ZetaChain is preparing to move ZETA and Anuma to Solana to focus resources on private AI business. Supporting this transformation is a user memory that can be extended across models and applications, as well as a re-established product and token economy around it.
On September 20, ZetaChain’s token holders voted to pass a proposal that determines the future direction: moving ZETA and Banner AI application Anuma to Solana, and gradually closing its own public chain after completing the migration of related assets.
The proposal has a support rate of 99.4% and a participation rate of 58%. The migration has not yet been implemented, and the specific arrangements need to be determined through a second proposal. However, this vote has made it clear that ZetaChain is preparing to end its independent Layer 1 operations and focus its resources on AI business.
A public chain that once hoped to connect to other blockchains is now preparing to move its business to other people's chains. Behind this change, there is an account that needs to be recalculated first.
After All in AI, ZetaChain no longer needs its own public chain
The groundwork for this transformation had already been laid long before the migration proposal emerged.
On June 1 this year, ZetaChain announced a comprehensive shift to AI and gradually stopped its original cross-chain interoperability functions. Since then, Anuma, ZetaChain’s first consumer-oriented AI product, and the private memory layer have become the new business core.
This strategic shift not only changed the track, but also changed the problems that the ZetaChain team needs to solve. Solana migration is a further choice on this transformation path.
In the original business, ZetaChain served the interaction of assets and applications between different blockchains. Its underlying network is an important part of the product, because maintaining its own chain is directly related to developing cross-chain capabilities.
Coming to Anuma's new stage, users of its services are more concerned about whether the model is easy to use, whether the context can be preserved when switching, and who controls the personal information remembered by the AI. Therefore, whether the underlying network still needs to be operated by ZetaChain itself and its contribution to these experiences need to be re-evaluated.
Public chains need to maintain consensus, validator networks and protocol upgrades, while AI products need to continuously improve model calling, memory retrieval and user experience. As a chain based on the Cosmos SDK, ZetaChain needs to handle upstream security bulletins and patches, while coordinating dozens of independent validators for each update.
From the perspective of resource allocation, ZetaChain is redefining the scope of necessary options. The team believes that operating an independent L1 is no longer conducive to promoting private AI business. The former is becoming a fixed task with limited returns and cannot stop investing.
If it is moved to other mainstream public chains, it can reduce the team's burden of maintaining an independent consensus network and focus more investment on products. On the other hand, ZetaChain still needs to be responsible for Anuma's application security, key processing and data flow. Changing a chain can change the division of responsibilities and cannot replace product-level security work.
In addition to its advantages such as high speed and ecological prosperity, Solana's appeal lies in its ready-made agency infrastructure that can shorten the path to product launch. It already has an Agent Registry for AI agents, which provides verifiable identity and reputation records; the x402 ecosystem allows network services to charge per call, allowing agents to pay for access to APIs, data and content.
These capabilities can be connected with ZetaChain's AI business. If an agent wants to do things for users, it needs to know what it is allowed to do, what information it can read, and how to pay for calling other services. ZetaChain hopes to use private memory and application capabilities to connect to this infrastructure, and use Solana's existing identity, wallet and payment facilities to complete the rest.
This also explains why the migration plan goes beyond just moving ZETA across the chain. As long as the original L1 continues to be retained, the dual burden of maintaining the network and operating AI products will still exist. What the complete migration pursues is to gradually integrate technology investment, asset circulation and developer cooperation into the same ecosystem.
The model will iterate and the memory will be left behind.
After understanding the motivation behind ZetaChain's decision, another question needs to be answered: What exactly does Anuma provide that is worthy of the team adjusting the entire business direction for it? The answer lies in a personal memory that persists across models.
As the capabilities of large models advance day by day, it is an irreversible trend for people to rely on AI. For a person who has used AI extensively for a long time, he often needs to choose different models to complete target tasks based on different needs. This requires users to reiterate desired effects, wording preferences, and align task progress.
If this information is saved independently of the specific model, the model can more easily be used as a tool for task selection, and the user's accumulated work context can continue to be used.
The multi-model aggregation application Anuma addresses this problem and makes it a product entrance. It puts multiple models into the same application, allowing users to continue to use existing memories when switching models, saving personal preferences and project background. According to its product design, users can manage these memories instead of reorganizing the context every time they change models.
Taking the widely used writing ability as an example, users can first use one model to organize information, and then use another model to modify the article. If both can use the same topic selection background and writing preferences within the scope of authorization, the cost of switching models will decrease. The information accumulated over a long period of time will not lose its effect when a certain model is replaced.
This provides an idea of competition that is independent of model capability ranking. Anuma does not need to invest huge amounts of money to train its own models and devolve model capabilities. It only needs to prove that it is more convenient and consistent for users to use different models through it.
Anuma, which focuses on aggregation models, has already gained recognition from some users in the early stage. Official data shows that as of September 22, Anuma has created a total of 306,900 wallet accounts and a total of approximately 1.27 million inference requests.
The next test is the user stickiness of these accounts and whether they can continue to attract new users to join. This is closely related to Anuma's own capabilities: the more useful the memory is, the more willing users are to use it; and if the memory often misses key points, references outdated information, or reappears in irrelevant tasks, it will directly damage the experience. Therefore, the competitiveness of this business ultimately depends on the quality of memory and whether users are willing to let it participate in more daily tasks.
In addition to memory, another "privacy" that is not perceptible to users is another ability that must be taken apart and understood. According to Anuma's technical description, the memory uses a local-first storage method, sensitive content is encrypted by a wallet-derived key, and the optional cloud backup uploads ciphertext. The wallet here assumes the role of accessing personal memory and becomes the identity and key tool in AI products.
However, encrypted storage of memory and how the model processes input need to be understood separately. Anuma's official website states that when calling a closed-source model, the current message and related context will still be sent to the service provider, and the service provider may apply its own data retention policy; the private mode uses an open weight model and corresponding inference infrastructure.
In addition to Anuma, there is also a developer business
If these capabilities only serve Anuma, ZetaChain's growth will mainly depend on one application. Only by opening up capabilities to third parties can the scope of this business be further expanded.
Take a company developing a travel assistant as an example. It is good at destination information and itinerary planning, but it also needs to deal with model access, user memory, supplier failover and call costs. These tasks are related to its main business, but if each company does it independently, the cost of capital and time will increase.
If Anuma could provide these capabilities as a service, developers could build products around their businesses. Users may also allow different applications to inherit their existing personal preferences after authorization, without having to create a new profile in each assistant.
In fact, Anuma has opened its development entrance to other applications and agents. Its currently public SDK and documentation provide development tutorials for web chat applications, mobile applications and agents, supporting developers to create applications, configure API accounts, and use capabilities such as model calling, streaming responses, and tool execution.
Further open plans involve the production system Anuma itself is using. An engineering article on September 14 disclosed that Anuma uses Maxim AI’s open source gateway Bifrost to connect different model suppliers. The team is exploring opening the routing layer to make provider failover, privacy model pooling and cost accounting available to other applications and agents, combined with user-granted shared memory.
The commercial value of this plan lies in allowing third parties to call the model and obtain some of the services required to run the model. Even if users are not interested in Anuma's aggregation model, other applications may still have a need for its underlying capabilities.
The more difficult step is to allow memory to cross application boundaries while maintaining clear authorization scope. The same personal memory should have different open scopes to different agents. The travel assistant may need travel preferences, and the work assistant may need project background. Handing over all the information at once is obviously not an ideal default option.
Among the research and development directions disclosed by ZetaChain in September, it includes granting and revoking memory access permissions to applications and agents, writing relevant records to the chain, and calling them through x402 settlement. The longer-term vision is to allow users to organize knowledge and methods into agents and get paid when they are called.
In addition, opening up will also bring new requirements. Revoking authorization can prevent subsequent access, but it cannot automatically recover information that has been received by external services; third-party developers will be concerned about charges, interface stability, and data migration arrangements. Anuma operates its own applications and developer services at the same time, and needs to give partners enough reasons to rely on it for the long term.
This is also something worth observing in ZetaChain’s transformation: it is preparing to withdraw from the operation of an independent public chain, but still hopes to provide basic capabilities for other developers. It’s just that the services provided this time are closer to the specific work in AI applications.
After the migration, ZETA still has to prove its worth
For a blockchain project, the token economic system is the foundation of the building, and business transformation affects the value logic of the token. After moving to Solana, the ZETA token needs to find its place again in the new business system.
According to the official announcement, native ZETA will be converted into SPL tokens on Solana at a ratio of 1:1, and the name, total amount, and original vesting period will remain unchanged. This plan does not involve ZETA that has been issued on Ethereum and BNB Chain, and the staking and reward arrangements after the migration have yet to be clarified.
However, specific implementation will have to wait for the second proposal. The team needs to coordinate the exchange exchange first, and then determine arrangements for snapshots, collection, and chain suspension; the original verification and pledge will continue to run before execution.
These arrangements address the issue of how assets are moved. But the longer-term question is how the tokens will play a role in new businesses. Or to put it more bluntly: Why do users still need ZETA?
Solana's network fees are paid in SOL, which means that ZETA cannot directly continue the positioning of the native public chain Gas token. It relies on application layer usage rules to establish requirements. Existing ZETA Access already provides a path: users lock ZETA and obtain points that can be used for AI calls. The team hopes to expand this access mechanism beyond Anuma as more applications access it, allowing ZETA to serve a wider range of private AI usage scenarios.
But from token lock-up to a sustainable business, there is still an account that needs to be settled. Lock-up will reduce the number of tokens that can be circulated in the market during the lock-up period, but it will not generate operating income. However, each time the model is inferred, there is an actual cost. Whether users are willing to adopt this method for a long time, how points are redeemed, and what income is used to cover reasoning expenditures will all affect whether this mechanism can continue to operate.
Not only that, the increase in third-party applications also needs to be translated into demand for ZETA through specific rules. Does developer access need to hold or lock tokens, how long will they be locked, and how will the fees be allocated? Only when these arrangements are gradually clarified will the relationship between application growth and tokens become clear.
The basis of all this is still whether the product itself can create continued demand for use. Anuma is trying to expand its usage scenarios. The recently tested Nearby function uses personal interests and preferences to find people nearby who may be compatible, trying to extend memories from chats to real life.
Whether it's social discovery or more third-party applications in the future, the team needs to prove that long-term memory can lead to features that users are willing to use again and again. For ZETA, these features must also form a clear and sustainable purpose for the token.
The choice of ZetaChain provides a concrete case for the relationship between public chains and applications: when the product direction changes, the infrastructure that once had to be operated by oneself can also be re-handed over to the external network. Ending the operation of independent L1 can relieve the team of part of the maintenance burden, and also allow the success or failure of this transformation to fall more directly on the product.
Turning off your own chain is just the beginning of this trade-off. Will users continue to use it, will developers be willing to pay for access, and can these needs find a long-term position for ZETA?
These answers truly determine how far ZetaChain can go after shutting down its own chain.