A whole family of crypto assets has quietly abandoned the peg. They do not promise you a dollar; they promise you an accurate reading. They go by a dozen names because nobody has named the family. This page names it, tours it, and shows the one thing every member is missing.
An index unit is a unit of measurement whose value is defined by a computed index. The subject is unrestricted: physical measurement, on-chain state, pure mathematics. The method is not, and has to be deterministic and repeatable by anyone. Not a redemption promise. Not a reserve defending a fiat price.
The inversion is the whole point. A dollar stablecoin like USDC tracks "one dollar" and holds reserves to defend that fiat peg; if you see a dollar sign, it is a promise. An index unit tracks "one unit of the index," and everything about it exists to defend the accuracy of the reading. If a dollar figure appears anywhere, it is display convenience, never the definition.
Two consequences follow. First, the coin is just the carrier: a transferable on-chain claim that tracks the index value. The unit is the thing being tracked; the token is how you hold it and move it. Second, an index unit is upstream of money: it is a unit of account and measure first, and only incidentally a store or medium of value. It can index anything deterministic and observable, and it does not even need to be tradable to be useful. A meter that modulates a system's own behavior is an index unit with no market at all.
These projects rarely cite each other and mostly do not know they are relatives. Read the self-descriptions in amber: reflex index, elastic unit of account, flatcoin, perpetual note, energy money. Different vocabularies, same move: value defined by a computed reading instead of a fiat promise.
Backed by ETH, pegged to nothing. Its redemption price floats, steered by an on-chain control loop responding to market price. Live since 2021, and the purest proof that a non-pegged unit can hold together: the index it tracks is its own controller's output.
A descendant of RAI's design with a basket of collateral types. Same thesis, second generation: the unit's value is whatever the control index says, and the market arbitrages toward it.
Targets an inflation-adjusted dollar and, instead of defending a price, rebases every wallet's balance daily so that the unit converges on the index. Live since 2019. The index it tracks includes an off-chain inflation reading, delivered by oracle.
Built from senior claims on AMPL's rebasing collateral, aiming at a steadier carrier for the same underlying index. An index unit derived from another index unit: composition arriving before the category has a name.
Pegged not to the dollar but to a consumer-price index, so the unit is meant to hold purchasing power rather than face value. The entire product is exactly as trustworthy as its inflation oracle, which is the point this page is building toward.
Same family as FPI with its own inflation methodology. "Flatcoin" is the name this corner of the field has been converging on: flat purchasing power, defined by an index.
Denominates value in units tied to the energy cost of computation, with issuance coupled to proof-of-work difficulty. The index here is physics plus protocol accounting: a reading of what a unit of computational work costs.
Wound down. Its postmortem is instructive rather than embarrassing for the category: the unit design was coherent, and the hard part turned out to be everything around the index: sourcing it, funding it, and making anyone able to trust it.
Targeted its own basket-derived value rather than a dollar. Quiet since its early experiments. Another data point that the bottleneck of this family is not the monetary mechanism; it is the reading underneath it.
Adjacent experiments (all-weather baskets, reserve currencies, algorithmic pegs) borrow parts of the idea; the graveyard of purely algorithmic dollar pegs is a different failure family and is not this list. What unites the cards above: none of them promises redemption into fiat. Each defines its unit by a reading.
Look at where each unit's honesty actually comes from. RAI trusts its market-price feed. AMPL and FPI trust inflation oracles. Nuon trusts its own methodology. Each project built or bought a private index pipeline, and each pipeline is the one part of the system its users cannot recompute. The monetary mechanics are on-chain and auditable to the last opcode; the reading they all obey arrives on trust.
That is the ceiling on the entire category: an index unit can never be more honest than its index. And it is where the failures cluster. The units that wound down did not die of bad monetary design; they died of the cost and fragility of maintaining a reading nobody could independently verify. Meanwhile every survivor pays the same tax separately: each maintains its own oracle plumbing, its own methodology disputes, its own trust story, for readings that are conceptually shared.
The Index Machine is not another entry in the gallery above. It is the engine and the standard those entries are missing: a deterministic computation layer where each reading is a pure function of public inputs, evaluated on a clock, sealed into a receipt, replayable bit for bit in anyone's browser, and anchored on Ethereum so the history cannot be quietly restated. Programs are content-addressed and the registry that carries them on-chain records only what exists, what depends on what, and how names resolve. An index unit built on it inherits, for free, the one property none of them has today: a reading whose honesty is checkable instead of promised, by anyone, without asking.
The banner is deliberately non-exclusive. RAI, AMPL, SPOT, FPI, Qi and their successors are kin, not competitors: anyone may compute against the standard, certify against it, or fork the implementation after the license converts. A category rallies its builders; a walled garden repels them. The adoption model is the one that made the VIX, the Fear and Greed index, and the CPI canonical: public legibility first, integration follows the number everyone already watches.
The projects in the gallery are kin, not competition. Every one of them carries a reading its own users cannot recompute. The standard is non-exclusive: anyone may compute against it, certify against it, or fork the implementation once the license converts. If you are building an index unit, an oracle, a flatcoin, a reference rate, or anything that has to answer where a number came from, there are two ways in.
Neither is a sale. The protocol has nothing to sell and no mechanism to sell it with. The economy runs on two instruments, and mining is the only tap either one ever comes out of.
INDX-Q is the work instrument and the liquid unit of the economy. One INDX-Q is one standardized quantum of verified network service. It is an index unit in the sense this page describes, instantiated on the standard itself: an index unit, not the index unit. INDX-G is the governance instrument. It is earned as a small fixed drizzle alongside the work emission, locked for voting weight, and consumed by governing: every governance action burns a fraction of it, and holders re-earn or re-acquire to keep voting. Those who do the work burn at half the rate of those who only hold. Reading and verifying the standard are free forever. The instruments gate power, never access to the numbers.
The founding cohort is a constitutional convention, not a standing council. It holds one class of authority over a whitelisted set of actions. Both standing vetoes apply to it throughout. It auto-sunsets on a fixed clock. Most slots are open to public and peer application. A seat carries governance weight and nothing else: never units, never emission, never standing. While the cohort runs, a seat holds nothing and burns nothing. At sunset the seat lapses unless its holder then holds INDX-G on the ordinary earned or held route. Seating opens when there is something to govern. Names go on the list before that.
Units come into existence one way: work-metered minting into the key that performed the receipted service, quantum by quantum. Work is the tap for both instruments, INDX-Q per quantum of service and INDX-G as a drizzle alongside it. There is no presale, no allocation, no faucet, no treasury, and no premine. Holding is a separate question from minting: both instruments are ordinary transferable assets once they exist, so either can be acquired from someone who earned it, and governance can be bought on the open market. The protocol is never the seller. A network that wants these readings runs nodes to afford them.
The methodology is still being written. The people who argue with it now shape what the readings mean. If you run infrastructure, maintain an oracle, or have been burned by a number you could not check, this is the window.