Proof-of-Governance as the Endgame for LSTs

Research
21

Great post – think this approach mitigates a lot of issues we’ve seen with PoS. Couple questions on implementation details:

  1. Will quorum weights (and thus inflation incentives) be fixed per operator across the valset?
  • Assuming yes, probably should mention the potential for CR improvements as measured by increasing the nakamoto coefficient (along with the more evenly distributed economics that come with it).

1a) Downstream of that, how do you decide what size the validator set needs to be in order for the super minority number to be considered sufficiently censorship resistant?

  • This is super subjective and a moving target, but imo demand side considerations should be weighted against operator profitability requirements. That said, it seems like the number could be much higher than current (6) but also should be much lower 33 if current rollup operators are happy with status quo (ie don’t need to copy/paste the existing set and making their quorum weights equal). This matters as it relates to what the updated emissions rate will be ofc, 0.25% might be enough for 10-15 operators but not for 33 (I have no idea to be clear).

What do you see as the tradeoffs in making the valset social consensus gated?

  • ie if Blackrock wants to run a rollup but would feel better if they also ran a validator for risk mgmt purposes, it doesn’t feel great asking them to write a post in some offchain governance forum vs having the set gated by collateral requirements (ie they can buy TIA or get a delegation from the foundation). Practically speaking, do you envision the devco/foundation stepping into the social consensus process on their behalf to get them in the validator set?

Last question: how does the demand side of the network get a seat at the decision making table, if/when they need to in the future? (This is not just a Celestia problem to be fair)

  • Lets say rollup adoption takes off (CLOBs on Blobs!), and to make it fun we’ll say it takes off before sticky features are implemented ie issuance for assets beyond TIA on Celestia + zk accounts, so these are by and large rollups with native L2 assets that require very low fees driving network usage. What happens when revenue hungry token holders want to go into profit mode prematurely and ask operator to jack up per-byte blob data fees, but the devco/foundation sides with the demand side in optimizing for long term growth maximization?
  • I think this is the champagne problem that @dankad is pointing to, and I’m not sure there’s a great answer unless there are 1) incentives baked into the protocol for the demand side to run a validator and 2) the valset is not social consensus gated by operators that do not want to share issuance revenue with demand side operators, and side with token holders by default
4 Likes
22

:waving_hand: Thank you so much for sharing this. The vision is super thought-provoking and makes a lot of sense, especially the insight that staking yields become redundant as LSTs push bonding to ~100%.

One point I’m curious about and hoping to understand better from the community is how a fully off-chain validator selection aligns with Celestia’s social consensus philosophy and the goal of credible neutrality.

Would it make sense to keep some lightweight, transparent in-protocol criteria (like a scorecard for uptime/performance) to prevent centralization risks, while still keeping issuance minimal and removing redundant stake yields?

I’m just exploring how to keep this aligned with Celestia’s social layer values. Would love to hear others’ thoughts before I think more concretely about it.

Thanks again for pushing this discussion forward. Really appreciate it!

23

Yes, that was the expectation. And yes, this would increase the Nakamoto coefficient (in fact, it would make the Nakamoto coefficient optimal given Tendermint as a consensus protocol).

That’s a good question. Not sure I have a good answer for that for a single validator set. However, I believe that censorship-resistance is better achieved through Multiple Concurrent Proposers, rather than just a large validator set, so perhaps the question is moot.

Don’t see any reason why anyone would want to run a validator for this reason. A full node sure, but a validator doesn’t really add anything. No, I don’t envision Labs or Foundation stepping in in such a situation (speaking for myself, not either entity).

I think this is the correct conclusion. Why don’t Ethereum validator jack up the fees today? It would be in their short-term rational best interest. Turns out they don’t, because validators and token holders have longer-term views.

1 Like
24

Yes, I suggest that possibility here:

It could work the same way jailing for downtime works today. That doesn’t have to change, because jailing for downtime is a consensus thing, not a staking thing.

25

Thanks for the clarification. To build on that, I wanted to share a draft idea I’ve been thinking. It is basically a “middle path” that keeps the insight you have shared.

I definitely love the idea that we can drop issuance to stakers to basically zero and rely on validators earning fees + future income. But removing staking and choosing validators fully off-chain or hardcoding them feels like it might lean too much on opaque processes over time. (Please do educate me if I misunderstood it).

I know we have consensus-level downtime jailing (which you clarified), but I wonder…

  • Could we keep that and add a super simple, open validator scorecard like uptime, inclusion, maybe infra diversity… so the community has a clear, shared signal?
  • And maybe tie issuance more directly to DA usage, so it scales naturally with demand?
  • And still have a protocol-native LST for capital efficiency, but with no governance power… so we avoid LST monopolies messing with selection (as you described)

The goal would be to to keep it transparent and merit-based while still hitting your original goal of minimum issuance.

So, the rough sketch I’m imagining is

  • Issuance only goes to validators, not stakers. And the amount adjusts automatically based on DA activity (may be ~0.25–5% band. Instead of fixed).
  • Consensus jailing for downtime stays exactly as today.
  • Optional scorecard for extra signals (uptime %, inclusion %, maybe community check-ins).
  • Protocol-native LST so people get capital efficiency without governance power attached.
  • Most fees burned, rest rebated to validators, node runners, or other contributors.

So basically, your core idea + more explicit transparency + usage-linked economics.

I’m genuinely not trying to push an alternative for the sake of it. :folded_hands: I’m still pretty new around here and I am just sharing where my head’s at in case it sparks ideas or helps refine PoG further.

If this direction sounds useful, I’m happy to write it up a bit more and can help polish the details together. Or if you think it overcomplicates things, totally fair.

Thanks so much for reading my wall of text :sweat_smile:

1 Like
26

Right, but Lido and Coinbase are actually very conscientious operators. Now imagine the case where a new operator is started completely anonymously and some rumours about ties to Lazarus. Do you think a lot of people would delegate to them? I don’t think they would, given slashing penalties. So people do use out of band operation to make this decision. The process is far from perfect, but it probably does work.

Yes, but the incentivization is global, not local, so it suffers more from the tragedy of the commons problem. If Lido went rogue, specifically those who “voted” for Lido by delegating their stake would be punished. If PoG had chosen Lido as one of the operators, it would not fall back on voters in the same way (they may lose some reputation but not their coins).

1 Like
27

This is a very pragmatic piece that we believe could mend some of the structural supply issues burdening Celestia, while also creating a clear path to accelerate the chain’s development.

From a business case, since Q1 2024 to Q2 2025 Celestia Validator Commission on TIA Issuance and Transaction Fees topped $120m, while across the same time REV has only garnered a grand total of $1.3m, 50% of which originated in Q1 24’ and was non-blob fee related.

1600×233 126 KB
Source: Blockworks Research

When annualizing the latest REV from Q2 of this year, you get 114,106 TIA. Assuming annual inflation of 0.25% on a 1.13bn supply, that’s 2.82m annually. Therefore, REV would need to ~24x from where it is currently for the token supply to go net-deflationary. We’ve outlined other REV growth multiples in the table below, all the way to 50x REV which would lead to a net deflationary rate of -0.25% annually.

REV growth × Tokens minted Tokens burned Net token change Net supply growth rate
1 2,825,000 114,106 2,710,894 0.24%
2 2,825,000 228,212 2,596,788 0.23%
3 2,825,000 342,318 2,482,682 0.22%
4 2,825,000 456,424 2,368,576 0.21%
5 2,825,000 570,530 2,254,470 0.20%
6 2,825,000 684,636 2,140,364 0.19%
7 2,825,000 798,742 2,026,258 0.18%
8 2,825,000 912,848 1,912,152 0.17%
9 2,825,000 1,026,954 1,798,046 0.16%
10 2,825,000 1,141,060 1,683,940 0.15%
11 2,825,000 1,255,166 1,569,834 0.14%
12 2,825,000 1,369,272 1,455,728 0.13%
13 2,825,000 1,483,378 1,341,622 0.12%
14 2,825,000 1,597,484 1,227,516 0.11%
15 2,825,000 1,711,590 1,113,410 0.10%
16 2,825,000 1,825,696 999,304 0.09%
17 2,825,000 1,939,802 885,198 0.08%
18 2,825,000 2,053,908 771,092 0.07%
19 2,825,000 2,168,014 656,986 0.06%
20 2,825,000 2,282,120 542,880 0.05%
21 2,825,000 2,396,226 428,774 0.04%
22 2,825,000 2,510,332 314,668 0.03%
23 2,825,000 2,624,438 200,562 0.02%
24 2,825,000 2,738,544 86,456 0.01%
25 2,825,000 2,852,650 -27,650 0.00%
26 2,825,000 2,966,756 -141,756 -0.01%
27 2,825,000 3,080,862 -255,862 -0.02%
28 2,825,000 3,194,968 -369,968 -0.03%
29 2,825,000 3,309,074 -484,074 -0.04%
30 2,825,000 3,423,180 -598,180 -0.05%
31 2,825,000 3,537,286 -712,286 -0.06%
32 2,825,000 3,651,392 -826,392 -0.07%
33 2,825,000 3,765,498 -940,498 -0.08%
34 2,825,000 3,879,604 -1,054,604 -0.09%
35 2,825,000 3,993,710 -1,168,710 -0.10%
36 2,825,000 4,107,816 -1,282,816 -0.11%
37 2,825,000 4,221,922 -1,396,922 -0.12%
38 2,825,000 4,336,028 -1,511,028 -0.13%
39 2,825,000 4,450,134 -1,625,134 -0.14%
40 2,825,000 4,564,240 -1,739,240 -0.15%
41 2,825,000 4,678,346 -1,853,346 -0.16%
42 2,825,000 4,792,452 -1,967,452 -0.17%
43 2,825,000 4,906,558 -2,081,558 -0.18%
44 2,825,000 5,020,664 -2,195,664 -0.19%
45 2,825,000 5,134,770 -2,309,770 -0.20%
46 2,825,000 5,248,876 -2,423,876 -0.21%
47 2,825,000 5,362,982 -2,537,982 -0.22%
48 2,825,000 5,477,088 -2,652,088 -0.23%
49 2,825,000 5,591,194 -2,766,194 -0.24%
50 2,825,000 5,705,300 -2,880,300 -0.25%
24.76 2,825,000 2,825,000 0 0.00%

Further, assuming all new issuance goes to validators, this would translate to 28,250 TIA per validator on an annual basis - this assumes that new issuance is split evenly and the validator count remains at 100. In dollar terms using the current price of $1.58, that is $44,635. If you view validator / operator commissions as a cost to the network, this would reduce that cost drastically, as in Q2 25’ alone operators were paid $6.4m (44% higher than the proposed annual issuance proposed in OP).

Token Supply 1,130,000,000
Inflation 0.25%
Amount inflation 2,825,000
TIA Price $1.58
Dollar Inflation $4,463,500
Validator Count 100
$ Per Validator $44,635

In our point of view, there is very little reason or rationale to justify the inflation that Celestia currently has. While CIP-29 is a step in the right direction, it still does not address the core question of why the network needs to issue new tokens at all. Additionally, REV would have to increase by ~495x to become net deflationary at 5% inflation. Even at the terminal inflation target rate of 1.5% in CIP-29 which would theoretically be reached in year 14, REV would still have to scale 149x to turn deflationary.

For added perspective, given Celestia does not have a native execution environment, we can assume non-blob fees will not be a forward looking driver of network REV, and the focus will be on blob fees. If you take Celestia’s best ever week for Blob fees which was 5,682 TIA, annualized it comes out to 295k TIA. This would only account for 10% of the newly proposed issuance. To get a more clear, realistic view point on how the community, investors, and other parties can underwrite their respective thesis in the network, it’s also important to understand how much DA usage is subsidized via grants, which at this time is unclear.

In summary, we think this approach is not only appropriate, but necessary in order for Celestia to realize its goals. We also believe offchain governance can lead to faster results, but transparency on decision making should be made public - whether it be open calls anyone can join, or transcriptions of meetings etc. From the validator side, we believe we can provide rational decision making grounded in reality to help steer the network in the right direction, to maximize token value accrual via REV, and dominate DA market share. We are one of the fastest growing validators on both Solana, and Babylon, and will be genesis validators of two high-performance networks soon to launch.

Our focus from the research side has always been around protocol mechanics and market dynamics, and we believe we could be a value add validator in the set if the PoG is to take shape.

10 Likes
28

Following up on the question whether validators would still be able to operate sustainably with only 0.25% issuance.

You mentioned that, ideally, validators should be paid in stablecoins.

Are there any forecasts or calculations available regarding the potential payment rate for validators if PoG is implemented? Thanks in advance!

29

I think we all agree that self-delegated stake is not a meme, so if we want to preserve some notion of economic security for the protocol, could we potentially keep self-delegated staking?

Not sure what the incentive would be for validators to self-stake unless it was part of the governance process to be elected or included some additional amount of rewards which might not be worth the cost.

Or does removing delegation but keeping self-staking just put us in the same awkward position as Ethereum?

2 Likes
30

The point of the 0.25% number is it’s the same number that validators get (on average) today. It’s not less. So any questions around if this number is enough also apply today, and are thus orthogonal to this proposal (even if they’re valid questions).

1 Like
31

Thanks for the answer! However, we’d still appreciate further clarification regarding how the proposed changes will affect validator revenue (if available, any supporting calculations, charts, or forecasts outlining the expected validator earnings after PoG implementation).

Additionally, we wanted to share a few concerns:

  1. Centralization of governance: off-chain governance introduces non-transparency and potential centralization in validator selection, deleting one of the core blockchain features.

  2. Reward reduction for validators: reducing issuance to 0.25% while maintaining the current validator set (100 validators), as we see it, would lead to decrease in rewards. 70+ validators from the current set, would see decreased rewards earnings. Only a small number of community validators might benefit from these economic changes, which over time could degrade the overall network quality.

  3. Removal of staking: closing the staking option removes an important mechanism for involving token holders. We believe this could result in the loss of active users, potential mass sell-offs, and a shift in validator incentives. This removes the competitive element between validators, which will negatively impact engaging with users, promoting the network or educational efforts.

  4. Loss of community participation. Without staking, token holders no longer contribute to network security or earn yields, also weakening decentralization and engagement.

  5. Security model: from our point of view, without staking and slashing mechanism, validator accountability is weakened. The proposal suggests replacing slashing with a mechanism that denies fees paid to validators. However, since this process relies on off-chain governance, it potentially introduces risks of manipulation. Therefore, clear requirements and guidelines need to be provided regarding this.

32

quite literally the exact same number of tokens that a validator would get now if they had 1% of the voting power. 80 validators would get more tokens than they get currently. For example, your validator would almost double since it would go from .64% of the voting power to 1%.

off chain might not be the best wording. the gov process will likely be posted on chain, just not in the celestia-app state machine. presumably a rollup etc. Also, all node operators including light clients ultimately choose the valset by choosing a binary.

Each accusation of “centralization” or “validator accountability” or “loss of community pariticipation” etc must be taking into context of the other inevitability, an LST with a large amount of voting power. Would this process be more opaque or more hidden or involve fewer people than an LST w/ a large portion of the voting power?

3 Likes
33

I’m curious what the point of holding TIA would be after staking is disabled. TIA isn’t like a stock that represents a share of Celestia and it’s not like an LST that captures rewards and has a price that directly reflects value outside of speculation. What is a governance token after you remove governance?

It’s not like we’re burning through TIA from fees and in the proposal you say those will be burned anyway, not part of any revenue sharing system.

After reading this I can’t find a reason to hold TIA and worse, it makes me very sceptical of the projects future because where is the revenue coming from for operational expenses if TIA goes to like 1 cent or even 10? It obviously wouldn’t take long before losses set in because the massive sell pressure would crush the price.

My other concerns are that 1) it’s usually the beginning of the end when a team suggest removing power from the holders 2) anytime off-chain is preferred to onchain there’s premeditated corruption gears spinning. Neither of these actions are in the spirit of blockchain that I’ve been apart of since neckbeards dominated the scene.

1 Like
34

we’re not meaningfully burning TIA in fees so far, no. nfa, but most of the time eth and sol inflate more than they burn :slightly_smiling_face: . The advantage here is inflation goes down to .25%, so that’s a lot easier to burn more than what is inflated.

the end game for LSTs is for everyone to use it and not use the native asset. if everyone is being inflated equally, then what’s the point? Some jurisdictions tax those rewards as income, which is further defeats the purpose. This proposal is just speed running Lido’s and Jito’s end game, alongside recognizing that social consensus must cover LST logic that has controls large portion of the voting power.

The gov mech is not proposed here, and similar to Lido would presumably account for all parties. If you believe LSTs will be the dominant form of holding PoS tokens, then this isomorphic. This proposal is simply skipping all the BS in between that. Beyond existing LSTs, celestia light clients will verify all protocol changes and rules, including changes to the validator set.

governance will almost certainly still be posted onchain for verifiability, just not in the celestia-app state machine! Sorry to be the bearer of bad news, but every blockchain inherently relies on social consensus for security against censorship or dishonest majorities.

fwiw everyone here care’s deeply about the mission. assuming malice arbitrarily, on the research forum of all places, is toxic. be well friend, but take that elsewhere.

3 Likes
35

IMO validator set election is too important to the core protocol to not be in the Celestia state machine. What’s the argument for why it shouldn’t be?

36

My understanding of this proposal is that the “allow list” would be hardcoded in the binary.

Who’s in the active set of that allow list would still be in the celestia-app statemachine.

This proposal doesn’t prescribe an allow list selection process, but if one follows the example of Lido, then a lot of difficult to prove onchain properties such as geo location, if they run bridge nodes, have a high bandwidth connection, or other desirable properties could be incorporated in the decision making process.

If we could move the allow list selection outside of celestia-app, we could remove a large portion of the cosmos-sdk and start a migration to a more performant and more provable stack.

Fwiw, I could see a longer term outcome where the allowlist process is proved or enshrined and it is no longer hardcoded. It would still involve the difficult to prove properties listed above, and presumably try to avoid the original downsides from token voting in PoS.

2 Likes
37

good job celestia. :face_blowing_a_kiss:

38

Just to confirm my understanding - am I correct in thinking that the 0.25% issuance will be distributed evenly among the 100 validators?

If so, with the supply of 1,134,118,575 TIA, the annual issuance at 0.25% would be 2,835,296 TIA, and that would be around 236,274 TIA per month.

Assuming an even distribution across 100 validators, that would come out to roughly 2,362 TIA per validator per month.

Please let me know if I’m missing anything here.

Thanks so much!

39

Overall I think this is a very interesting take on staking, validation and security (or the assumption thereof) for POS blockchains.

Some pointers that you make that I think should be considered “truths” at this time:

  • Value leakage to custodians, LSD providers, validators and consortium by high issuance is significant
  • Theoretical impact of inflation is limited to the net difference in staking APR → inflation but practically there is a lot of liquidity impact that should be considered.
  • In delegation models security is largely a meme, the total staked assets is almost irrelevant as only branding and offchain contracts/SLAs are at risk.
  • High staking yield is a drain on any DeFi activities with the same token.
  • offchain governance in a way can be stronger than on-chain governance

However, I am not sure everything in this model works or makes sense.

  • LSTs carry risk, native staking has certain benefits (airdrops, custody support and what not) so assuming 100% bonded rate is ever reached is unlikely. As issuance drops its actually more logical for people to be willing to lend out or just hold the base asset and it happens on mature systems a lot. The bonded rate increases and increases until a point where there is less and less locked/OG holders and it starts decreasing. I dont believe 100% will be reached any time soon. This means staking APR != inflation so there and will remain a real cost/benefit on whether you stake or not.
  • Paying every validator exactly the same will mean there is truly no alignement for them to do any work for the chain besides the bare minimum to remain in the set through governance. Celestia would basically be fully POA as you mention where governance is the authority setter. This is a fine security model (see Noble for example) but in this model validators should really only be infrastructure providers and not expected to hold or perform other duties in the ecosystem. This means other real costs will appear from RPC, bridge nodes, testnet, marketing and so forth. Likely outsourcing these things separately is indeed more cost efficient but its something worth considering. You are reducing the total number of aligned parties on the chain.
  • Governance on a user by user basis is known to be incredibly faulty, even in the OG world there are examples of this (hi Brexit). Governance nowadays is largely a popularity contest between the few active voices and has created turmoil in and outside of Cosmos for a long while. I don’t necesarilly believe that removing validators from voting (although it would be great not to have to be a decider on non-technical matters) is the ultimate solution here.

Nonetheless, This is a concept worth iterating on. It is drastic in many ways but also spot on with regards to the (economic) problems plagueing POS. There is no need to make certain parties rich in other to be succesful, minimizing leakage is always the best option for the chain in the long run.

best,
Ertemann

1 Like
40

Dynamic emissions + low inflation + burn mechanism + keep onchain governance.