Aethir co-founder: AI agents don't need blockchain, but blockchain can't do without AI

Author: Mark

Translation: TechFlow

Unpopular opinion: AI agents don’t need blockchains, but blockchains desperately need agents.

Agents are here now, and in my view, it might only take a few months before we enter a "decomplexification era" for crypto — an era where AI agents seamlessly solve the long-standing user experience (UX) hurdles that have kept blockchain from going mainstream. But before we get there, the industry will go through a chaotic phase.

No pain, no gain — to make agents truly useful in crypto, the entire industry may first undergo a kind of “self-collapse.” Just as the democratization of meme coin launches (e.g., via pump.fun) triggered a meme coin supercycle, significantly reshaping market liquidity distribution and causing sideways markets for months, the same phenomenon will unfold with agents. As the cost and barrier to building agents drop, thousands of agents will flood online.

Initially, these agents will primarily connect to social media, acting as one-way tools to broadcast tokens — sharing, promoting, nudging, or trolling. But this model will quickly become boring. Then, agents will begin executing simple on-chain transactions. While that sounds promising, teaching agents to recognize a “good trade” is extremely difficult (just ask anand iyer).

Thus, we’re entering a “chaos phase.” During this time, low-value or low-complexity agents will frequently transact low-value assets on cheap blockspace. These agents will manipulate surface-level metrics, clouding our ability to identify real on-chain data. Crypto Twitter will be flooded with agent-generated content, and cheap blockspace will be heavily consumed. The entire industry could descend into disorder.

Many will develop fear, uncertainty, and doubt (FUD) toward agents. But that’s not necessarily bad — building truly great agents takes time. Just 12 months ago, it was nearly impossible for agent developers in crypto to attract venture capital attention — I know this firsthand, having backed several teams at the time. Back then, all the interest was in infrastructure (e.g., AethirCloud was one beneficiary), while the potential of agents remained largely unrecognized.

Now, we're at the peak of early hype around a transformative technology. So what comes next? What does it mean for established blockchains like ETH if cheap blockspace gets saturated by agents? Will “luxury blockspace” make a comeback? Will agents differentiate based on the chains they operate on? How will crypto UI evolve when agents can both aggregate information and act as tools to move assets from point A to point B? What separates paid agents from free ones? How do we cost-effectively scale agent reasoning? And from an application standpoint, what business models can properly cover these costs?

To me, the most exciting part of these new trends is the unexpected surprises. Who would’ve thought that a crowdsourced DAO launch platform — Daos.fun (daos.fun, baoskee) — would become a catalyst for asset management agents? In hindsight, it seems obvious, albeit slightly ahead of its time, yet it provided much-needed momentum to this trend (with contributions also from ai16z.ai and Shaw). So, what do agents need next?

Still, I firmly support building more things that people actually use, rather than piling up more infrastructure. That’s exactly why I’m excited about the agent wave — it centers on end users. We all believe in this technology because it has the potential to manage our assets more efficiently than we can ourselves.

Finally, support agent builders — even amid fear, uncertainty, and doubt (FUD). It’s okay. This is the beginning of a whole new Web3, where user experience (UX) is no longer a barrier to creating exceptional things. Do your own research (DYOR), but stay optimistic — the future will be amazing.