ImBananas (τ,τ)

Bittensor is a crypto-ponzi; unsustainable nonsense! TAO has no utility or PMF; it is all driven by token inflation: $328M worth of new tokens are printed annually, yet only $15M in annual revenue was generated! Subsidies from holders pay for subnets; economically bankrupt: 🧵 Token inflation is used to give people the illusion of low cost. As the truth is that creating AI models in a "decentralized" way is far more expensive. While offering no additional utility or benefits It is all theatre; subnets are not created as competitive products. They are created simply to exist & extract as much value out of TAO investors as possible Ponzinomics & Extraction: For example, the Pine Analytics data proved that unsubsidized inference on the Chutes subnet would cost up to 3.5x as much as centralized competitors such as Deepseek or TogetherAI! What makes it all so much worse is that token holders pay for these subnets through inflation. Yet, none of the revenue actually flows back to the token holders. The subnet owners get to keep 100% of the revenue! On top of 18% of emissions, just because... That is a borderline scam, extremely profitable for subnet operators, but setting up token investors for extreme loss when the system inevitably collapses As TAO has a 21M supply limit, which might be appealing to ignorant token investors. But also implies that the network will entirely collapse, as it is fundamentally unsustainable, just like BTC Inefficient & Expensive: The problems run even deeper than that, as is the case with most DePin projects that rely on subsidies rather than real-world value accrual: The reason why decentralized computing is so inefficient is that it requires verification & replication. Within a trustless environment, we cannot simply trust the work done by individual nodes. Instead, the work must be replicated multiple times over, introducing extreme inefficiencies. This is not so bad for simple TX's, but for serious, large computing tasks, this becomes a deal breaker This is why 41% of the rewards go to "validators" whose sole task is to verify that the work being done is legitimate! This only adds to the massive inefficiency already introduced by latency within a distributed network. There are several good reasons why AIs are trained in massive data centers with cards equipped with extremely low-latency, high-bandwidth connections. Something TAO is unable to directly compete with in technical & economic terms Product Theatre: In some cases, it is worth paying a premium for decentralization; one example of this is decentralized storage However, this is not the case for the training of AI's, as this is a one-off cost usually carried out by a centralized for-profit organization. As running the AI itself is much cheaper & even achievable by individuals on a single consumer-grade machine... So, what is the incentive for this centralized organization to use a more expensive method? That does not even result in a commercially viable product, due to the lack of scale... The answer is that there is no legitimate incentive! As there are only so many people they can fool into such a ponzinomic scheme, creating an upper bound on the size these subnets can grow to, which is nowhere near what large centralized AI companies can achieve today Conclusion: There is no future in such a bankrupt design! It is all theatre to extract as much as possible; subnets are not created as competitive products. They are created simply to exist & to extract as much value out of token investors as possible! There is much I did not cover in this critique, including "decentralization theatre", modularity, bad governance, perverse incentives & terrible UX. This critique was purely economic, which is bad enough to reject TAO on that basis alone! As value investors, we have to avoid such nonsense. It is not only dangerous from an investment perspective, but it also harms the industry as a whole. The more we prop up nonsense like this, the more difficult it will be for outsiders to take our industry seriously That is why we must speak out, as we care about crypto's ultimate goals. Financial freedom, censorship resistance, privacy & more Reject the nonsense, as the numbers & facts speak for themselves. Crypto already presents us with such a beautiful dream for the future; let's not spoil our opportunity by wasting our energy on half-baked ideas like TAO Crypto deserves better than that. So, help us spread this message far & wide. As the truth will set us free! 🔥

“The bomb he dropped at the end: - New algorithm for training "arbitrarily large" models on Bittensor - Targeting dropping in the next week or two - Positioning to go beyond what Templar did with dense 72B models - Jon: "to actually compete, like if you look at xAI, they're training 10 trillion parameter models now... we need to sort of cross that chasm... I think there's a way to do it” And just like that Templar will be just an old war story..