This is a discussion thread for a candidate parameter change LIP related to inflation in the Livepeer Network.
Background
Inflation on the Livepeer network is adjusted dynamically each round. The protocol targets a certain percentage of the overall supply of LPT staked (50%), and if participation is below that amount, inflation ticks up incentivizing more people to stake through greater rewards. If it’s above that amount, inflation ticks down, ensuring the network isn’t issuing more LPT than necessary. Since the migration to Arbitrum a couple years ago, the participation rate has hovered in the 40-50% range, which has lead to steadily increasing inflation rate on the network.
In annual terms, inflation currently sits at about 25.6%, and since this is distributed to the 44% of stake on the network, the rewards for staking LPT is currently 58% annually. Over 23,000 LPT are available to be claimed by orchestrators and delegators at the moment.
In a purely theoretical sense, there are very few drawbacks to this. Those who are powering the network are increasing their ownership of the network at a faster and faster rate. And those who are sitting on the sidelines and not participating, are seeing their stake in the network diluted. However in practical terms, there are a few negatives that have been pointed out.
Namely, this is a significant amount of perceived value that is being issued out into the market, there are risks of negative perception, those operating on the network (or those who are staking but are not necessarily helping to power a global video and AI infrastructure) may be being “overpaid”, relative to the amount of work that is actually coming in to the network in fees. Without diving too far into speculative details, it presents risk to the perception of LPT, and therefore to the project itself, should inflation rate run away indefinitely.
The disaster scenario is a death-spiral type effect, where so much inflationary LPT is hitting the market, leading to very little perceived value in the token, and therefore less interest/belief in the project, less participation, and onwards down the line. There is so much positive substance behind this project, network, and ecosystem, that we wouldn’t be doing the project justice by allowing that scenario to occur.
The first two steps towards a solution
For discussion, and future LIPs, I’d like to suggest a simple starting point solution.
- In the short term, we vote on a parameter change proposal that changes two params:
targetBondingRate
andinflationChange
. This would have the effect of getting inflation trending downwards from its current rate, instead of increasing every day as it has. - When the inflation rate comes into an acceptable range, likely in a 9-15 month timeframe, we consider a protocol upgrade LIP that introduces two new parameters:
inflationCeiling
andinflationFloor
. This would bound inflation within a range that the community deems healthy and acceptable, without worrying about runaway inflation or inflation hitting zero. It’s hard to say what the ideal range is, but let’s assume it’s somewhere between 2%-10% annual inflation, based upon what we observe in other healthy, fee generating, stake based ecosystems.
In the future, as fees tick up and become a bigger part of orchestrator and delegator income, we can also consider more dramatic changes to the participation incentives in the network. Or even alternative economic models than what Livepeer has been operating under for years. But this proposal aims for a quick and understandable solution that gets things moving in the right direction, without disrupting the expected economic assumptions under which people are participating in the network.
Initial values, proposed values, and impact
What are the current values of targetBondingRate
and inflationChange
, and what should the newly proposed values be?
targetBondingRate
is currently 50% - represented onchain as 500000000.inflationChange
is currently 500. This has the impact of of adjusting the inflation rate each day in one direction or another by 500/1000000000, or .0000005. A small change each day, but it adds up over time significantly.
Part of the purpose of this discussion thread is to capture input and suggestions as to what these values should be adjusted to. But I’ll share some candidates and some modeling so that people can observe the impacts and play around with the values to see what happens for themselves.
As for targetBondingRate
, we should think about setting it to 33%. This is:
- Below the current participation rate of 44% so inflation would start to trend down rather than up.
- Represents an understandable target #, of 1/3rd of LPT staked and participating at any given moment - with the rest available for participation in onramps/offramps, DeFi integrations, and liquidity venues.
- Is justifiable in terms of security provided to the network based on perceived market cap of LPT. When the 50% target was set initially, the market cap was under $10M USD, with the justification that $5M of value at risk would be enough to secure streaming use cases. Given the current market caps, 33% of that value seems to be enough to secure streaming and AI workflows for users of the network.
As for inflationChange
, there is a whole range of possible values above 500 that this could be set to, depending upon how quickly we want to feel the effects of reduced inflation. I’d consider 1000 to be a reasonable candidate value, represented by 2x the current level. View this chart to see the impact of leaving it where it is (1x), doubling it (2x), or tripling it to 1500 (3x). You can see the dates in the future that the network would hit these thresholds of 10%, 5%, and 0% annual inflation, and how many LPT would be in circulation at that time. There are currently 37.9M LPT in circulation.
This chart was pulled from an imperfect spreadsheet that assumes inflation rate rises until the proposal passes, and then drops indefinitely. You can view the sheet here, modeling doubling (2x) the inflationChange
value, and assuming about 33% of newly issued LPT unbonds each round.
You can see that if we do not change this parameter at all (first row), then it will take about 2 years for the annual inflation to trickle down to 10%. Whereas if we double the rate of change, it will take about 1 year to hit that milestone, and another 4 months beyond that to hit 5%. Tripling it has the dramatic effect of trending inflation towards zero in only a little over a year’s time.
The time to consider enacting a further proposal to introduce a ceiling and floor, if that is of interest to the community, would be the time period between hitting 10% and then 0%.
One nice thing about the impact of this potential change is that it will only be felt gradually over time. The day following the proposal there will be almost the same exact # of LPT minted and distributed as the day before. The model shows about 24,600 LPT being issued/day at that theoretical point. 1 month later, it will still be issuing over 24K LPT/day. It will take 5 months before the issuance falls below 20K LPT/day. So there is little risk of this being immediately disruptive to any network participants, or their assumptions of income within the next 6-12 months.
Finally, don’t forget that inflation is distributed only to those who stake. So if you assume a 44% participation rate is maintained, a 10% inflation rate still equates to a 22.7% annual gain relative to stake - a fairly nice incentive that should help continue bootstrapping the hardware on the GPU network for AI and video transcoding.
Counter Arguments or Alternative Paths
There are a number of valid questions or concerns that have come up around this in the past. Let me get ahead of some of the obvious ones in Q&A format.
The network is overpaying to node operators who aren’t transcoding. Don’t we need to fix the whole incentive and participation model, rather than just change a couple inflation parameters?
This has been discussed frequently, and certainly has some merit. Many proposed solutions run into challenges however around identifying “real” vs “fake” work intended to game the rewards systems. It’s also been recognized that non-performing-nodes aren’t really harming the network significantly, and would themselves be better off as delegators supporting other active nodes. Dramatic changes to the economic model would likely require major protocol upgrades - essentially a Livepeer 3.0. That’s certainly possible, but I think it should be addressed outside the small scope of this proposal - potentially via public goods funding via a protocol R&D SPE.
What about less issuance through a change to the inflation curve?
There’s an interesting case going on in Solana right now, where they feel the inflation is overpaying the network operators, and are proposing an updated issuance formula. I actually like that formula and think it’s reasonable. But it would present an “overnight shock” to the Livepeer ecosystem - where rewards would dramatically fall immediately as the proposal was implemented. We can of course discuss and consider whether a dramatic change is welcome, but I like the idea of a very gradual change over time without any jarring impact that violates anyone’s assumptions.
Why is Livepeer changing the rewards on us?
It’s important to point out that a change like this can only occur through the years-old LIP process which outlines community governance of the Livepeer protocol. It will be discussed, proposed, and voted on by the tokenholders. Only if it passes will it take effect. Bitcoin has set a standard in the space for immutability, which is important in the case of something trying to be “global money” - people need to trust that the economic policy won’t change. But Livepeer isn’t trying to be global money. It’s trying to be the world’s open video infrastructure - and the governance process is set up to enable flexible iteration of the protocol in service of this mission.
What happens if inflation hits 0%?
I suggest that the community consider adding an inflationFloor
parameter prior to this happening, so that they can determine the lowest acceptable inflation value. Perhaps 2%? That would be enacted in a different LIP, perhaps 6-12 months out.
What happens if participation rate drops below the new target?
This is a good question. If participation rate drops below the new target (33% suggested as a candidate), inflation would continue to rise once again. Either the community would determine this is ok, and all that very high reward would be distributed to those powering the network until participation rate rises again - or it might disprove the target participation rate experiment all together. The latter would lead us to consider adopting a more predictable inflation decline over time, such as the Solana example mentioned above, or used within other networks.
What’s next?
Please share any and all feedback here in the thread. We can discuss in an upcoming water cooler chat. And when it feels like all voices have been heard, we can move to an LIP process which includes the draft, last call, and proposal phases.