Skip to main content

Scaling AI on Ethereum

Scaling AI on Ethereum — research from Castle Labs.

Castle Labs4 min read
Scaling AI on Ethereum

Welcome to edition 35 of our weekly AI newsletter!

While usually we place a strong emphasis on Solana and Base as emerging AI ecosystems, this week we have the pleasure of diving deeper into Ethereum’s AI strategy, following the recent developments.

But first. How are we doing?


Sector Overview

The market cap is still hovering around $12.16B.

Article figure

Up until now, there have been 3 key levels which are indicative of the specific trends experienced in the sector:

  • Below $10b: bearish outlook.

  • Around $12b: strong neutral channel.

  • Above $14b: bullish outlook.

While, as we said, $12B has always been a key level, this week’s market movements leave little room for positivity.

In fact, many have highlighted how Altcoins OI was higher than Bitcoin OI, traditionally a sign of an oversold market and a potential signal for a broader retracement.

Last but not least, AI coins are also suffering from reduced mindshare.

They have, in fact, failed to capitalise on the initial attention received by the sector at the beginning of the year.

Among the defining factors for this are a strong rebound of DeFi and a change in regulators’ attitude towards the sector, which has pushed niches like stablecoins and Real World Assets (RWAs), much higher than others. Last but not least, many CT native investors have shifted towards robotics.

Nonetheless, teams continue cooking in the AI world.

In particular, last week we had an incredibly interesting news, about how Ethereum is starting a new team, with a focus on AI.

What’s this about? Can ETH even compete with other ecosystems which have proven to be particularly fit for AI projects (such as Solana and Base)? Let’s dive into it.


Scaling AI on Ethereum

Ethereum caught most by surprise when they announced that they were doubling down on AI, with the launch of the dAI (decentralised AI) team.

Article figure

The team, led by Davide Crapis, has the vision of “making Ethereum the preferred settlement and coordination layer for AIs”.

In practice, they’ll focus on two specific areas:

  • Decentralisation of the infrastructure: With the emergence of AI behemoths, this raises the issues of avoiding over-reliance on a few centralised entities and ensuring AI operates on a decentralised tech stack

  • Decentralisation of the AI economy: Aside from the underlying infrastructure, there is a strong need to ensure executions, coordination and alignment between agents happen in a decentralised and permissionless manner without third parties

Two main aspects of Ethereum’s unique value proposition are its credible blockspace neutrality and censorship resistance; as such, it will be interesting to observe the creation of AI agents which are aligned with this ethos.

In their own words, they aim to “connect two communities that have too often worked in parallel”.

By doing so, they can benefit from each other:

  • AI agents become safer, more trustworthy and decentralised

  • Ethereum becomes the main settlement layer, unleashing a new era of efficiency led by AI agents

So, not only another way to launch AI agents…. but fully neutral, verifiable, and censorship-resistant ones.

The dAI team will not be working in isolation, but in close collaboration with the Protocol and Ecosystem teams of the Ethereum Foundation, ensuring deep impact.

This can open up an interesting design space, where AI agents can finally be trusted with asset management, among others, and applied across more use cases.

This initiative has already created a ripple effect, as we’ve seen some related announcements.

As one of the first developments, the dAI team has launched a complement to the ERC-8004, a proposal for trustless agents.

Article figure

The newly deployed x402 extension adds native crypto payments to the A2A agentic communication protocol recently launched by Google

Article figure

For AI agents, this translates into:

  • Onchain settlement

  • Leveraging ERC8004 trustl and verifiable credentials

  • Universal standard for crypto payments between agents

Article figure

This proposal has been broadly embraced by ecosystem participants such as Metamask, which announced its support and close collaboration with Google.

“We believe a single, interoperable payment standard is essential for agents to buy and sell digital resources and services”.

Article figure

How would Metamask leverage this?

Article figure

Consesys is also committed to making “x402 multi-asset, multi-chain, and self-custody grade across our product stack”.

One such instance is already live, with DINBuild supporting “x402 micropayments so agents can autonomously purchase RPC access on demand”.

Some hackathon projects are already working on building tools around ERC8004.

Article figure

You thoughts we went a whole dispatch without talking about Virtuals? Think again.

Ethereum partnered with Virtuals to launch a series of official workshops for devs, and using the Virtual platform to deploy new agents.

Article figure

While this was by far the most interesting update for the week, here’s some broader news you might have missed.


Weekly News Recap

  • Calling all Solana builders! In collaboration with Superteam AE, Ignyte launched a Solana challenge, with 10k USD as prize.

    Article figure

  • Cod3x just launched its revamped trading terminal.

Article figure

Dive deeper into what this means and how it brings some changes to the CDX token.

Article figure

  • Robotics keep eating mindshare from AI projects, make sure to not be left behind. Here’s an interesting update from Xmaquina, specifically about humanoid robotics:

Article figure

Also really good aggregation of interesting robotics resources by Cookie to get you started, make sure to check it out!

Article figure

  • REI recently announced its improved domain-specific training. Leveraging Interence time Training can greatly improve how AI agents acquire domain knowledge:

    Article figure

The REI team is one of the most interesting to observe from a tech and development perspective. Looking forward to the launch in production.

  • We’ve already mentioned A2A in the context of Ethereum, but this is also a big news itself. The launch of an agentic commerce protocol by Google is a great confirmation signal for the sector. How many crypto partners can you find in the picture?

    Article figure

  • Nonetheless, while price action is pretty stagnant, AI continues to be the centrepiece of the development of most crypto projects. Here’s a great post by S4mmy putting things in perspective:

    Article figure

That’s it for this week. We hope you had a good time reading!

These developments on Ethereum are going to keep us busy for the next few weeks. Expect more updates.

Originally published in the Castle Labs newsletter. Subscribe at research.castlelabs.io.