Curation is a real mechanism with real capital in it. It is also, on the evidence of the protocol’s own proposal repository, a mechanism that core researchers tried to abolish.
GIP-0058, “Replacing Bonding Curves with Indexing Fees”, was created on 2023-08-08 by Justin Grana and Howard Heaton. It carries an amendment dated 2023-12-01. As read on 2026-08-30 its stage is Withdrawn.
It is worth an entry precisely because it did not ship. An accepted proposal tells you what the protocol does. A withdrawn one tells you what its designers thought was broken, which is information you cannot get anywhere else.
The charge against curation
The proposal does not hedge. Its executive summary opens: “The Graph’s current curation mechanism is inefficient for several reasons”, and lists three.
Indexers cannot predict their reward. “Subgraphs cost varying amounts to index, making it difficult for Indexers to predict GRT rewards.” An indexer commits real resources without knowing what the work will return, because the return is not a function of the work.
The payment is volatile for reasons the indexer does not control. “Indexer payments are uncertain and volatile due to staking decisions in disparate parts of the network.” Your reward on a subgraph depends on what other indexers stake on other subgraphs. This is the same complaint GIP-0051 made about Cobb-Douglas rebates, in a different part of the protocol.
Signal is a poor proxy for what it is meant to signal. “There is a noisy relationship between curation signal and the quality of indexing service for a subgraph.” Signal is supposed to route capacity toward data worth serving. The proposal says the correlation is weak.
What it proposed instead
A market with prices in it.
- Indexers publicly post a price per unit of subgraph gas.
- A consumer, or an algorithm acting for them, selects indexers at the posted price.
- The selected indexers index, then submit a POI that “verifiably and deterministically states how many units of work it took”.
- The consumer pays price times units of work, secured by on-chain collateral.
The claimed properties follow directly: compensation “depends directly on how resource intensive it is to index a subgraph”, revenue per unit of work is “perfectly predictable and is not volatile”, and the relationship between what a consumer pays and how many indexers they get becomes “perfectly predictable”.
And the deep change, stated plainly: “the incentive to index does not come from indexing rewards, but rather from a direct transfer of GRT from a consumer to an Indexer.” Issuance would have moved to subsidising protocol security rather than incentivising indexing.
That is a different protocol. The proposal calls it “a radical departure from curation” in its own words.
What actually happened
Curation is still here. The deposit charge is 1%, and on Arbitrum the curve is Flat, which addressed the bonding-curve complaint without touching the mechanism.
So one of the three objections was answered by a smaller change, and the other two, the unpredictability of reward and the noisiness of signal, were left standing.
Why this belongs in a library
The protocol’s marketing describes what works. Its proposal repository records what its authors thought did not, in their own words, with their names on it. There are 85 GIPs and several are withdrawn or stalled.
If you want to know where the real problems are, read those.
Before reading on: does this mean curation is a bad mechanism?
No, and the entry deliberately does not say so.
It means curation is a mechanism with documented weaknesses, which is different, and better than the alternative of a mechanism whose weaknesses nobody has written down. A protocol where researchers can publish a detailed argument that a live mechanism holding real capital should be abolished, and have it sit permanently in the public record, is behaving well.
The practical reading for a curator is the one in curation strategy: your return is contingent on a forecast, the relationship between signal and outcome is looser than it looks, and the mechanism is under active reconsideration by the people who built it. None of that says do not curate. All of it says know what you are participating in.