Every so often, a statement will be made which puts everything into perspective. An excellent example of this is this comment from the Gnosis AMA on Reddit a few months ago, made by Martin Köppelmann (the co-founder and CEO of Gnosis):
“Augur could not call itself a decentralized prediction market if they would control >50% of their tokens. However — Gnosis is decentralized regardless of the distribution of tokens. Even if we would keep 100% of the tokens and not sell any we could not interfere with markets on Gnosis. Gnosis is decentralized because it runs on Ethereum.”
If this quote seems like a boldfaced dismissal of a serious concern, that’s because that’s exactly what it is. Just because something “runs on Ethereum” does not automatically make it meaningfully decentralized (in fact, if that were the only criteria that needed to be met, then Augur would qualify just as well as Gnosis). Yet here we have the creator himself saying that he would be perfectly fine calling Gnosis “decentralized” even if he personally controlled 100% of the tokens and no one else were able to use the system in any way whatsoever.
The worst part is, the above claim is false. The distribution of GNO tokens matters very much in the Gnosis protocol, and if the majority of the tokens are concentrated into too few hands, it introduces critical vulnerabilities into the system (and completely prevents the model from being sustainable in the long term).
Let’s dive into why.
Poisonous Token Incentives and Perpetual Profit Machines
Gnosis is designed so that users of the platform can pay fees in three different tokens: WIZ, BTC, or ETH. WIZ tokens are Gnosis-specific, and are pegged to represent $1 of value (though this peg isn’t sustainable in Gnosis, which we’ll explain further momentarily). A fee payment with WIZ is the most straightforward of the three options, and these fees would not be critically vulnerable in Gnosis, because they are sent directly from a user to a service provider. This is why we felt comfortable adopting the dual-token architecture that Gnosis championed (although our equivalent of “WIZ” tokens are called “PHI” — pronounced “fee” — which are Delphi-specific but otherwise identical to WIZ in terms of purpose and functionality).
However, things get much more complicated and big problems arise if users ever try to pay their fees in BTC or ETH. Gnosis themselves take a moment to acknowledge a subset of the risks here in their whitepaper:
“It is expected that WIZ will be the overwhelmingly predominant method for paying fees in the Gnosis ecosystem. In the unexpected event that this is not true, and users are paying in BTC or ETH, the platform may become vulnerable… [which] may logically cause erosion of the Gnosis userbase, subsequently triggering justified loss of developer confidence…”
The team has attempted to protect against such risks with complex and unproven new market-parallel fee-bidding mechanisms:
“All fees paid in BTC/ETH/Tokens go to an auction contract outside the control of the Gnosis team. If fees exist in the auction contract, any GNO token holder can submit a bid, bidding their held GNO against some amount of fees contained in the auction contract. If the bid is accepted, the GNO will then enter the auction contract and the user will receive the fees specified. When the user’s GNO enters the auction contract, the fee reduction mechanism will be triggered causing a reduction in fees on Gnosis proportional to the total amount of GNO held in this auction contract. The auction contract is one-way and GNO cannot leave this wallet.”
If you’re starting to feel a little bit confused (or suspicious) at this point, that’s a perfectly natural and reasonable reaction. Gnosis accompanies this description with the following figure, presumably in an attempt to clarify the Goldbergian mechanism:
There are a few problems here, but on a higher-level, it is worth taking a moment to acknowledge the risks of haphazard, indiscriminate approaches like this in general. Chains of such dubious assumptions simply expose the platform to other classes of vulnerabilities (perhaps even more serious than those originally in question). It seems that Gnosis might suspect this is the case, too, because as it turns out, they are ultimately assuming (and hoping) that the mechanisms do not end up being widely used:
“NOTE: It is unlikely that this mechanism will be used as game theory and expectations point to users predominantly paying fees in WIZ. In the event this mechanism is triggered, we expect the occurrence to be extremely rare.”
Even assuming that these game-theoretical assumptions of Gnosis (which are certainly subject to considerable doubt) manage to hold true, and assuming furthermore that the complex and currently-underspecified technical scheme for their new fee-bidding mechanism manages to be implemented devoid of bugs or errors, there is still a problem here: non-WIZ fees can always be poached by someone controlling enough GNO. With a monopoly on the GNO supply, no other system participants can realistically bid on the “locked fees” of those users who paid via BTC or ETH. This represents a persistent risk vector for any such payments, which should ultimately be expected to manifest in higher market prices being charged by service providers in the network. Intelligent users who recognize this basic fact would rationally choose to use less-expensive competing platforms, like Delphi or Augur, meaning that this phenomenon would act as a continual deterrent against using Gnosis at all. §3.3.2.1–3.3.2.3 of our whitepaper expounds on this subject in much more detail, for those who are interested in reading further.
As if that weren’t bad enough already, there’s actually another problem here which reveals a profound economic oversight at Gnosis. In their system, WIZ tokens are supposed to be pegged to $1 of value, and GNO tokens are able to generate WIZ tokens via a token-locking mechanism. While this can theoretically work in a sufficiently competitive market where the GNO tokens are meaningfully distributed (which would allow true price discovery of the generator tokens to take place), it is fundamentally absurd to argue that this represents a sustainable arrangement in contexts where the majority of GNO tokens are concentrated into a very small number of hands. If such a system could possibly work, then Gnosis would have invented a “perpetual profit machine” which allows them to print themselves money at will, regardless of whatever else is happening in the market at the time.
If someone is trying to convince you that they have invented a smart contract that can print out infinite free money no matter what, usually it is a good idea to reevaluate whether or not to believe the claims that they are making. Either Gnosis is deliberately deceiving their users, or they didn’t spend enough time evaluating their proposal to spot their logical breakdown; neither possibility inspires much confidence in the future of the project or team.
What Are The Numbers?
So, how much GNO does the Gnosis corporation own, anyway? The answer is shocking:
Even worse, if you factor out just two addresses that bid in the Dutch Auction, and then count what percentage is left for the actual Gnosis community, it is only 2.25%. In other words, the Dutch Auction resulted in almost 98% of the GNO tokens being concentrated into the possession of three total entities:
This can reasonably be termed a “catastrophic distribution failure”, and it means that from the average prospective user’s point of view, it would be economically irrational to use or build on the Gnosis platform: “The game is rigged,” so to speak.
We apologize for repeating information that can be found in our whitepaper, but the takeaway here is that Gnosis relies upon a fair and balanced GNO token distribution, which prompted the team to expend significant resources and effort towards attempting to ensure their token auction resulted in a balanced and well-spread-out issuance profile. However, the results of their token auction ultimately served to demonstrate how egregiously mistaken the team’s assumptions and game theoretical foundations truly were, and the ICO culminated into a radical economic imbalance that jeopardizes the underlying platform’s market integrity in a number of ways. When all is said and done, the GNO token supply lacks any meaningful economic distribution, which unfortunately has profound implications in terms of limiting the ultimate efficacy of the platform. This means that simply by inverting the distribution model that Gnosis employed (granting the developers only 5% and the community 95%), Delphi has managed to eliminate multiple systemic vulnerabilities in one fell swoop.
Token distribution matters, after all. Delphi will be truly decentralized (not just superficially “because it runs on Ethereum” but in actuality, too). We don’t think we’ve invented a perpetual profit machine, but we do think we’ve built a superior alternative to Gnosis, and for us, that’s good enough.
