Usman Sam

Volto sempre a um teste simples que uso há uma década: se um protocolo desaparecesse amanhã, alguém perceberia além das pessoas que estão mantendo o token? Na maioria dos projetos de "IA + cripto", a resposta honesta é não — e foi essa a lente que eu usei para o Newton Protocol antes de deixar que eu me interessasse pela narrativa em torno dele.
O argumento é coerente o suficiente para eu entender por que as pessoas são atraídas. A Newton está construindo um rollup dedicado para permitir que agentes de IA gerenciem carteiras e executem estratégias sem nunca manter suas chaves diretamente. O mecanismo são contas inteligentes baseadas em ERC-4337 e EIP-7702, combinadas com ambientes de execução confiáveis e provas de zero conhecimento, para que as ações de um agente sejam, supostamente, tanto privadas quanto verificáveis. Em termos simples: o agente faz algo em seu lugar e, em vez de confiar na palavra do operador, você recebe um comprovante criptográfico mostrando que a ação correspondeu às regras que você definiu. Esse é um avanço real em relação aos robôs de negociação de “caixa-preta” que circulam no DeFi há anos, em que você não tinha como saber se o robô fez o que alegou ou apenas o que era conveniente para quem o operava.

Tenho ficado encarando um padrão de gráfico na NEWT que algum trader chamou de rompimento de triângulo simétrico, e toda vez que vejo essa frase penso em quanto dessa indústria funciona com base em formas em vez de substância. Um triângulo em um gráfico de preços te diz algo sobre a psicologia das pessoas que estão negociando o token. Não diz nada sobre se o protocolo por baixo dele realmente funciona. Então deixei o gráfico de lado e fui procurar a coisa por trás.
O Newton Protocol está tentando resolver um problema que eu vi atrapalhar todos os ciclos de DeFi desde 2017: como você permite que outra coisa — um bot, um script, agora um agente de IA — atue em seu lugar com o seu dinheiro, sem entregar suas chaves e torcer pelo melhor. A resposta inicial sempre foi uma variação de "confie em nós". Bots de negociação centralizados, scripts offchain, chaves de API guardadas em algum arquivo de configuração. Eu já perdi dinheiro com esse modelo antes, não exatamente por um hack, mas por falta de transparência — uma estratégia que silenciosamente mudou o comportamento e ninguém conseguia provar o que aconteceu, porque nada era verificável. Então quando um protocolo diz que quer tornar a automação "verificável" em vez de apenas "automatizada", eu presto atenção, porque isso é uma afirmação arquitetural de verdade, não um slogan.

I've been doing this long enough that when I read a whitepaper now, I skip straight to the parts that sound uncomfortable. Not the vision paragraph. Not the "we are building the future of X" section. I go looking for the part where the team admits something is hard, because that's usually the only honest sentence in the whole document. Newton Protocol has a few of those, which is more than I can say for most things that land in my inbox this cycle.
For anyone who hasn't run into it, Newton is trying to build a rollup specifically for AI agents that trade, rebalance, and execute financial logic on your behalf, without you handing over your keys. The pitch is permissions instead of custody. You say "only trade if volatility exceeds X" and the agent can act inside that boundary but never outside it. Trusted execution environments do the actual computation, zero-knowledge proofs verify that the computation followed the rules, and a keystore rollup manages all of this across chains. There's a model registry where developers publish these agent templates like they're apps in a store, and a token, NEWT, that pays for gas, staking, and collateral when operators run these agents.
I want to be fair here before I get cynical, because I think fairness matters more than people give it credit for in this space. The architecture is not lazy. Using TEEs alongside ZK proofs instead of picking one is a real design choice, not a marketing choice, because it acknowledges that neither tool alone solves the trust problem cleanly. TEEs give you speed and privacy but ask you to trust hardware manufacturers. ZK proofs give you cryptographic certainty but are slow and expensive for anything computationally heavy, which AI inference definitely is. Combining them is the kind of decision that comes from people who've actually hit the wall both ways before, not from people reading a stack overflow thread on zero-knowledge proofs the week before drafting the deck.
So credit where it's due. This isn't vaporware dressed up in AI language, which is most of what I see these days. Half the "AI agent" tokens I look at are just a cron job with a Twitter account. Newton at least tried to solve the actual hard problem, which is: how do you let something act on your behalf without trusting it completely and without making every action cost twelve dollars in gas.
But here's where the fatigue kicks in, and I mean actual fatigue, the kind you get after watching six or seven versions of this same idea come and go.
We've had "agent marketplaces" before. We've had "verifiable automation layers" before, just wearing different words. Every cycle produces its own vocabulary for the same underlying hope, which is that if you build sufficiently elegant infrastructure, demand will show up because the infrastructure deserves it. It rarely does. Infrastructure doesn't get used because it's well designed. It gets used because someone standing in front of a real, annoying, expensive problem finds it and thinks "this saves me time" or "this saves me money," and they don't care at all about the zkML architecture underneath it. They care that it works, that it's cheap, and that they don't have to think about it twice.
And this is where I get stuck with Newton. Not because the tech is bad. Because the tech being good has never been the thing that decides these outcomes.
Think about who actually needs a verifiable automation layer for trading agents right now. Institutions might, eventually, once compliance requirements force them to prove that automated systems followed rules, and Newton's MiCA positioning suggests they're already angling for that audience. That's smart, honestly, probably the smartest part of the whole strategy. But institutional adoption moves at institutional speed, which is to say glacially, through legal review and risk committees and pilot programs that take eighteen months to greenlight a thing that a retail trader could just start using tomorrow if they wanted to.
Retail, meanwhile, doesn't really want verifiability. Retail wants yield, or wants a number to go up, or wants to feel like they're doing something clever with less effort. The people already running trading bots aren't sitting around wishing their bot's decisions were cryptographically provable. They're wishing the bot made more money. Verifiability solves a trust problem that most users haven't actually experienced as painful yet. You're building the seatbelt before anyone's had the crash.
That's the tension I keep circling back to with projects like this. There's a real difference between something that looks meaningful, that reads well in a deck, that architecturally makes sense to anyone who's spent time in cryptography, and something that becomes necessary in the way that, say, a stablecoin or a DEX became necessary because people were bleeding money or time without them. Necessity is rarely elegant. Necessity is often ugly, patched together, held up by duct tape, because whoever built it first was solving a problem they personally had, not a problem they anticipated other people might have someday.
Newton feels anticipatory. Which isn't an insult. Being early isn't the same as being wrong. But being early in crypto usually means dying quietly two years before the thing you predicted actually happens, and then somebody else builds a worse, uglier version of your idea at exactly the right moment and captures the whole market because timing beat design.
There's also the token unlock overhang, which nobody likes talking about in the technical writeups but which matters enormously to whether any of this ever gets the runway to prove itself. A large chunk of supply unlocking while the actual product, the marketplace and the multichain rollup, are still "upcoming," not live, creates this uncomfortable arithmetic where the market has to believe in the roadmap faster than the roadmap can deliver anything real. I've seen this pattern enough times to recognize it without needing to check a calendar. Sell pressure doesn't care about your architecture diagrams.
And the friction problem is the part I think gets underweighted the most, even by people who are otherwise skeptical. Even if Newton nails the tech, even if the Keystore rollup ships clean and the marketplace fills up with legitimate agent models instead of the usual copy-paste garbage that floods every "developer marketplace" within a month of launch, someone still has to configure permissions correctly, understand what "only trade if volatility exceeds X" actually means in practice, trust an agent enough to let it run unattended, and then actually check on it periodically because nobody fully automates anything they care about without checking. That's a lot of cognitive steps for a user who could otherwise just... trade manually, or use a centralized bot that already exists, already works, and doesn't ask them to think about zero-knowledge proofs at all.
Good design reduces friction in theory. In practice, good design often just relocates the friction somewhere less visible, and calls it solved.
I don't say any of this because I think Newton is a scam or a waste of engineering hours. It clearly isn't. The team seems to understand what they're building and why, which already puts them above a lot of the field. What I don't know, what nobody knows yet, is whether the world they're building for actually exists yet or whether they're constructing a very well-made bridge to a shore that hasn't formed.
Maybe the agent economy becomes real in the way people keep promising it will, and Newton ends up being the boring, necessary plumbing everyone quietly depends on without thinking about it, the way nobody thinks about TCP/IP. That would be a genuinely good outcome, and a rare one.
Or maybe it stays a well-built answer to a question the market hasn't gotten around to asking yet, and it fades the way a lot of "ahead of its time" infrastructure fades, not with a collapse, just a slow draining of attention toward whatever narrative shows up next quarter.
I don't know which one this is. I'm not sure the people building it know either, and I'd trust them less if they claimed to.
@NewtonProtocol #Newt $NEWT

Eu volto repetidamente a um teste simples que peguei em algum lugar por volta de 2018 e nunca abandonei: quando um projeto muda o que diz que é, eu dou menos atenção ao novo pitch e mais atenção ao porquê o antigo deixou de funcionar. O Newton Protocol é um estudo de caso útil para esse teste agora, porque ele foi lançado como uma coisa e, apenas alguns meses depois, está sendo descrito como algo visivelmente diferente — e eu acho que essa lacuna diz mais do que o roadmap.
A formulação original, aquela que deu ao NEWT sua atenção inicial, era sobre agentes de IA negociando e gerenciando carteiras em seu nome dentro de um rollup seguro, com um marketplace onde desenvolvedores poderiam publicar e monetizar esses agentes. É uma narrativa forte, que atravessa dois ciclos de hype ao mesmo tempo: IA e finanças autônomas onchain, e atraiu capital real antes do lançamento, incluindo uma rodada estratégica liderada pela PayPal Ventures. Mas em algum momento nos últimos meses, a descrição pública mudou para "compliance-as-code" (conformidade como código), uma camada de aplicação de políticas que permite que instituições e emissores de stablecoin verifiquem transações em relação a regras antes que elas sejam liquidadas. Esses não são o mesmo produto. Um é sobre gerar retornos por meio de automação. O outro é sobre infraestrutura regulatória. Uma equipe pode construir ambos, muitos projetos de infraestrutura ampliam seu escopo, mas quando o centro de gravidade na comunicação muda tanto, tão rápido, eu quero saber se isso é clareza estratégica ou se é uma resposta ao fato de a primeira ideia não ter ganhado tração. Eu ainda não tenho uma resposta confiante para isso, e ficaria desconfiado de qualquer pessoa que afirme que tem, tão cedo.

Passei algum tempo esta semana voltando ao Newton Protocol porque, sinceramente, eu tinha quase esquecido dele.
Alguns meses atrás era mais uma dessas ideias de IA x cripto que pareciam interessantes no papel, mas ainda pareciam distantes do uso real. Todos nós já vimos vários projetos prometerem "finanças autônomas" sem mostrar muito além de diagramas e roadmaps ambiciosos. Depois do último ciclo, fiquei bem cético em relação a qualquer coisa com IA no título.
Mas agora as coisas parecem um pouco diferentes.

Se você tem acompanhado o mundo das criptos ultimamente, provavelmente já percebeu que quase toda semana surge um novo projeto de IA fazendo manchetes. Vou admitir que, quando eu me deparei com o Newton Protocol (NEWT) pela primeira vez, pensei: "Lá vamos nós. Mais um projeto jogando IA na mistura."
Depois de passar algum tempo lendo sobre isso, percebi que ele tenta resolver um problema que realmente importa.
Se a IA vai gerenciar cripto no futuro, como você permite que ela faça transações sem dar a ela controle total da sua carteira?

Tenho pensado em como o mercado cripto se apaixona rapidamente por novas narrativas e como raramente ele pausa para perguntar se a economia subjacente realmente merece essa atenção. Eu vi soluções de camada 1 prometerem substituir tudo, protocolos de infraestrutura afirmarem que se tornariam a espinha dorsal da internet, e tokens de IA chegarem carregados de expectativas que nenhum projeto em estágio inicial conseguiria satisfazer de forma realista. Essa história me deixa mais lento do que a maioria quando olho para algo como o Newton Protocol. Eu continuo voltando à mesma pergunta que me salvou em múltiplos ciclos de mercado: se o mercado deixasse de recompensar histórias amanhã, o que ainda valeria a pena estudar?

I have been paying attention to how different blockchain projects are trying to solve practical problems, and Newton Protocol is one that keeps drawing my attention back for an interesting reason. Instead of focusing on a single application, it seems to be thinking about what happens when AI begins handling more decisions on-chain, especially in areas like automated trading and strategy execution.
As I spent more time reading about the project, I started to see a clear theme running through it. Newton Protocol is built around the idea of creating a secure rollup where AI-driven strategies can operate with greater confidence. That stood out to me because AI is becoming more capable of making complex decisions, but those decisions still need an environment that people can trust. The protocol appears to place that foundation at the center of its design rather than treating security as something added later.
Another part that caught my attention is the way the project connects different participants. Instead of limiting itself to infrastructure, Newton Protocol also includes a marketplace for AI developers. I find that interesting because useful AI tools often depend on the people creating them just as much as the technology itself. Bringing developers into the same ecosystem as users and automated strategies makes the project feel more connected, with each part supporting the others instead of existing independently.
The focus on AI-driven strategies and automated trading also made me think about how blockchain and artificial intelligence are gradually becoming more intertwined. Rather than seeing automation as a separate feature, Newton Protocol seems to treat it as something that should be supported by the underlying network itself. That approach feels consistent with the broader vision the project is presenting.
I also noticed the presence of the NEWT token within the ecosystem. While every blockchain project has its own token, I tend to look beyond the name and consider how it fits into the overall structure. In this case, it feels like part of a larger system that is being designed around secure AI activity instead of existing as an isolated component.
The more I reflected on Newton Protocol, the more I found myself thinking less about individual features and more about the direction it is trying to explore. The combination of a secure rollup, AI-driven strategies, automated trading, and a marketplace for AI developers creates a picture that feels coherent rather than fragmented. That sense of alignment is what has kept the project in my thoughts, and it is the reason I am interested in following how it continues to develop.
@NewtonProtocol #Newt $NEWT

The more I study $NEWT , the more I think the next major challenge in Web3 is no longer building faster blockchains or cheaper transactions it's building systems that can safely manage autonomous decision-making.
For years, blockchain security has been built around one simple principle: cryptographic ownership.
If you control the private key, you control the assets.
Every wallet, multisig, and smart contract ultimately relies on cryptographic signatures to determine whether an action is authorized. This model has served the industry remarkably well because it solved one of the hardest problems in digital finance: proving ownership without relying on centralized intermediaries.
But the environment is changing.
The rise of AI agents, smart accounts, automated treasury management, algorithmic trading systems, and autonomous on-chain applications introduces an entirely different security problem.
The question is no longer just "Who signed this transaction?"
The more important question becomes:
"Should this transaction be executed at all?"
That distinction is subtle, yet it fundamentally changes how we should think about blockchain security.
A valid signature only proves that an authorized entity approved a transaction.
It does not prove that the transaction follows organizational policies.
It does not verify treasury rules.
It does not enforce spending limits.
It does not check jurisdictional restrictions.
It does not evaluate operational risk.
And it certainly cannot determine whether an autonomous AI agent is acting within the boundaries originally intended by its owner.
As AI systems become increasingly capable of initiating transactions without direct human involvement, relying solely on cryptographic authorization becomes an incomplete security model.
Ownership and intent are no longer the same thing.
This is where Newton Protocol ($NEWT ) introduces an interesting architectural perspective.
Rather than replacing wallets, custody providers, or existing authentication mechanisms, Newton Protocol positions itself as a decentralized programmable policy layer that evaluates transactions before execution.
That distinction matters.
Instead of asking only whether a transaction carries a valid signature, Newton asks whether the transaction also satisfies a predefined set of programmable policies.
Think of it as introducing governance logic directly into transaction execution.
Developers can define rules that determine how assets should move before any transaction reaches the blockchain.
These policies can include spending thresholds that prevent excessive transfers, jurisdiction restrictions that limit where assets can be sent, counterparty validation to ensure interactions occur only with approved addresses, AI execution limits that constrain autonomous agents, and protocol-specific operational controls tailored to individual applications.
Rather than relying on humans to manually review every action, these policies become programmable infrastructure.
This reflects a broader shift in security philosophy.
Historically, blockchain security focused primarily on protecting ownership.
Tomorrow's security may focus just as much on protecting behavior.
Those are not identical objectives.
An AI trading agent may possess valid credentials while simultaneously making decisions that exceed acceptable risk parameters.
A treasury automation system may execute correctly according to its programming while still violating governance policies established by a DAO.
A smart account may authenticate perfectly while interacting with contracts that fall outside an organization's compliance framework.
In each case, authorization alone is insufficient.
Policy enforcement becomes equally important.
One of the more compelling ideas behind Newton Protocol is the separation of ownership, permission, and execution into distinct layers.
Ownership determines who ultimately controls assets.
Permission defines what those owners allow.
Execution determines whether a specific action complies with those permissions under current conditions.
Separating these responsibilities creates a more resilient architecture because it reduces the assumption that authorization automatically implies correctness.
This layered approach resembles security practices that have evolved in traditional enterprise systems over decades.
Financial institutions rarely rely on signatures alone.
Internal controls, spending approvals, compliance engines, risk monitoring, and policy enforcement all exist alongside authentication.
Blockchain infrastructure is gradually reaching a similar stage of maturity.
As decentralized applications become more sophisticated, similar operational safeguards become increasingly valuable.
This trend becomes even more relevant in the context of AI.
AI agents excel at speed, automation, and optimization.
However, they also introduce unpredictability.
Even well-designed systems can encounter unexpected market conditions, adversarial environments, or conflicting objectives.
The challenge is not necessarily preventing AI from acting.
The challenge is ensuring AI acts within clearly defined boundaries.
Programmable policies create those boundaries.
Instead of assuming autonomous systems will always make desirable decisions, protocols can continuously evaluate whether every proposed transaction aligns with predefined operational rules before execution occurs.
That changes the security conversation from reactive monitoring to proactive verification.
In many ways, cryptography solved the problem of identity.
Programmable policies aim to solve the problem of intent.
Both are necessary.
One verifies who initiated an action.
The other verifies whether that action should proceed.
As autonomous finance expands, these two dimensions increasingly complement one another rather than compete.
This may ultimately represent one of the most important infrastructure shifts of the coming AI era.
Future Web3 systems will likely involve thousands or even millions of autonomous agents interacting continuously across decentralized networks.
At that scale, human oversight alone becomes impractical.
Security architectures must evolve beyond signatures toward automated policy verification that operates at machine speed.
Newton Protocol appears to be building toward that future by introducing programmable decision layers that sit between authorization and execution.
Whether this architectural model becomes a widely adopted standard remains to be seen, but the underlying problem it addresses is difficult to ignore.
As blockchain applications evolve from simple user-driven transactions to increasingly autonomous systems, security can no longer depend solely on proving ownership.
It must also prove responsible behavior.
Perhaps the most valuable infrastructure in the next generation of Web3 will not be the protocols that simply verify signatures, but the ones capable of proving that every authorized action also complies with predefined governance rules, operational policies, and risk constraints.
Because in an AI-powered financial system, the strongest security may come not only from cryptography but from programmable intent verification.
@NewtonProtocol #Newt $NEWT