Editor's note: The dTAO mechanism of Bittensor was originally designed to allocate TAO release amounts more fairly, but issues arose just one month after its launch. The SN28 subnet exploited a loophole in the mechanism, driving TAO releases into meme coin speculation, ultimately leading to centralized intervention by the foundation. As decentralization progresses, similar incidents may be unavoidable, and Bittensor could become a general incentive network rather than an AI project, with the core issue being a lack of unified goals.

The following is the original content (for the sake of readability and understanding, the original content has been edited):

I am someone who is irresistibly drawn to novel token economics. Watching crypto protocols continually adjust their incentive mechanisms can sometimes seem very clever—until they inevitably run into problems. This process itself is quite interesting. So when Bittensor launched the dynamic $TAO (dTAO) system on Valentine’s Day (was this a Valentine’s Day gift from @const_reborn?) I was immediately captivated.

The idea is simple: to provide a new and more "fair" distribution method for releasing higher trade volumes of TAO, allowing each subnet to obtain TAO more reasonably.

But less than a month later, problems began to surface. It turns out that a design that seems reasonable does not necessarily survive in the free market.

The Operating Mechanism of dTAO

A brief overview of how dTAO operates:

1. Each subnet has its own subnet token ($SN), existing in the form of a TAO-SN based UniV2 type native liquidity pool (LP). However, the "staking" of TAO for SN here is essentially the same as "exchanging" TAO for SN. The only difference is that you cannot add additional liquidity to the liquidity pool, nor can you directly trade between different SN tokens (for example, SN1 → SN2), but you can exchange indirectly through TAO (SN1 → TAO → SN2).

2. The TAO release amount is allocated based on the SN token price of each subnet. They use the moving average price to smooth out price fluctuations and prevent artificial price manipulation.

3. The release volume of SN tokens is also very high, with a supply cap of 21 million, similar to TAO and BTC. A portion goes into the TAO-SN trading pool, while the remainder is allocated to stakeholders in the subnet (miners, validators, subnet owners).

4. The number of SN tokens entering the TAO-SN trading pool depends on how much SN is needed to "balance" the TAO entering the pool, ensuring that the price of SN remains stable under the TAO valuation while increasing liquidity.

5. However, if according to the calculations above, the number of SN obtained by the subnet exceeds its maximum release amount (which depends on the SN release curve), then the SN release amount will be capped, resulting in an increase in the price of SN when priced in TAO.

The core assumption of the dTAO mechanism

The ( point in the dTAO mechanism is based on a core assumption: subnets with higher market capitalization create more value for the Bittensor network and therefore should receive a higher release amount of TAO.

But the reality is that in the crypto market, the tokens with the highest prices are often those that receive the most attention, are the most hyped, have the strongest marketing, and even exhibit characteristics of a Ponzi scheme. This is why the valuations of L1 public chains and meme coins are always relatively high.

The original intention of mechanism design is good because it assumes that the subnet that truly creates value will buy back SN tokens through generating revenue, thereby driving up the price of SN and obtaining more TAO release. However, I feel that this logic is a bit naive.

meme coin subnet & broken token economy

Before the launch of dTAO, I discussed the obvious flaws in the dTAO token economic model with several crypto analysts—high market capitalization does not equal high revenue, nor does it truly create more value.

But I didn't expect that this theory would be validated by the market so quickly. The way a free market operates is always unexpected.

Just before the dTAO upgrade, an anonymous user took over subnet 281 and directly transformed it into a meme coin subnet named "TAO Accumulation Corporation", abbreviated as "LOL-subnet". This is clearly unrelated to AI.

On the (now deleted) Github page, it used to say... No mining required, just hold the coin - completely turned into a Ponzi scheme.

In the LOL-subnet (subnet 281), miners do not need to run any code, and the validator's scoring mechanism is entirely based on the number of subnet coins held by the miners. The more SN28 coins held, the more TAO released.

The actual situation is: speculators buy SN28 tokens → the price of SN28 rises, the price of SN28 rises → more TAO release is obtained, if the TAO release exceeds the token release limit of the subnet → the price of SN28 continues to rise, the released SN28 tokens are distributed to "miners" according to the holding ratio → the more SN28 held, the more received, in order to obtain more TAO, more people buy SN28 → higher price → the Ponzi cycle continues.

Ultimately, the release of TAO officially flowed into... meme coin! At one point, SN28 even became the 7th ranked subnet by market cap on Bittensor.

Why hasn't SN28 completely taken over Bittensor? Centralization intervened to save Bittensor.

During the rapid expansion of SN28 over a few days, the Opentensor Foundation directly used their root stake to run custom validator code, encouraging everyone to sell SN28. Ultimately, SN28 plummeted by 98% within a few hours and was completely liquidated.

![])https://img.gateio.im/social/moments-ecd6561d9903c5516d01bd5c5751ecb8(

![])https://img.gateio.im/social/moments-dbca16d65ee3da5f5201c12ee7ac27d8(

After the Opentensor Foundation took action, SN28 plummeted by 98%.

Essentially, the foundation acts as a centralized entity, preventing the free market from operating according to the dTAO mechanism. However, they are able to do this because we are still in a transitional period—the release mechanism of TAO is gradually shifting from the old model to the dTAO mechanism.

Old TAO Release Mechanism & Transition to dTAO

Under the old mechanism, the 64 validators holding the most TAO staked in SN0 ("root subnet") can vote to decide the direction of the released amount of TAO.

But this mechanism also has significant problems driven by interests, especially since large validators (such as the Opentensor Foundation, DCG Yuma, Dao5, Polychain, etc.) hold too much power.

For example, potential conflicts of interest include:

They can prioritize allocating TAO to the subnets they invest in or incubate.

We can guide the TAO release volume to our own running validators and receive TAO rewards from the subnet.

These problems have existed for a long time, and dTAO intended to solve this decentralization issue, but the SN28 incident proved that the new mechanism still has significant flaws.

![])https://img.gateio.im/social/moments-7bef9dfbb6ac041bf65e0d64acd35b36(

Moving in the right direction towards Decentralization, but the risks remain.

Breaking away from the old mechanism is indeed the right step towards Decentralization. Although this may mean that the team will lose some control over the TAO release volume, I still recognize that they have chosen a more decentralized reward mechanism.

However, when the SN28 event occurred, the dTAO mechanism had only just been launched for a week, and SN0 (the root subnet) still controlled about 95% of the TAO release volume (as seen in the blue line in the figure below), which allowed the Opentensor Foundation to quickly intervene and prevent the funding pool from continuing to expand.

But the problem is:

About a year later, the power of SN0 will gradually decline to around 20%, at which point it will no longer be able to directly control the majority of TAO release volume.

·If a situation similar to SN28 arises in the future, it is likely that no one will be able to intervene through SN0.

In this case, Bittensor may no longer be a "Decentralization AI" project, but instead become a meme coin incentive network.

![])https://img.gateio.im/social/moments-96deb3b674d220fa9a0c1e464cf2b6d1(

Bittensor is still in the transition period of the release mechanism, with control gradually shifting from the old mechanism (SN0 or "root prop") to the new mechanism (dTAO or "alpha prop").

Not just a meme coin, Bittensor could become a general incentive network.

Even if we assume that in a bear market environment, people will not rush into meme coins, Bittensor still has a high probability of becoming a completely "general incentive network" unrelated to AI.

For example: if someone launches a subnet for decentralized Bitcoin mining (this concept is not new), it can incentivize more efficient BTC mining methods, and then use the mined BTC as a continuous income to repurchase SN tokens, thereby obtaining the TAO release volume.

If this model holds, TAO will transform from a Decentralization AI project into a generalized incentive project, where the release of TAO will no longer be used to promote AI development but will instead become a means of subsidizing various operational expenses (OpEx).

Technically, there's nothing inherently wrong with this, as the Yuma consensus mechanism is designed to reach consensus on "subjectivity" and is not necessarily limited to AI. However, without a clear goal, the entire network may become... meaningless.

Final Chapter: The cracks in the dTAO mechanism have already emerged

The dTAO mechanism has only been online for 1 month, and issues have already emerged.

According to the incentive logic of a free market, if there is no centralization force intervening, Bittensor may no longer be an AI project, but rather an "attention network" dominated by meme coins in the subnet, or it could become a generalized incentive network where various enterprises use TAO to release subsidies for operational costs, without promoting the development of the Bittensor ecosystem.

I believe that Bittensor needs a true "objective function" that allows all subnets to move in the same direction. But the problem is that it is difficult to define an absolute goal in the AI field (AGI?), as we have already seen, even the LLM evaluation framework struggles to be completely fair... This is also why the Yuma consensus mechanism was initially designed to reach a consensus on "subjectivity".

"Tell me the incentive mechanism, and I can tell you the result." Peace!

"Original link"