Last night I couldn’t sleep and looked through Newton’s npm packages—@newton-xyz/sdk, v1.0.5, released just on June 25. It’s a TypeScript SDK, built on EigenLayer AVS, with a pretty clear module breakdown: HPKE encryption, zkPermissions, policy evaluation, and identity registration—everything you’d expect is there. The docs even include a Quickstart—“simulate your first policy evaluation in 5 minutes.”

It looks like a legit project. Then I clicked into the code repository and browsed it for half an hour—the hair on the back of my neck started to stand up.

Bitget’s report is written in an upbeat tone, saying Newton wants to build a “verifiable on-chain automation layer,” packaging TEE, ZKP, and zkPermissions into developer tools. Foresight’s data looks great too—Magic Labs raised nearly $90 million, and PayPal Ventures, Tiger Global, and Polygon are all at the table.@NewtonProtocol

But after I read the SDK Reference, I found a very practical problem: for a developer to get started with Newton, they have to figure out how to operate the TEE hardware, how to design ZKP circuits, how to integrate with EigenLayer AVS, and also the permission syntax of zkPermissions—each layer is hardcore cryptography. No matter how beautifully the documentation is written, for an ordinary Web3 developer to run a first AI agent end-to-end from scratch, the learning cycle is at least counted in weeks.

What makes me even more restless is that the community has already dug into the data: Newton’s initial circulating supply is only 21.5%. The community allocation is 60%, to be released gradually, while the TVL and the number of active users have not been disclosed to this day. The token has fallen 94% from its peak. A protocol without real ecosystem support—no matter how pretty the SDK is—is just a toolchain, not a moat.#Newt

Newton wants to create a unified logic for underlying convergence. I understand—that fragmentation in the AI agent space does need a standard. But the premise for a standard to be adopted is that someone actually uses it. When developers find the learning cost is high, ecosystem incentives are unclear, and the token price is still going down, how many people will choose to build the first agent on Newton?$NEWT

Magic Labs’ GitHub account has 0 stars. An infrastructure project that claims to “bring automation fully onchain” has a code repository with zero attention. Not even people just watching out of curiosity.

I’ll keep watching Newton’s progress—especially the adoption of the SDK and the quality of PRs submitted by community developers. But since the tools are still new and the ecosystem is still empty, I won’t put large amounts of money in. Let developers first teach other developers how to build and use it, let the ecosystem grow things, and only then talk about value capture.