This proposal and design change needs to be done/decided before the TAO halving in December 2025 and will not be backwards compatible from that point going forward. The current design has issues that will cause today’s subnet to reach their maximum supply in ~10 years instead of ~60 years and thus they won’t have any incentives/emissions to provide. ALPHA halvings will be compressing (get shorter), and asymmetries would compound with each TAO halving.
I think that the issue is alpha halvenings accelerating over time. and I agree that it is an issue.
I dont think syncing the halvenings fixes it… in my analysis (in the paper), you get 9 months more emission through this fix - with other consequences.
Im not saying my solution is perfect - far from it - it will require adaptation, and improvements. But it has no impact today, and the impacts that occur in the future seem (IMO) manageable (redistribution of liquidity)
While it doesn’t fix the asymmetries, the ALPHA halvings aren’t compressing and keep the supplies at 21 million. One of the “asymmetries” that remains though is that ALPHA halvings will take longer for each subnet after successive TAO halvings, because the minting rate will still decrease compared (which isn’t ideal).
Something else to have in mind is that with the excess TAO in mind to keep the injections balanced one may have to inject more ALPHA than the TAO block emission would be, and I am not sure if that means we have to “give up” on the subsidy mechanism that encourages the sum of prices to stay at/above 1.
Thanks for the thoughtful research on each of your proposals @dougsillars@mcjkula
From what I understood, the main drawback of synchronizing TAO & local ALPHA halvings is that total alpha supply will be different across subnets right? Is there any other negative consequence I might have missed?
Ps: the fact that the 1st alpha halving will come much faster is a drawback for subnet participants, that will earn les revenue, but I’m not sure it’s such a problem given the current emission burn rate on so many subnets.
My proposal syncs the injections and emissions with the TAO block emissions, given that ALPHA emits twice of that the supply differs based on TAO that remains to be minted, so 2 * (21,000,000 - Current TAO supply). Thus the OG subnets would have a supply of ~25 million given the when DTAO launched.
Thus all subnets would continue to get injections and emissions for another ~100 years (instead of 9-30 years) until the last TAO halving, with Zeno-Halvings emissions could continue to halve without having to increase the supply of TAO.
That means the only thing that my synchronization proposal changes is the ALPHA supply (where the highest supply of a subnet would be ~25 million ALPHA), all other asymmetries listed in the paper are eliminated across all cohorts. @dougsillars@tuta3689
@dougsillars to clarify my understanding of your proposal, you would not impact the alpha_out emissions to align with the tao halvings, and only mess with alpha_in/tao_in?
After this change, is it correct that the emission curves (how fast alpha tokens are issued) across subnets stay the same for every subnet between the moment they register and the end of issuance, and only the max total supply decrease? Or would some be faster than others?
With the Zeno tail emissions (which is just increasing precision from 9 to 18 orders of magnitude right?), would the max supply be fixed and easy to determine depending on subnet registration time?
Here are my thoughts, as someone that started investigating the topic only recently: this proposal seems to address the main issues while not being overly complex nor having any critical drawback. Losing the 21 million cap for subnets is a major change from a narrative perspective, but is probably much less critical that changing that of TAO. Having a different max supply between subnets doesn’t seem like an issue, as long as the max supply of each one is fixed and straightforward.
Given that this proposal would essentially remove ALPHA halvings, and ALPHA emissions would be based on the TAO block emission (like the ALPHA injections are), all would halve equally (in December ALPHA emission would go down to 0.5). Thus the ALPHA emission would be the same for all subnets, to halve proportionally the same (in sync) with the injections.
The subnets will all approach their maximum supply equally, while the supplies will shrink the later a subnet registers.
(The blue line is a subnet from today, while the other’s are registered 3, 6 and 9 million TAO later)
The maximum supplies without accounting for burn, are predictable through the remaining TAO to be minted (2 * (21,000,000 - Current TAO Supply).
Examples (Current Supply = Max Supply):
At DTAO launch = ~25,000,000 ALPHA
At 10,500,000 TAO = 21,000,000 ALPHA
At 11,500,000 TAO = 19,000,000 ALPHA
At 15,500,000 TAO = 11,000,000 ALPHA
In regards to the Zeno-Halvings, they wouldn’t change the supply, instead it could be implemented so that there would never be a “last TAO halving” and the block reward could drop below 1 RAO → 0.5 RAO through the higher precision (a unit below RAO). Thus new subnets could register even after 100-200 years, would continue to receive proportionally decreasing injections and emissions, and wouldn’t run out of emissions until we would decide to stop (having a last halving).
They although aren’t needed to be considered for the next 90-100 years, so the Zeno-Halvings are rather for the next generation to decide about. (Reason for Zeno-Halvings is that if TAO block emission runs out, then there would be no emissions in 100 years for existing and new subnets).
if we look at my proposal - even subnets registered in the 14th tao halving will emit 21M alpha, and we do not run into OG subnets exceeding 21M alpha.
Few downsides also otherwise meaning that for newer subnets root proportion will take longer (given the lower minting rate from lower injections), longer ALPHA halvings, higher liquidity/price impact and worse liquidations (given increasingly higher ADR).
That against a single “downside” of having different supplies for different subnets, while keeping everything else synchronized/equal across cohorts.
The fix doesn’t do anything about the asymmetries, all it does is slowing down running out of emissions from 10 years so 30-50 years, given it eliminates the high compression from (injections > emissions).
(where we move tao_in and alpha_in to tha alpha_halving when alpha_halving > tao_halving)
* Right now -the excess_tao is awarded to newer subnets (subnets registered at later tao halvings will have significant liquidity issues). We need to work out the math to make sure that the excess is awarded fairly:
If a “new subnet” has higher price/emission than an OG subnet, should it be subsidized? (probably not - but we need to work out the math)
Can a new subnet be subsidized and then get deregistered in a way that benefits the owner? (maybe - but we need to model, and work out the math)
(i am sure there are more complications - but I believe they can be mathed out without significant changes to emission.)
Upsides:
No “we have to do something now before tao halves” urgency. As long as this is implemented in the next 2-3 years (before an OG subnet hits their 2nd alpha halving), all is fine.
No change in alpha supply. All subnets (from OG subnets to subnets that register in the 12th tao halving) will emit 21M alpha tokens.
Emission of alpha is extended at least 24 years beyond today’s mechanism (OG subnets will emit their last alpha in 2034 if we do nothing.)
OG subnets with really large emission pools will “share liquidity” with newer subnets.
Alpha emitted to the pool remains in the same order of magnitude to the alpha emitted to participants.
Only for existing subnets, while newer subnet will start with 1 ALPHA emissions while the injections will be lower due to TAO halvings, so they won’t be the same.
