peg — no CPI link | WOUND DOWN 2025 | Low volatility without a legible unit ≠ a product; people must understand what 1 unit is |
| Nuon | Truflation-pegged, overcollateralized — index by the same founders | DEAD | The index must be independent of the issuer (Spec 03) |
| Truflation | Private real-time CPI oracle | LIVE · bruised | Real-time feeds nowcast only; official statistics anchor (Spec 03) |
| USDi | Daily CPI-interpolated fund share, TradFi-run | LIVE · niche | The legal wrapper is a fund — design distribution around that from day one (Spec 07) |
The meta-lesson across all six: nobody died of peg mechanics; everybody died of no demand. Mechanism design is necessary but the binding constraint is a reason to hold — which for GLU is the real yield passed through to holders (the ladder earns CPI + 1–2%; a nominal stablecoin's issuer keeps ~5% nominal while your purchasing power melts).
Under the US GENIUS Act (2025), a "payment stablecoin" must redeem for a fixed monetary value. GLU's redemption value floats with the index by design, so it falls outside the stablecoin regime entirely and lands in securities law — functionally an investment-company share. That is precisely why USDi, built by TradFi inflation veterans, launched as an accredited-investor fund. Under the EU's MiCA, GLU fits explicitly as an asset-referenced token (ART) — a legal path that exists but that nobody has yet walked for an inflation index, with reserve, own-funds, and usage-cap obligations.
So the honest sequencing: v1 is a tokenized global real-return fund with a transferable share (Reg D/S or an offshore wrapper, KYC'd primary market, free secondary transfer where lawful), and the MiCA ART application runs in parallel for EU retail. Calling it "a fund share that behaves like the world's most stable currency" is not a concession — it is the accurate description of every fully-reserved stablecoin that ever worked.
Purchasing-power volatility of the measuring instruments themselves — the five CPI baskets — is the floor below which no asset can go. GLU is designed to sit on that floor. These are the residual risks, stated plainly:
Official statistics understate or lag lived inflation, and methodology changes move them. Mitigation: five independent national agencies, median aggregation, gold sleeve for coordinated-debasement tails. Not fully solvable.
Inflation-linked bonds sold off with everything in March 2020. The liquidity sleeve and redemption-in-kind option absorb runs, but a fire-sale window can open a temporary NAV discount.
2021-style −1% real yields would make the ladder earn slightly less than the index accrues. The equity buffer absorbs this; a sustained decade of it compresses the buffer. Disclosed, not hidden.
Prices are quoted in dollars; a unit that drifts +2%/yr against the dollar has UX friction. Mitigation: wallets display live fiat equivalents — you hold GLU, you read dollars. RAI proved illegibility is fatal; this is the counter-design.
| Stage | Ship | Proves |
|---|---|---|
| v0 | Live index + simulator. Publish W(t) daily from public APIs (the backtest pipeline behind this document already computes it), with a dashboard: "what your salary/savings are worth in GLU." No token. | The index is real, auditable, and people care — demand signal before a dollar of legal spend |
| v1 | Fund-wrapped token. Tokenized ILB ladder (start with the Ondo TIPS wrapper + short linkers via a custodian), daily NAV = W(t), KYC'd mint/redeem, USDi-style accredited launch | The matched-ladder mechanism holds NAV through real months of CPI prints |
| v2 | MiCA ART + public secondary. EU authorization, AMM liquidity along the drip path, real-yield pass-through to holders | Retail demand for the first asset whose chart is flat in the only frame that matters |
| v3 | Unit-of-account layer. Invoicing, payroll, and long-term contracts denominated in GLU — the endgame is not a coin, it is the ruler other things are priced with | — |
The dollar's chart only looks flat because we plot everything else in dollars. Plot the dollar in groceries and it is a ski slope. GLU is the asset whose chart is flat in groceries.