Decentralised governance is a sensitive and unexplored topic of technology, yet many of the current instances slack at the core upon closer examination. This blog acts as a medium for thought on the many instances and experiments of governance in the field of web3.

It is recommended readers compliment each edition with a soothing hot cup of tea of their preference, as the road ahead is dark, uneven, and cold.


Acquisitions are nothing new, still in of themselves require nurturing with - time, capital, and compliance. Dealt in the private domain, shareholders fit the demographic of investor or employee and authority is hierarchical. These actors shape the nature of the deal relative to the counterparty, transferring all rights for monetary gain. Where there were companies, there are now protocols.

Some “fat” facilitate a data layer, others more “thin” populate an application layer within, as described in the fat protocol thesis. The stakeholders in these systems differ, fat protocols rely on nodes to process and constitute the validity of network actions derived from the underlying consensus algorithm. Thin protocols tend to define contributors via a native “token”, often streamlined as incentives to bootstrap participation and contribution, allocating governing power to users and actors alike. These share the topology of a flat organisation, where no mere actor is more significant than the other.

It is common for thin protocols to use a weighted voting model, which has the potential for such structure but only if distributed equitably, if not hierarchies can form. Consensus is sufficiently achieved through fulfilling quorum, which more than often can be a low value to strafe from poor voter turnouts. Fat protocol consensus mechanisms differ in each instance, the native “coin” commonly upholds no authority but the resource committed to secure the network does. An example being Bitcoin’s Proof of Work (PoW), where computational power secures and directs the longevity of the network. Other models like Proof of Stake (PoS) function through a consortium model, where each peer stakes a constant denomination of the native asset. Consensus in each instance is enforced through Byzantine fault tolerance (BFT), and in the presence of foul play, adversaries are exonerated at the cost of their commitment.

Legacy organisations regardless of their top-down formation, exemplify some autonomy in the early stages but eventually diminish upon scaling to corporates. Evaluations either public or private, are derived from an entity’s revenue streams, employee count, share float, and presence. The economics and voting dynamics of company shares, rarely refract from the dividend and weighted model, retaining benchmarks of this asset class consistent. Aspects of which can be abstracted to decentralised protocols, but for the most part are inadequate quantifications of relative value. Native assets across thin and fat protocols, drastically diverge in their economics, utility, and authority.

Publicly traded assets throughout their lifetimes can inhibit value accrual from speculation which can be fuelled by economical theory, but more often roots from social sentiments. Arguing that an asset’s “value” is derived from the free market alone, disregards all qualitative reasoning and is susceptible to market manipulation.

That said, how can one approach deriving relative value across these cooperatives of open collaboration, we know that governance is a necessity at the core of any form of coordination. While monetised votes have always been controversial to the architecture of a governing body, it is nothing new in the world of capitalism. In 1946, a revelation to governance and game theory was formalised by Lionel Penrose. This was then applied upon by John F. Banzhaf the third, to create a methodology distinguishing the probability of a singular actor i in a set of n voters that could skew the outcome of a ballot, known as the Banzhaf power index.

On the 8th of November 2021, discussions were unveiled across the public discourses of Gnosis and xDAI about a potential merger authored by the internal management of each organisation. The basis point for this benchmark relied on 14-day time-weighted average price queries to central markets to compute a pre-defined exchange rate, which was criticised because of GNO’s recent all-time highs in the market versus STAKE’s lows. It was retracted for a model with a multiplier which had more acceptance. A month later, the ballots were live and coordinated through the vote signalling solution Snapshot, enforced with no quorum.

As we walk into a new era of commoditisation and public goods, open initiatives are not refrained by the shackles of top-down droit. Therefore, all acts of founder interaction should digress from conflicts of interest.

For the sake of simplicity, this analysis only constitutes the top 100 EOAs (including multisigs) of a circulating supply and the remainder being represented as a singular coalition representing all minor actors for each distribution. This data concerns the timeline precedent to 16-12-2021, and was sourced from Etherscan, Blockscout, and using a custom BigQuery script to derive STAKE holdings converted to GNO. View the raw data, calculations, and charts here.


One of the first tangible scaling solutions presented to Ethereum, one that did not wish to compete but instead foster the components of its parent network to help negate transactional costs. Achieved through specifying the base currency with the stablecoin DAI, through bridging it from Ethereum. Originally conceptualised as a Proof of Authority (POA) test network for developers, xDAI blossomed into a landscape for creatives and experiments. Incubating innovations such as the real-time zkSNARK strategy game Dark forest, Omnibridge, and the xDAI bridge providing cross-chain solutions with swift finality. Like any iterations today, are facilitated by multi-signature wallets.

The project eventually pivoted to a Delegated Proof of Stake (DPoS) consensus algorithm through issuance and application of the STAKE token. The core utility being its affordance in consensus but also was preached as a governance token. Although when a multisig wallet manages the protocols treasury, it exerts doubts about the governance capacity of the token itself. While Snapshot is a useful tool for mitigating financial costs to partake in governance, it still is a system that depends on the assumption of trust between voters and executors to follow through on the underlying decisions polled. It is an ongoing issue with the widespread adoption of this tooling, let alone the theatre that is usually tallied with it.

It is becoming the norm to hold all ballots through this method begging critical inquiry. Context is important, states of criticality or contentiousness should be balloted on-chain. Furthermore, one should ask, what if the topic is contradictory to such executors - there is no backstop to prevent disobedience nor does a Snapshot poll verify prolonged commitments to any ballot, as it is merely a photograph of balances at a select block height. Hypothetically, if orchestrated subtly, an adversary with insider knowledge could purchase voting power prior to a proposal and then liquidate upon voting, reducing price exposure.

Top 100 addresses holding STAKE across xDAI and Ethereum

While STAKE has pockets of concentration, it is impressive in such a loose analysis to uncover that when the minor actors are aggregated, they act as the second-largest body. It is apparent that this asset’s demographic is vast, as precedent to the buyout it was diversified through a merger with POA network. Consensus here is defined at 51% amongst all participants, as it would be in BFT. There are over 19 different elected cohorts facilitating the role of validators on the network, which one could argue is an improvement in comparison to some larger-scale networks with fewer nodes.

Address STAKE Banzhaf Index
0x5e…6ad7 1372626 0.2444441863631056
REMAINING 1165637 0.1425486309512017
0xe9…241a 378916 0.11162407310432244
0x48…170d 193980 0.04764046071842599
0x96…4b59 191899 0.04710927926530592
0xb0…84af 138356 0.033634227664410724
0xd1…6d2c 131594 0.0319518370363876
0xa9…22ba 121977 0.029569718758782004
0x26…c81d 74997 0.018069970555026467
0x14…e68a 50387 0.012118857528735977
0x24…55b9 47314 0.011377663546395638
0x2c…7e55 46229 0.011116060497923111
0x86…40f9 43978 0.0105734750362733
0x64…e879 42382 0.010188891027342429
0x00…05fa 40000 0.009615082184289463
0x1c…a0e2 39346 0.009457573205846137
0x32…c5c6 37464 0.009004395504897542
0xe1…bd6e 32723 0.007863290142249136
0x0d…92fe 31323 0.007526454959667192
0x9f…aec4 30098 0.0072317690201023094
0x28…5ed1 25697 0.006173384591246561
0x18…bc1e 24865 0.005973351324965386
0xef…f8db 24466 0.005877427384666152
0x5d…413a 23784 0.005713475175518873
0xbc…5ca1 21269 0.005108954441632169
0x3e…03ba 20996 0.005043341972182873
0xbe…950a 20100 0.004828008198809365
0x16…1d0d 18568 0.004459858915724145
0x34…12f7 18009 0.004325537540373836
0x0a…c6fb 17258 0.004145088619893044
0x45…0162 17256 0.004144608075286944
0x67…521b 16865 0.0040506627799887255
0xad…dce9 15487 0.0037195896903293797
0x9c…91b1 14454 0.0034714223575922462
0x64…3ed3 14157 0.0034000738402902336
0x32…ee6b 13127 0.003152644717154841
0xf9…080a 12758 0.0030640057118445055
0x44…8cf9 11711 0.002812509844398831
0x70…41c7 11137 0.002674636433708613
0x34…6ece 10461 0.002512267062842477
0xe9…1cb0 10292 0.0024716753985762185
0x02…246e 10201 0.002449818457696532
0x4f…98c1 10018 0.0024058646177008
0x19…ecf0 10006 0.0024029824092218492
0x65…9d6b 10001 0.0024017814894025163
0xb0…effc 10000 0.0024015413054654733
0xe8…d8a6 10000 0.0024015413054654733
0xee…9131 10000 0.0024015413054654733
0xf8…b52a 9981 0.0023969778123590748
0x4b…4c9c 9975 0.0023955367099422966
0x7c…d7d6 9874 0.0023712782006223318
0xb5…d125 9586 0.0023021059128831996
0x90…986f 9208 0.0022113183605104044
0xfe…1ffb 9000 0.002161361693591051
0xf6…2da7 8878 0.002132060349511339
0xdc…ed97 8704 0.002090270111946481
0xe1…0f53 8527 0.0020477595947344463
0xbb…6dc4 8517 0.002045357877772434
0x46…e642 8454 0.0020302270783846246
0xc3…3443 8275 0.00198723655736115
0x05…ba39 8117 0.0019492898141427134
0xe3…6cac 8003 0.0019219106311174083
0x25…5ec3 7942 0.001907260404790542
0x83…d483 7705 0.0018503409218864973
0x0b…6601 7685 0.0018455376078661094
0x26…916b 7683 0.0018450572766153482
0x45…be99 7631 0.00183256867371887
0xf3…090d 7604 0.001826084214114598
0x17…dcdf 7533 0.0018090325106134763
0x89…c181 7149 0.0017168097930337818
0x14…23d9 6834 0.0016411590509800296
0x91…678b 6792 0.001631072331791892
0x16…6476 6629 0.001591926355801861
0x0c…27e7 6625 0.0015909657203497336
0xf0…e88b 6437 0.0015458159600184956
0xa2…5169 6354 0.0015258828862300077
0x6d…39e0 6274 0.0015066703220361216
0x22…b9ff 6244 0.0014994656197441839
0x44…e480 6073 0.00145839891190472
0xb8…4901 6022 0.0014461509773511922
0x79…e902 5997 0.0014401470929868562
0xdb…4298 5954 0.001429820419711905
0x4b…f762 5690 0.0013664196619794862
0x4d…3591 5595 0.001343605082989271
0xb8…1850 5567 0.0013368807946730257
0x7b…f7a1 5394 0.001295334384800477
0x70…3310 5325 0.0012787638919104724
0x4e…b6ed 5297 0.001272039640408809
0x01…09ef 5284 0.0012689176677607784
0xe4…2431 5261 0.0012633941796460321
0x43…5e5f 5001 0.0012009549192554467
0xed…66c1 5000 0.0012007147688475905
0x3a…e43c 4971 0.00119375040896099
0x15…4924 4881 0.0011721369019841183
0xbd…85a0 4653 0.0011173828410420876
0xfb…c4bc 4466 0.001072475067013981
0xb5…2bbf 4302 0.001033090824181557
0xde…4821 4111 0.00098722272029753
0xa9…1959 4000 0.0009605664496889018
Figure 2: STAKE power distribution

Computing the power distributions shows us that the largest singular actor is actually the treasury (as expected), as it is internally managed through a multisig, therefore, isn’t a piece of infrastructure. The disparity between the top 3 actors and the remainder seems equitable at first glance but is diminished by the restraints of computing the Banzhaf power index for large n player finite games, and the attempt to negate this by clustering all minor actors into a singular coalition.

When we look into the specifics of the proposal, that dictated the future of the network, it seems the turnout wasn’t quite what is deemed as consensus in distributed systems. According to the xDAI proposal; a total of 1,210,318 STAKE out of the circulating amount of 4,991,236 STAKE was only cast to the ballot. A ratio of 1:2.1 to the quorum, which is quite the divergence.

Figure 3: Voter demographic of the STAKE proposal

On light inspection three addresses stand out in this demographic, the major actor 0xe96…0d12 was delegated weights from the multisig address 0xbc…26bb that has direct association to the Gnosis partnership with xDAI and provided the majority of liquidity to GNO/STAKE Uniswap V3 pool. Then 0xFe…b7F4 that purchased approximately $425,000 of STAKE and 0x454…be99 that bought $100,000 a day before the ballot went live.

Not that any of these actions are unethical, this is just the nature of monetised voting but it does show us the sheer influence the counterparty had. From these three addresses alone, they held approximately 30% of all voting weight in the ballot, and this is without a thorough analysis of all actors. The fact that this proposal did not adhere to the standards of fat protocol consensus is a matter of concern.


The owls are not what they seem, some say. Gnosis, originally a prediction market on Ethereum, incepted amidst the gold rush of 2017 and catalysed from a “poorly designed” Dutch auction where 95% of the supply was retained by the internal players. From day one, with 250,000 ETH raised it was a behemoth. Although that amount is measly today compared to the many wallets that Gnosis operates from serial investing and trading - with, of course, the money raised that was supposed to build out a protocol of value. While the premise of prediction markets was endorsed as revolutionary back in the dissonance of 2017, we know today that this was a fallow statement - from an inability to capture traction, the concept being subject to regulatory scrutiny and the delusion of futarchy having extensive benefits to a governing body (more on that later).

Gnosis Ltd, knew this all too well as overcast foreshadowed the markets and eventually were criticised. They quickly pivoted and found their niche in bringing open-source tooling to Ethereum, notably of which was multi-signature wallets. An essential form of tooling to mitigate risk in fund management. The accessibility granted through developing and publicising such tooling has had immense benefits for individuals and open organisations. The question still remained of their token utility, and still seemed to be an ongoing narrative up until the acquisition.

While we could talk all day about the many ventures Gnosis ignited, whether successful or failures, for internal financial gain or public goods. It is worthy to note their resilience. Furthermore, while the organisational structure is less like a protocol and more like a corporation, let’s investigate how their journey towards decentralisation is going.

Figure 4: Top 100 addresses holding GNO on Ethereum

We see a relatively similar topology to xDAI’s stakeholders, the contrast being the remaining minor coalition is only the 5th largest and versus contracts (which is outside the scope of this analysis), are approximately 30% more concentrated. Where quorum is defined at 51%, as if we were to use gnosis’s own value at 4% - it would be infeasible to compute the index due to probability constraints.

Address GNO Banzhaf Index
0x2e…6a13 282543 0.31079299006291866
0x9d…53d5 133364 0.07898748651781151
0xaf…c0d6 75000 0.05124487198407653
0xfc…a2e8 71618 0.04897624251003832
REMAINING 66750 0.04568859217933548
0xa4…46c1 50251 0.03448033366956244
0x1d…a341 50001 0.034309766469972795
0xad…76e3 41168 0.02827310059328169
0x62…3112 38438 0.026404028670435904
0x97…a73b 38230 0.02626156728736583
0xae…d7a9 34951 0.024014807533713222
0xfe…b7f4 31224 0.021459052732673515
0x2d…5cdd 26999 0.018559628052894864
0xb1…c269 25000 0.01718711058280377
0x39…ebdf 23928 0.01645090864231301
0xe9…40e8 19562 0.013451479099828452
0x9e…ca61 16627 0.011434344111027056
0xbc…26bb 16591 0.011409598867211022
0xde…6870 13779 0.00947649192556307
0xfb…bb98 13492 0.009279170462733315
0x94…a3a3 13018 0.008953271939631054
0x95…1dc3 11857 0.008154983270337854
0x57…593e 11111 0.007642012773520306
0xf5…2764 10000 0.006878017735743157
0xa8…0bb6 9320 0.006410384125780234
0x40…3ddd 7756 0.005334770954634757
0x55…f224 6955 0.004783870788886505
0x7d…41b4 6103 0.004197877469347734
0x28…1d60 6016 0.004138039215161724
0x41…4c6b 5946 0.004089893377794901
0x53…efc4 5942 0.004087142184002658
0x8f…3b0c 5797 0.0039874111866138725
0x49…9ec6 5680 0.003906938277956872
0x29…0858 5000 0.003439227290678738
0xe3…0e18 4691 0.003226691185925732
0x0c…56ec 4665 0.003208807816477484
0x81…f1e1 4392 0.003021031785892139
0x67…1c87 4379 0.0030120900411596306
0x00…0497 4207 0.0028937836404990076
0xf0…8461 3922 0.002697751741029259
0x75…8608 3557 0.0024466917361390915
0xec…63fb 3537 0.0024329349743849983
0x44…2a3f 3502 0.0024088606294971077
0x46…26e9 3479 0.002393040337584361
0x64…e5f1 3414 0.0023483307825139083
0xcd…944e 3408 0.0023442037441190728
0x55…da65 3330 0.0022905522064550965
0x7d…9530 3116 0.0021433540413478285
0x42…c62f 3042 0.0020924536241649673
0xea…b9a4 3013 0.00207250614751555
0x1f…2aef 2922 0.002009912284721754
0x5d…6905 2896 0.001992028308371022
0x49…e462 2891 0.001988589081368263
0xfe…ab3e 2722 0.0018723430638347715
0xdd…3544 2455 0.0016886875694740904
0xce…250a 2446 0.001682496923875317
0x57…a4a2 2382 0.0016384745354854299
0x39…04f7 2104 0.0014472519046249308
0xf5…dcb3 2038 0.0014018536403934995
0x12…8034 1991 0.001369524554601292
0xa4…8aa9 1989 0.001368148848481798
0x22…cb32 1947 0.0013392590136376353
0x12…109c 1731 0.0011906825375149863
0x08…9ffb 1660 0.0011418448366306796
0xee…a1e5 1605 0.001104012794799654
0x90…de31 1598 0.001099197806437207
0x65…e6b2 1513 0.0010407300692846717
0xd0…549d 1508 0.0010372907894268864
0xb0…00ec 1424 0.0009795108686037708
0xe5…c948 1411 0.0009705687348340199
0x74…c129 1384 0.0009519966082164706
0x70…87dc 1369 0.0009416787585790292
0x7c…55cb 1321 0.0009086616326170941
0xe6…8ac1 1309 0.0009004073494621078
0xed…c15e 1228 0.0008446909213655568
0x5f…9bef 1219 0.0008385002053739057
0x18…b73a 1205 0.0008288702020404591
0xf7…08a2 1200 0.0008254309149366431
0xd4…9988 1147 0.0007889744653245416
0x1d…b9fe 1093 0.000751830146683271
0x8f…0ec3 1039 0.0007146858169883382
0x8a…fa6f 1026 0.0007057436619533626
0x00…78d8 1005 0.0006912986410144979
0x10…48b9 997 0.0006857957754888846
0xb8…f2c9 980 0.0006741021855144276
0x00…cff3 966 0.0006644721695030913
0x05…fb7d 964 0.0006630964528762203
0x24…8e66 957 0.0006582814445770462
0xea…ec99 953 0.0006555300111901419
0xe9…837b 866 0.0005956863223222633
0x1f…109b 864 0.0005943106050607677
0xf7…4d0c 836 0.0005750505621540512
Figure 5: GNO power distribution

With observation to the remaining minor coalitions across both power distributions, GNO’s minors have approximately 80% less influence in the system compared to STAKE. This value may be saturated due to the discrepancy in computing large finite games, and grouping all minor actors to negate this. Furthermore, if we were to assume that there is a commonality amongst a selection of actors in the demographic; such as multisig addresses with a majority holding in GNO, a different picture emerges.

Figure 6: Top 100 profiled addresses holding GNO on Ethereum
Address GNO Banzhaf Index
0xae…d7a9 34951 0
0xfe…b7f4 31224 0
0x2d…5cdd 26999 0
0x39…ebdf 23928 0
0xbc…26bb 16591 0
0xde…6870 13779 0
0xfb…bb98 13492 0
0x94…a3a3 13018 0
0xf5…2764 10000 0
0xa8…0bb6 9320 0
0x40…3ddd 7756 0
0x55…f224 6955 0
0x7d…41b4 6103 0
0x28…1d60 6016 0
0x41…4c6b 5946 0
0x53…efc4 5942 0
0x8f…3b0c 5797 0
0x49…9ec6 5680 0
0x29…0858 5000 0
0xe3…0e18 4691 0
0x0c…56ec 4665 0
0x81…f1e1 4392 0
0x67…1c87 4379 0
0x00…0497 4207 0
0xf0…8461 3922 0
0x75…8608 3557 0
0xec…63fb 3537 0
0x44…2a3f 3502 0
0x46…26e9 3479 0
0x64…e5f1 3414 0
0xcd…944e 3408 0
0x55…da65 3330 0
0x1f…2aef 2922 0
0x5d…6905 2896 0
0x49…e462 2891 0
0xfe…ab3e 2722 0
0xdd…3544 2455 0
0xce…250a 2446 0
0x57…a4a2 2382 0
0x39…04f7 2104 0
0xf5…dcb3 2038 0
0x12…8034 1991 0
0xa4…8aa9 1989 0
0x22…cb32 1947 0
0x12…109c 1731 0
0x08…9ffb 1660 0
0xee…a1e5 1605 0
0x65…e6b2 1513 0
0xd0…549d 1508 0
0xb0…00ec 1424 0
0x74…c129 1384 0
0x70…87dc 1369 0
0x7c…55cb 1321 0
0xed…c15e 1228 0
0x5f…9bef 1219 0
0xf7…08a2 1200 0
0xd4…9988 1147 0
0x1d…b9fe 1093 0
0x8f…0ec3 1039 0
0x8a…fa6f 1026 0
0x00…78d8 1005 0
0x10…48b9 997 0
0xb8…f2c9 980 0
0x00…cff3 966 0
0x05…fb7d 964 0
0x24…8e66 957 0
0xea…ec99 953 0
0xe9…837b 866 0
0x1f…109b 864 0
0xf7…4d0c 836 0
Figure 7: GNO profiled power distribution

What we see above is this assumed profiled coalition holding all governing power, and every other actor being disenfranchised or more commonly defined as dummy candidates. It goes without mentioning, that this benchmark does not include the unallocated supply that Gnosis Ltd still manages with no strategic plan how to distribute for an equitable governance landscape. Only recently, did they launch “GnosisDAO” which is in fact, yes you guessed right - a multisig. Not only does this constitute the means that GNO is not a governance token at the time of writing but if it was, it would be anything but democratic, one could even argue a dictatorship.


Protocols architected poorly set a bad precedent and standards for the industry, they are paradoxical to the true means of distributed systems and attempt to cash in on the value provided from so. Individual sovereignty is not only compromised but non-existent. These systems should not be designed in the faith of prolific actors, they should be equitable and diverse, or else they are just reflections of the inequitable societal structures we all know too well.

Decentralised autonomous organisations at scale are not multi-signature wallets, they are smart contracts directly governed by all stakeholders. This opens the field to many possibilities and potential attack vectors, yet the benefits of doing so far extend that of a conglomerate. They are agile and open to all actors. Furthermore, open organisations should have robust policies to define consensus providing defence from external; or case in point, internal exploitation.

Huge thanks to Dennis Leech, Professor of Economics at Warick University for his research in deriving more effective computations of power indices, Ruben R. Puentedura who formalised the theory into a python script and Heinrich Hartmann optimising the algorithm with vectorisation. I hope to contribute to this field myself as this blog progresses, upon realising the potential application of parallel programming.