Dynamic TAO

Comparatively to other decentralized AI related tokens, such as Akash (AKT) and Render (RNDR), Bittensor relates to a broader portion of the decentralized AI stack. For example, while AKT and RNDR relate most strongly to compute, while Bittensor has 32 subnets, which attempt to handle a variety of different parts of the AI stack, from data to finetuning, and pretraining to inference. One issue that arises with owning many parts of the AI stack within the network, as subnets, is that it can lead to dilution for TAO, both in emissions and in attention.

In his Dynamic TAO article, Constance, argued that since the amount of TAO emitted per block will remain the same (1 per block), TAO holders will not be further diluted, which was previously a concern. However, 1 TAO will be distributed amongst all subnets (currently 32) each block based on the TAO and Dynamic Token price curve. In addition to not diluting TAO holders and dramatically increasing TAO emissions, Constance has argued that the Dynamic Token system will incentivize Subnets to "attract demand" to their subnets and Dynamic Token. 

Another concern was that subnet tokens would dilute attention away from TAO, by encouraging users to allocate into subnet tokens instead. The BIT001 proposal addresses this concern to an extent, by designing the subnet token system such that users can exit and enter into Dynamic Tokens only through TAO. However, this system will likely still lead to attention and capital being diluted away from TAO and instead allocated to specific subnet tokens. I view this as an unsolvable problem. At the base of Bittensor is the idea of creating a peer-to-peer intelligence market/digital commodity marketplace. The network wants to incentivize the development and growth of the entire AI stack, from data to inference, inside of Bittensor. In my opinion, this naturally comes with the trade-off of having subnet tokens that will inevitably take away some attention from TAO. TAO holders can still allocate their capital into pools to earn and hold subnet tokens and exit into TAO if they wish. Additionally, as Constance also mentioned, the Dynamic Token system and subnet tokens in general are an ideal tool to incentivize subnet owners. Exiting is important for founders and teams, and to be able to provide the opportunity to exit within the TAO ecosystem can incentivize founders and teams to build on Bittensor (on top of rewards they already earn), instead of on another network, since they can launch a token on Bittensor as well.

As a final note, it’s also important to consider that it was previously difficult to custody TAO. TAO could only be custodied natively through the Bittensor network, which was not supported by hardware wallet providers, or through wTAO, wrapped TAO on Ethereum, which comes with its own trust assumptions. However, TAO has recently been listed for spot trading on Binance, meaning that institutions can access to TAO.

Although Bittensor has some vocal critics, our research recommendation is to closely watch TAO, given that it’s poised to continue to be a narrative leader for decentralized AI. BIT001 is a necessary upgrade to continue to foster and incentivize founders to build on Bittensor, while preventing further dilution of TAO, both in emissions and in attention.