The token VVV of Venice AI reached a new high today, briefly surpassing $25. This rally was triggered by a mathematical research controversy that highlighted the importance of "private reasoning" for many.
New York University mathematician Tristan Buckmaster wrote in a public statement that he and his collaborators had input the entire draft of their project into Codex. After learning that OpenAI’s internal team had also made related progress, he asked whether their model had accessed these conversations or been trained on them. He received a response that the model did not review user data, but his follow-up questions regarding training were unanswered.
OpenAI's response denied that researchers or agents accessed specific user data to solve problems, while acknowledging that, although unlikely, it cannot be ruled out that de-identified data from private conversations between the team and Codex may have helped improve the model.
The community is beginning to question whether, when the stakes are high enough, these labs can see all of your work and beat you to the results.
Trader based16z has shared his trading position, disclosing long exposure to VVV through spot, perpetual futures, and an over-the-counter call option with a strike price of $25. He believes this event has widely highlighted the importance of privacy-focused narratives, and that Venice is the most suitable liquid asset to capture this narrative. He also compares VVV’s circulating market cap to ZEC’s at $50, reflecting his assessment of a revaluation of privacy value.
As AI moves from answering common-sense questions to participating in papers, code, product development, and trading strategies, the content users input becomes increasingly valuable.
Who can let you maintain control over your work content and execution process while using AI? Three Web3 projects have their own answers.
VVV: Turning Privacy into a Business
Erik Voorhees founded Venice AI, which provides consumers with a chat application and developers with an API to access models. It aggregates multiple models, differentiating itself through privacy and fewer restrictions.
Venice AI's privacy features are divided into three levels. Venice's "Anonymous Mode" hides user identity, but the upstream model can still see the request content. "Zero Retention Mode" relies on the service provider to honor its commitments. "TEE Mode" runs inference within a protected hardware environment; "End-to-End Encryption Mode" encrypts data from the user's device until it reaches the protected environment, where it is decrypted.
The business has already reached a significant scale. Banyan, which invested in Venice, disclosed that Venice's annualized revenue increased from $14 million in January to over $100 million in August.
VVV is a token issued by Venice on Base. For every $100 spent on Venice API credits, $5 is used to purchase and burn VVV. On the supply side, the annual issuance of VVV tokens is also decreasing: on September 1, the annual emission dropped from 3 million to 2.5 million tokens, and is scheduled to further reduce to 2 million tokens on October 1.
Another source of demand for VVV comes from DIEM. Holders can lock their staked VVV to mint DIEM; for each DIEM staked, they receive a daily updated $1 API credit. This allows developers and agents to hold an asset that continuously generates usage credits, preparing for future model calls.
On September 14, DIEM’s target supply will complete its phased expansion from 38,000 to 40,000, creating room for additional minting. This move also reflects the Venice team’s optimism regarding user growth.
The bullish case for VVV is straightforward: under the privacy threats posed by AI, Venice has carved out its market position, leading more users to pay for its services and driving increased buybacks; greater demand for continuous inference capacity further boosts staking demand.
NEAR: Verifiable Private Reasoning
Following the Venice AI lead, we also encounter a familiar name: NEAR. This blockchain is expanding the use cases for the NEAR token through confidential computing and agent services.
In March this year, Venice announced its integration with NEAR AI, allowing users to select verifiable privacy inference services provided by NEAR AI. When consumers make requests on Venice, the underlying privacy computation capabilities are supplied by service providers such as NEAR AI.
The core capability of NEAR AI Cloud is to run models within a TEE—Trusted Execution Environment—hardware-isolated and secure. By design, plaintext data during computation is confined to this protected region, making it inaccessible to infrastructure operators; users can verify hardware attestation to confirm that their requests have entered the intended environment. This provides teams seeking to protect research and business data with a secure cloud computing option.
Open-weight models can be deployed in this environment. When invoking proprietary models such as Claude, GPT, or Gemini through the gateway, requests are still handled by upstream service providers; NEAR’s confidential computing ensures that neither the server nor its data can be compromised. NEAR’s value along this path lies in its ability to secure the computation process and manage model services.
This capability has been integrated with the NEAR token. Launched on July 30, staking payments allow token holders to convert their NEAR staking rewards into inference credits, with credit amounts varying based on staking volume, token price, and yield. Users retain ownership of the underlying tokens and can unstake at any time. For teams that frequently invoke models, this adds a new utility to holding NEAR.
Another opportunity for NEAR lies in payments. For an AI agent to complete a task, it must not only invoke models but also purchase data, pay for services, and move assets across different chains. NEAR Intents’ intent-based execution allows users to submit desired outcomes, enabling solvers to compete to fulfill the swaps and executions. This infrastructure enables complex cross-chain operations to be seamlessly integrated into AI agent workflows.
According to DeFiLlama, NEAR Intents generated approximately $9.32 million in total fees during the second quarter of this year, with protocols retaining approximately $1.5 million; the retained revenue is used to repurchase NEAR tokens in the market.
Therefore, NEAR’s bullish narrative is supported by two business use cases: providing computation for sensitive tasks and enabling execution and payments for cross-chain tasks. The former can reach users through applications like Venice, while the latter has the potential to grow alongside AI agents.
TAO: An Open AI Model
Bittensor has been the most anticipated AI project in crypto. TAO is the native token of the Bittensor network.
Bittensor organizes different tasks into independent subnets, where miners provide services such as inference, storage, and prediction, and validators assess service quality; the network distributes rewards according to predefined rules. A team can compete around a specific need, while the entire network supports multiple such markets.
As demand for AI grows, applications require more interchangeable suppliers, and small teams need to find customers, computing resources, and funding. Bittensor attempts to organize this supply through an open incentive market, enabling different teams to compete on specific tasks.
Some subnets have already begun generating external revenue. For example, Chutes, which provides model inference services, earned approximately $1.37 million in revenue during the second quarter of this year, sourced from subscriptions, pay-per-use, and instance services.
How does TAO capture this growth? Each subnet has its own alpha token, paired with TAO to form trading pools. Staking TAO in a specific subnet effectively converts it into the corresponding alpha token, positioning TAO as the foundational asset for capital allocation between subnets. As more competitive services emerge on the network, demand to participate in these markets has the potential to expand.
On the supply side, TAO retains the scarcity design familiar to the crypto market. The total supply of TAO is capped at 21 million, with the first halving completed in December 2025. Currently, 0.5 TAO is issued per block, totaling approximately 3,600 TAO per day.
The surge in VVV has provided us with a window for observation. As models grow more powerful, people entrust them with more sensitive information, making "encrypted/privacy-focused AI" a natural fit for product-market fit.