JOSEPH DESOZE

Artificial intelligence is becoming more capable every day. It can study markets, compare opportunities, manage strategies and react to changing conditions faster than any human. The next stage is already taking shape, with AI agents beginning to execute transactions and manage digital assets on behalf of users.
That future sounds efficient, but it also introduces a serious question.
What happens when an AI agent makes the wrong decision while controlling real money?
A bad answer from a chatbot can be ignored. A bad blockchain transaction may be impossible to reverse. If an agent misunderstands its instructions, relies on inaccurate data or interacts with a malicious contract, the damage can happen before a person has time to respond.
Newton Protocol is being developed around this problem.
Its purpose is not to make AI perfect. Instead, Newton aims to place clear and enforceable boundaries around what an AI agent, automated system or asset manager is allowed to do. The agent can continue operating independently, but only within the limits approved by the user.
This is a meaningful difference.
Many AI projects focus on making agents more intelligent. Newton focuses on making their actions more accountable. Intelligence may help an agent identify an opportunity, but authorization determines whether it should be allowed to act on that opportunity.
When Newton Protocol and the NEWT token launched in June 2025, the project was mainly presented as infrastructure for secure and verifiable onchain automation. Its original design included a specialized system for managing user permissions, automated instructions and a registry where developers could publish agent models.
The broader vision also included a marketplace where developers could create, share and potentially monetize automated strategies. Users would be able to discover agents designed for tasks such as recurring purchases, portfolio management, trading and other forms of onchain activity.
That original vision remains part of Newton’s story, but the project has developed into something larger.
In early 2026, Newton updated its whitepaper and began presenting the protocol as an authorization layer for the onchain economy. This wider direction means Newton is no longer limited to AI agents or automated trading. It can potentially evaluate actions initiated by individuals, institutions, applications, vault managers and automated systems.
The change feels natural because AI agents are only one part of a much wider problem.
Blockchains are excellent at settling transactions. They can verify signatures, confirm balances and execute smart contract instructions. What they usually cannot understand is whether a technically valid transaction follows the real rules behind the money.
A transaction may be correctly signed and still violate a company’s spending policy. A vault manager may have the technical ability to allocate capital to a risky market, even if doing so breaks the vault’s published strategy. An AI agent may be authorized to trade tokens but accidentally receive enough access to transfer the entire wallet.
In each case, the transaction may be valid at the blockchain level while still being unauthorized according to the user’s actual intention.
Traditional financial systems perform many checks before approving transactions. Banks and payment companies may examine identity, spending limits, fraud signals, sanctions exposure and internal risk policies. In crypto, these controls are often handled through private servers, website restrictions or manual processes.
These methods can work, but they leave important gaps.
A website restriction can sometimes be bypassed by interacting directly with a smart contract. A centralized compliance server can fail or become a single point of control. An internal policy may exist in a document without any technical mechanism forcing the manager to follow it.
Newton attempts to move these rules closer to the transaction itself.
Before value moves, the proposed action is checked against a policy. If it follows the approved rules, it receives authorization. If it violates those rules, it is blocked before settlement.
Imagine giving an AI agent permission to manage a portfolio.
You may want the agent to trade automatically, but you probably do not want to give it unlimited control. You could allow it to trade approved assets, use selected protocols and spend within a daily limit. You could also require human approval for unusually large transactions.
Newton allows these instructions to become enforceable policies.
When the agent proposes an action, the transaction is checked against the relevant policy. Network operators evaluate the request and determine whether it follows the user’s conditions. Their decision is then turned into a cryptographic attestation.
This attestation works like a verifiable approval receipt. It confirms that a particular transaction was evaluated against a particular policy.
The smart contract checks that approval before executing the transaction. If the request is compliant, it can continue. If the request falls outside the allowed boundaries, it stops.
The agent still has the freedom to perform its assigned task. It simply does not have the freedom to do anything it wants.
This approach is important because intelligent systems can fail in many different ways.
An AI agent may misunderstand a command. It may receive manipulated information through a prompt injection attack. It may depend on inaccurate market data. It may also perform exactly as programmed while operating under permissions that were poorly designed.
Newton cannot prevent every error, but it can reduce how much damage an error is able to cause.
An agent authorized to swap tokens does not need permission to transfer contract ownership. An agent managing a small portfolio should not be able to spend beyond its daily limit. A system designed to use trusted protocols should not be allowed to send funds to an unfamiliar contract.
These restrictions may sound simple, but they represent the kind of practical security that automated finance needs.
The goal is not to remove humans completely. It is to decide where human control must remain and then protect those boundaries.
Newton Protocol reached an important stage when its mainnet beta went live on Ethereum and Base in June 2026. The first major use case focuses on DeFi vault management.
A DeFi vault may hold assets belonging to many depositors while a curator decides where that capital is allocated. The curator may publish a strategy and promise to follow certain risk limits. However, depositors often have to trust that those promises will be respected.
Newton’s VaultKit is designed to turn those promises into enforceable rules.
A vault could set a maximum allocation for any single market. It could block unapproved protocols, restrict risky counterparties and require certain collateral conditions. Every management action would need to pass the policy before reaching the vault.
If the action follows the vault’s mandate, it can proceed. If it breaks the rules, it does not execute.
VaultKit does not take custody of user assets. It is also not a new vault that requires everyone to move their funds. It adds a policy check to the curator’s existing process.
This may be less dramatic than the idea of fully autonomous AI traders, but it addresses a real problem. When people deposit money into a vault, they should not have to depend entirely on the manager’s promise. They should be able to verify that the strategy’s rules are actually being followed.
Newton policies can also use information from outside data providers.
For example, a vault may require collateral to remain above a certain level. To enforce that rule, the policy needs reliable price information. A transaction that involves an unfamiliar wallet may require a risk score. An institutional application may need identity or sanctions information before approving a transfer.
Newton can bring these signals into the policy evaluation process. This allows rules to respond to real market, security and compliance conditions rather than relying only on information already stored inside a smart contract.
However, this also creates an important limitation.
A policy is only as good as the rules and data behind it.
Newton may prove that a transaction followed a specific policy, but it cannot guarantee that the policy was intelligently written. It may verify that a price check occurred, but it cannot turn an unreliable data source into an accurate one.
Cryptographic proof can make a decision verifiable. It cannot automatically make that decision wise.
Human judgment therefore remains essential. Developers must select reliable data providers. Managers must define reasonable limits. Applications must decide which risks are important enough to block.
Newton does not remove responsibility. It makes responsibility easier to see and enforce.
Privacy is another important part of this system.
Some policies may need to check personal or commercially sensitive information. An application could need proof of identity, location or compliance status. Publishing all that information permanently on a public blockchain would create serious privacy risks.
Newton aims to prove that a required check was completed without revealing every private detail used during the process.
Sensitive information can be evaluated inside protected computing environments. The final result can then be supported by cryptographic evidence. This creates a record showing that the policy was followed while keeping the underlying personal data private.
The system can also include challenge periods and financial penalties for operators that produce provably incorrect evaluations. This gives network participants an economic reason to behave honestly.
The NEWT token plays a coordination role within this ecosystem.
NEWT is an Ethereum based token with a fixed total supply of one billion tokens. Its intended functions include operator staking, delegation, payment of policy evaluation fees, network challenges and future governance.
Operators that perform accurate evaluations can receive fees. Participants who produce incorrect or dishonest results may face penalties. Token holders may also be able to delegate their NEWT to operators rather than running infrastructure themselves.
At launch, 60 percent of the total supply was assigned to community related categories, while 40 percent was allocated to internal categories. These included core contributors, early Magic Labs backers and Magic Labs. Internal allocations were placed under lock and vesting conditions intended to support longer term alignment.
Still, some parts of the token model and governance system are developing. Newton’s governance remains in an early stage, so future community control should not be confused with complete decentralization today.
The protocol also uses EigenLayer based economic security alongside the role assigned to NEWT. How these mechanisms work together at scale will become clearer as the network matures.
Newton’s mainnet beta is an encouraging milestone, but it is only the beginning of the real test.
The protocol still needs applications to integrate its policy system. It needs enough independent operators to make evaluations resilient and decentralized. It must remain fast and affordable when policies depend on several outside data sources.
Adoption will not happen automatically.
Developers will only add another layer to their transaction process if it solves a meaningful problem without creating too much complexity. Institutions will need confidence that the system can operate reliably under pressure. Users will need clear proof that Newton protects their assets without taking away their control.
The strongest signal will not be social attention or short term token activity. It will be repeated protocol usage. Applications paying for evaluations, developers publishing useful policies and institutions relying on Newton to protect real capital would show that the system has practical value.
This is why Newton should be judged by its execution rather than its ambition alone.
The project began with a strong idea: AI agents should be able to automate financial activity without receiving unlimited authority. That idea has now expanded into a broader attempt to bring programmable authorization to onchain finance.
Newton does not claim that AI will never make mistakes. It starts from the more realistic assumption that mistakes are unavoidable.
Its answer is to define limits before the agent acts, verify every important action and block transactions that cross those limits.
That may become increasingly important as AI agents, automated vaults and institutional systems gain greater control over digital assets.
The future of onchain finance will not depend only on faster blockchains or smarter AI. It will depend on whether users can trust automated systems when real money is involved.
Trust in that environment cannot come from promises alone. It must come from visible rules, accountable execution and evidence that can be independently verified.
Blockchains already prove that a transaction happened.
Newton Protocol is trying to make sure it was allowed to happen.
@NewtonProtocol $NEWT #Newt

$BTC LONGS – CẬP NHẬT TRỰC TIẾP
Vẫn đang giữ quanh vùng vào lệnh, với Điểm Kiểm Soát Vốn (POC) là mốc quan trọng cần theo dõi.
Bối cảnh thị trường
BTC đã quét xuống cận dưới của biên độ và kể từ đó giao dịch trong khoảng $62.3K đến $63K. Mức đỉnh biên độ $63.7K và giá mở cửa theo tuần vẫn chưa bị chạm đến, qua đó giữ nguyên cấu trúc tăng rộng hơn.
Triển vọng hiện tại
Kỳ vọng chính vẫn là một đợt đi từ $63.7K lên $66K, nhưng nhịp tiếp theo phụ thuộc vào việc giá phản ứng như thế nào tại POC toàn cầu:
Giành lại POC: Động lượng có khả năng tiếp tục hướng tới $66K.