Continuing discussions on Inflation

[b3nnn]

Hey everyone, I wanted to use this post to reinvigorate the inflation discussion led by @dob in this thread earlier this year.

As a member of the Foundation, and as chair of the Capital Markets Advisory board, I think it’s important to keep us moving forward on this as it part of broader perceptions of the Livepeer project, is part of the broader industry focus on ‘fundamentals’, and is a key component of how capital is allocated within our ecosystem.

From previous discussions (and some new ones), it seems there is broad consensus on the need for small and incremental action. I see my role as helping give a little nudge so we take that small but important first step.

The previous draft from Dob got us to the starting line of what a proposal could look like. My personal tldr of the thread was that:

• There was general alignment that we should start taking some action

• There’s alignment on using existing parameters, which avoids risks or delays from new protocol or smart contract work

• But.. the sticking point was whether to do that using targetBondingRate or inflationChange , or both, and how to do it in a principled, risk aware way rather than using something that might feel a bit arbitrary

Reinforcing all of this, during the Livepeer summit Doug and Arunas /@Jonas_Pixelfield completed a hackathon project that both modeled parameter changes and surveyed a sizeable set of Orchestrators and Delegators on their perceptions. A short summary is that:

• Simple modeling shows small parameter changes lead to effects over a fairly long time horizon (in the range of 12+ months to reach something that might be considered major change). This gives ample time to start, observe, and learn and adapt as necessary as we go

• The survey and interviews further reinforced the consensus from Orchestrators and Delegators that they see the need for action, but sometimes struggle to find confidence with any given approach

With all this in mind, I want to share what we plan to do to help the community move forward:

• Firstly, we want to keep discussing the Inflation topic with Orchestrators and Delegators. Two ways to do this include:

  • Completing a short survey to expand on your thoughts about the topic

  • Book a meeting where we can discuss or asnwer any questions in real time

• Secondly, we intend to try to quantify the risks involved with some additional modeling. I’ve asked Andrew from Shtuka Research (who is a member of the Capital Markets Advisory board) to take the lead on this. Andrew is a mathematician with a long career in academic and applied research, who will help quantify the risks of different change scenarios. He’ll also be helping us build out a framework for continual risk monitoring and adjustment in the future, so that we can all have confidence to move forward to voting on any proposed changes.

Hopefully you agree that these goals are a relatively simple way to make that last important push and build on the broad consensus reached so far. This is not a one-and-done topic so we will share a bit more about what the path ahead could look like as we get more information.

I’m going to sign off here so that Andrew can share a bit more about the survey and modeling, and I’d encourage anyone who wants to chat on this topic to reach out to me direct via DM on Discord or by using my calendar link share above.

[awma]

Hi all, Andrew from Shtuka Research here. I’ll be leading the research effort to evaluate possible actions on inflation.

Who are we? Shtuka Research provides quantitative consulting services to DAOs in economic modelling and market design. It’s been operating since 2023. I’ll be leading this project, with quant research support from @eager1999.

Our approach. Our approach to addressing this issue will be to draw a clear line between goals and mechanism. At the risk of stating the obvious, goals are where the community wants to get to, and mechanism is how we get there. We need to avoid getting those conversations jumbled together.

The survey. In this thread, we’re talking about goals. That’s where we need input from the community: you are best placed to know your own interests and vision for where you want Livepeer to be in the future. The intention behind our survey is to elicit those interests in the form of objectives for issuance, staking yield, and participation. We’ve included questions about picking out target ranges for participation, dilution, and yield over the next six months, as well as open-ended questions about vision for the medium to long term. For the yield question, we included some information on yields available across the industry to ground the conversation in the broader market context.

Over the coming fortnight, I’ll also be arranging calls with community members to hear their input and answer questions about our approach. If you want to talk, book a call with me here within the next two weeks:

https://calendar.app.google/MbzkXGP1tDTR5xrL8

Mechanism. Once we’ve got a more complete understanding of where the community wants to be, we move on to optimising the mechanism that gets us there. I’ll share more information about our approach next week, but to summarise briefly:

• We’re trying to get something passed on quite a short timescale, so simplicity is key. We think there’s more work that can be done to overhaul the adjustment mechanism over the coming year, but for now we’re only looking at tweaking things within the existing framework.

• By construction, issuance/yield is controlled by participation rate. But participation can also react to yield — indeed, that’s the principle behind the design of the issuance system. We’ll be doing statistical modelling and simulations to forecast how the Livepeer economy will react to our parameter choices and rule out settings that, according to our models, run too high a risk of missing the goals.

• Forecasts lose predictive power the further we try to look ahead. That’s why we need to (a) revisit things in mid-2026 to update our priors, and (b) don’t try to change anything too quickly, so that even if something unexpected happen the system state can’t get too far off course.

[awma]

Hi folks, little update on the work we’ve been doing to evaluate actions on inflation.

Firstly and most importantly: we’ve got a few survey responses in so far but we need more! We’ve got some nice information from the LIP-100 discussion, the Livepeer summit survey, and the first few responses to the current survey, but we’d really love to hear from everyone. If you have opinions on what ought to happen to inflation over the next 6–12 months but haven’t been in touch, please do go and fill in the survey here or set up a call with me or Ben:

• Me: https://calendar.app.google/MbzkXGP1tDTR5xrL8
• Ben: https://calendar.app.google/6hefSncthQRVsfPL9

To recap on possible actions, we’re currently working to model the impact on staking %, LPT emissions, and yield of tuning the targetBondingRate and inflationChange parameters, as discussed in previous threads.

Although the approach of tweaking these parameters to manage emissions may appear ad hoc, in fact our methodology — full details of which will be shared before pushing forward on any changes — places it on a firm empirical footing. Briefly, that methodology is:

1. Based on the responses to the survey and our conversations with the community, capture and communicate our shared interim objectives for LPT emissions and participation.
2. Build a statistical simulation model that takes in parameter tunings for targetBondingRate and inflationChange and spits out (randomised) projections for emissions and bonding rate until mid-2026. We can use these projections to quantify how confident we are that a given tuning will result in hitting those interim objectives.

We’ve been hard at work building the model for (2) over the last couple of weeks. It’s getting there, but we still want to capture more input from everyone before we do final validation of results and start sharing visualisations. Please do get in touch if you haven’t already, or drop any notes here to help continue the discussion. We’ll follow up with more information soon.

[awma]

Quick update on the survey:

We have a good number (25) of results now, so I’ll be closing the survey at midnight UTC on Thursday. That means that the last day to fill out the form is Wednesday (in the morning if you’re in Argentina).

I’ll then synthesise the answers and share a summary here this Friday.

---

Edit: The survey is now closed. And the results are in!

Respondents

A total of 30 respondents filled out the survey. The vast majority of them were Orchestrators. The five respondents who did not self-identify as an Orchestrator all said they were Delegators. Among those, we had the two inc. team members, one builder, and two that only put Delegator.

Most Orchestrators also self-identified as Delegators. Eleven did not.

Role	Count
Orchestrator	25
Delegator	19
Liquid LPT holder	2
Builder/developer	3
Inc. team member	2
Community contributor	1

Participation

Low and critical participation thresholds.

The following figure plots percentages against the number of respondents that, according to their responses, would consider a fall in currentBondingRate below that figure to be a concerning (blue) or critical (red) scenario.

All but two respondents would not find a bonding rate of as low as 40% a cause for concern in and of itself. Just over half said they would be OK with 30%. Opinions about “critically low” bonding rates are more spread out.

Reasons that low participation may be a cause for concern.

Several respondents admitted that they don’t know a definite reason that low participation would be a problem. The respondents that did state a definite reason had the following to say:

• Signals a holder community preoccupied with speculation rather than fundamentals. Lower participation means that a greater weight of LPT holders are skewed towards speculative activity rather than long-term engagement and commitment to the project. A rapid drop would be a signal that delegators are longer committed to the protocol, which is a problem. Too much LPT unstaked could lead to excessive selling pressure.
• Collateralising the service. Less security via stake available for slashing. (Note that slashing is not currently implemented, but it still may function as a signal of commitment to security.)
• Stake liquidity. If participation is too low, small stake movements would change the leaderboard too much, leading to difficulties for Orchestrators.
• Voting power concentration. If few tokens are staked and used in voting compared to those being actively traded, the DAO has limited defence against hostile governance actions.
• Signals generally bearish sentiment. Low participation could be taken as a signal of low interest in LPT-denominated yield.

Note. A significant fraction of respondents answered that low participation is undesirable “because it would lead to high inflation.” Since the intent of the question was to consider the protocol objectives for participation in isolation from the mechanism (adjusting emissions) used to achieve it, we cannot use these answers to establish principles for participation objectives. Next time we do a survey on this topic, we need to find a better wording that makes the intention clear to all respondents!

Reasons that high participation may be a cause for concern.

On the whole, respondents were far less sure about this question. Only 11 respondents answered that high participation is a problem, and even some of those seemed rather unsure of the conclusion.

• Liquidity crunch. Too large a part of LPT supply locked in stake would negatively impact liquidity, leading to price volatility and creating a barrier to entry for new entrants to accumulate stake. This was the most common reasoning, with eight respondents arguing this view.
• High participation does not guarantee high decentralization. In other words, the supposed benefits of high participation diminishes at high numbers.
• Optics. May resemble an unsustainable financial scheme where the only thing to do with the token is lock funds with expectancy of profit.

A few respondents gave numbers. The lowest number was 60%. Many answered 70, 80, or 90%. Since high participation does not appear to be a pressing concern, I didn’t spend much time analysing through these figures.

Collected additional remarks about bonding.

• I think the critical indicator is in the decentralization of token holders bonding, rather than the total tokens bonded.
• A system should be put in place to prevent overly abrupt upward or downward movements, and perhaps establish minimum and maximum thresholds.

Yield

What types of opportunity should be considered peers for the purpose of comparison with LPT staking?

Probably the most enlightening data to come out of this section are the responses to this question.

Category	Count
DePIN projects with staking	26
POS blockchains	4
DeFi products	3
DeFi products excl. stablecoin	2
Staking tokens with similar market cap	15
Similar TVL	6
Similar risk profile	14
Other: Similar tokenomics	2

Key observations:

• By far the most popular category for comparison was DePIN projects with staking.
• DeFi products were not high on people’s lists for peer comparison.
• Two respondents called for comparison based on similar tokenomics. One argued that judging whether another project was suitable for peer comparison “really depends on the tokenomics/mechanism design of the protocol.” The other was more specific, asking that we compare with “fee generating protocols where delegator earnings are a combination of fees + inflationary token.”

Should Livepeer aim to reduce staking yield in order to improve the perception of LPT vis-à-vis its competitors? How close to that of its nearest competitor should Livepeer staking yield be in mid 2026? What about end of 2027?

The community is divided on this issue. The majority (22 of 29 respondents) did call for yield to be brought down. Twenty respondents wrote a number for mid 2026, and 18 wrote a number (11) or something along the lines of “close to nearest competitors” (7) for end of 2027. Given the difficulty of identifying which projects ought to be considered “peer competitors,” it is hard to assign concrete numeric targets for yield based on these responses. Nonetheless, they give us indications of the thinking of community members about how (or whether) Livepeer should compare its yield to that of other protocols.

Judging by the respondents (20) who gave a number, those who call for yield to be brought down believe it should be brought down sharply — many called for it to fall to within batting range of its best peer competitors already by mid-2026. For example, combining the highest numerical response (3x nearest competitor) with the upper limit of the range of yields quoted in the question for other DePIN projects would set the mid 2026 target yield at 45%, far lower than the rate of over 65% we see today. If we adopt a more conservative picture of DePIN yields available elsewhere, we are led to interpret many of the answers as calling for yields of well under 30%. (Due to the complexity of the question, these calculations ought to be taken with a grain of salt.)

A significant minority were either unsure (3) or entirely against (5) the idea that bringing down yield is a goal at all. (We identified a respondent as entirely against the idea of lowering yield as a target if they expressed a negative opinion about it and left the numeric target questions blank.)

What were the main objections to lowering yield as an objective? Subject to varying levels of paraphrasing and aggregation, here are some of the common themes that came up in the responses:

• The system is functioning as intended. We should let the market forces do their work, and emissions will come down naturally with increased participation.
• High yield can be a valid part of a financial or GTM strategy. It is not necessarily a problem in and of itself.
• Setting a specific numeric target is wrong. We should ask more fundamental questions such as how many LPT is the protocol “paying” to whom for what?
• Would a significant drop in the yield rate really guarantee a meaningful increase in the token’s market value?
• This question/section seems to be treating LPT as a currency. For as long as I’ve been involved in this project, it’s been stressed that LPT is a work token.
• This question/section considers only “nominal” yield and does not take into account dilution. LPT rewards are not directly comparable with stable-denominated DeFi yields that are not subject to dilution.

Budget

As a percentage of market cap, how much should Livepeer aim to spend on across-the-board stake subsidy in 2026?

The answers to this question were quite wide-ranging, and even the least objectionable figures were considered too low by some respondents and too high by others. Given the proximity of the question to complex issues of targeted redistribution and budget, the diversity of responses is unsurprising.

In terms of minimising the absolute number of objections, a target range of 10–15% stands out as least unpopular. To achieve this figure in 2026 would require inflation to come down rather sharply. For example, if emissions were to trend down in every single round from now until the end of next year, the rate of change would have to be increased by a factor of at least 2.7x, corresponding to a setting of inflationChange = 1350, to achieve a net annual emissions of less than 15%. Of course, with any setting of targetBondingRate close to the current value, it is far from guaranteed that emissions will decrease in every single round.

Other notes about the data:

• We received 24 responses to this question. Five respondents abstained.
• One respondent wrote 75%. If taken literally, it’s unlikely to be possible to satisfy. I dropped it from the data when plotting the graph.
• Three respondents gave a single figure instead of a range. To plot this on the graph, I adopted the assumption that those respondents would not object to a final emissions budget within ±10% of the given figure.

Clarification. The framing used in the question presents the annual emissions rate — the percentage of supply at year end that was newly issued during that year — as an expenditure on subsidising staking. This framing does not capture the full picture of a nuanced topic. It’s perhaps clearer to say, for example, that over the course of 2025, 24% of unstaked supply will have been redistributed, via dilution, to Orchestrators and Delegators. With participation around 50%, that means around 12% of the total supply is being redistributed; equivalent to roughly $24M if priced at todays market cap.

Collected remarks about budget and what it should be spent on.

Respondents expressed a wish for emissions to go to Os that invest into actively providing a high quality transcoding or AI service (3 respondents), or to redirect them to the treasury to fund other programmes (1 response). One respondent asked for more transparency about how treasury funds — which when they were actively being replenished, were funded by emissions — are being used.

• Stake should go to Os that do real work.
• I’d like to have a bit more visibility on how and where the funding has gone and how it’s being used. It kinda gets lost sometimes and not easily trackable.
• Give less of it to delegators / orchestrators, give more of it to treasury and fund innovation.

Summarised general comments about the emissions mechanism as a whole.

• Half of the Livepeer token supply is unbonded and I think we need to ask ourselves why. Why are unstaked LPT holders not delegating despite high yield?
• I don’t like the free money aspect of LPT inflation. I once said that if we ever reach 100M LPT in market cap and we still are unsustainable in ETH fees, maybe there’s just no market for LPT.
• Fees from PMF solves all the problems. We wouldn’t even be talking about this if we had plenty of revenue from fee-paying customers.

[Wiser]

I’d like to add a perspective to the ongoing discussion about inflation.

When we talk about “50% of LPT in circulation,” we usually assume that all minted tokens are actively available. But in practice, some portion of LPT may be permanently lost — similar to how estimates exist for lost BTC.

For example, if we estimate that ~8% of LPT are lost (whether through early mechanisms like Merkle mine or other circumstances), then the effective circulating supply is lower than the nominal one. In that case, “half of the supply” is not really 50%, but closer to 46%.

This suggests that instead of changing broader parameters, it might be more accurate to adjust the baseline calculation of circulating supply. The inflation mechanism could then adapt the threshold percentage accordingly. To avoid disrupting dynamics or alarming investors, such adjustments could be introduced gradually, in several steps.

I believe this approach would refine the model without requiring major structural changes, while also aligning incentives more closely with the real circulating supply.

Please note that I don’t have any particular expertise in finance — I’m just sharing my personal reflection here.

[MikeZupper]

Andrew, thanks for putting together the survey and collecting everyone’s input. When you have a chance, could you share your high-level read on the strongest signals from the responses? A distilled summary from the research side would really help clarify what the community is saying before we discuss mechanisms or next steps.

And if possible, could you also share how those themes are shaping your early thinking about the modeling work?

[awma]

My read on the situation: the survey results show a nearly unanimous view that it’s time for change. The majority of respondents agreed that Livepeer staking yield does not to be so much larger than that of its peers. Not everyone agreed that “optics” was a good reason or that yield should be an objective in and of itself, but among the respondents who pushed back on that framing, most (3 of 5) agreed that controlling the emissions budget is a genuine concern. And in practice, putting the brakes on a runaway budget will involve bringing down yield.

In terms of concrete numbers, one area where we do see a pretty clear consensus is on participation: nearly all respondents would be OK with participation rates to settle into a range between 40 and 50%. So I think there is room to explore proposals that tolerate participation rates in this range. (That’s fortunate, because even leaving things as they are runs a pretty high risk of participation dipping back down below the 50% target.)

If we do introduce a measure to limit the emissions budget for 2026, it’s not totally clear from the survey alone what the objective number should be. Many respondents called for dilution of at most 15% of supply; but based on where we are now, getting there in one year would require a sharp increase of the inflationChange parameter. I don’t see enough evidence the community would align behind a precipitous change right away. A more modest, gradual change looks more likely to work out, at least for the time being. Getting total dilution over the course of 2026 below 20% looks to me like an achievable goal.

One reason it’s a bit tricky to tease a concrete budget figure out of the survey results is that the question really ought to be considered in the context of a broader strategy that also includes decisions about what other things the emissions could be spent on. Although I wanted to avoid making the survey about those decisions, unsurprisingly that (important) topic crept in in some of the answers to the open-ended questions.

A significant subtheme among the asks was:

1. Funding the treasury
2. Redirecting emissions towards “productive” Orchestrators

In practice, solutions to (2) could go via the treasury anyway. So perhaps the next step is to start aligning around an emissions budget that incorporates a plan to redirect some % to the treasury, as has been the case before. Then I think we might start finding it easier to see agreement about the specific numbers.

[awma]

Following up to address @MikeZupper’s second question: the simulation models we’re working on are designed to quantify the risk of falling short of given econometric objectives over some forecast period. Because of the limitations of forecasting far into the future, our forecast period is H1 2026 rather than the whole year.

The survey results inform the choice of objectives. Roughly speaking, we’re currently looking at the following:

1. Keep participation rate above 40%
2. Get dilution below some target figure (specific number as yet undecided).

(Since our forecast period is H1 2026, we’re actually looking at semiannual dilution rather than annual. So the numbers don’t match up straightforwardly to the answers given in the survey.)

We use the model by plugging in various parameter tunings and see if it predicts we hit the configured objectives with high confidence. The community consensus gives us space to maneouvre in order to hit (2) as long as (1) is satisfied. Fortunately, based on the results we have so far it looks like we are pretty confident of hitting (1) under a wide range of tunings — evidence from historic trends in stake movements just doesn’t support forecasting a larger drawdown.

Given that there doesn’t seem to be sufficient alignment for a drastic change, a sensible goal for (2) would probably not be far from the actual dilution seen in the past six months. That figure is about 12.9%, meaning that one LPT held now represents a 12.9% smaller share of supply than it did in late May.