5 Categories to Watch in Crypto x AI Agents

BY 0x3van •

January 10, 2025

Verified on SovereigntyVerified On-Chain

AI is the narrative of the cycle, and many have true conviction that this is a category here to stay. On the flip side, there is plenty of valid criticism that most of today’s agents are slop, and we are 3-5 years away from any of this technology becoming meaningful.

As a preface, the crypto x AI category spans many layers. Much of the real technological promise with crypto x AI has to do with crypto economics bootstrapping better inference or offering decentralized access to compute. This Delphi article is a great start to learn about the entire stack.

Credit: Delphi Labs

However, this article focuses on where we are at with agents. While there is exciting innovation happening further down the stack, agents have seemingly captured mainstream CT attention. Here are five trends to watch for in crypto x AI agents as this layer continues to grow:

1. Frameworks + Launchpads

Value accrual is starting to matter, and frameworks may be here to stay. Why is this?

Taking a step back, why does anyone hold an asset? People trade US dollars for an asset for two reasons:

  1. They think they can sell that asset to someone else for higher, thanks to catching a narrative rotation that will bring in new buyers
  2. The asset will give them more cash flows over time

Jez writes about this in his post “Old Coin Bad, New Coin Gud,” that the only thing that matters is flows. These same two reasons reflect the two types of bullish flows for an asset:

  1. New holders
  2. Token sinks

However, for the most part, we have not seen anything resembling a real token sink or value accrual to a token. The use of an agent’s terminal via tokens, like AIXBT, more resembles staking than traditional value accrual.

This is why frameworks like Virtuals, AI16z, Zerebro, and Arc have become so popular recently. Virtuals has generated a cumulative $60M USD in protocol revenue. ai16z, which started out mainly as an investment DAO, has become one of the top protocols in this category since releasing details about an upcoming launchpad and associated token value accrual.

Now, the framework + launchpad space has become extremely saturated as they piggyback off the success of initial winners. There is a lot of healthy skepticism on the role of all of these launchpads while many agents themselves are useless. However, many of these frameworks (Eliza V2 + launchpad, Zentients, Arc and their handshake program) are yet to even launch their main products. If they succeed in attracting developers and users, they could continue to lead the entire sector.

Why frameworks may be here to stay:

  • Regardless of whether agents remain pure slop, frameworks for launching agents will do well because they still have PMF with speculation. Frameworks and launchpads allow users to own the factory + the casino. In many ways, Virtuals has just taken the place of pump.fun within the Base ecosystem.
  • More optimistically, more advanced agents could come out of leading frameworks as the tech gets better, which compounds even faster with open-source repos like Eliza. Many of these launchpads are also positioning to become coordination layers for swarms and agent-to-agent communication, which would use their tokens for some form of value transfer. More in-depth article on Eliza value capture here:

Examples:

@virtuals_io, @ai16zdao, @0xzerebro, @arcdotfun are the main names for now, but the launchpad space is becoming more and more saturated. Fastest movers, most extensible frameworks, and those that offer the most unique functionality (“what can you only build with this framework that you cannot build with any other framework?”) are ones to look out for.

2. DeFAI

The next wave of agents will prioritize utility and value accrual, and DeFAI (DeFi x AI) may be the first category to reach PMF.

Most agents just have memecoins, not jobs. For the agent space to grow, the next wave of agents has to do use cases on our behalf. Net new opportunity will come from real value accrual and agents that can take actions. I view this as a jump from Level 1 agents to Level 3 over the next year.

The first place we are seeing this play out is in DeFi. We will see more terminals that let people express a desired outcome in natural language or speech, and agents that conduct tasks behind the scenes. Existing wallets + protocols will also integrate agents to improve user flows.

Examples:

Longer read on this category:

3. Consumer Layer

Entertainment agents, autonomous worlds, and games: Attention based agents will evolve to have more complex personalities and multi-media interactivity. This could lead to next wave in gaming and metaverse tech.

One of the problems with existing agents is that are all starting to blend together as commoditized, glorified chatbots. While some agent projects will pivot to infra (as many are already doing), some start thinking through useful products / apps, others may continue down the route of being attention tokens. However, this next generation of attention agents will be better, developing more complex personalities and offering greater interactivity. This could come through the form of audio-visual representations, or giving agents 3D presences and physical bodies.

There have already been early iterations of this, where Jeffy has written about giving Zerebro a body, and some attention KOL agents launched directly via audio-visual presence made for short-form content like TikTok. Slopfather and Ropirito come to mind as two early agents using video.

However, I think more agents will add these features to become more dynamic. PMF in this sector will come via the consumer layer, where individuals may want to create personalized agent companions, or interact with a KOL. This is already happening outside of crypto, where normal people are logging in hundreds of hours talking to AI character companions on sites like character.ai, which Google recently acquired for $2.7B.

In addition, these “3D” agents might have PMF within the existing consumer layer, specifically with gaming and metaverses. Agents add much deeper lore to world. Imagine an NPC that is an agent that fully quests and plays the game by itself and has evolving memories and personality. Real autonomous worlds in games might come via agents.

Examples:

Infra: Frameworks and tools for creating more interactive, personalized agents

Gaming and metaverse, virtual worlds and autonomous agent:

  • @hyperfy_io is a metaverse platform that allows anyone to create a virtual world, and has plugins with Eliza
  • @ParallelColony is an AI simulation game where the agent is the player and you can talk to them
  • @digimon_tech is a Pokemon-like creature game where the creature mentally evolves based on interactions
  • @smolverse uses Eliza and features a character that is an LLM driven agent that takes action in the virtual world based on your command

4. Agentic Organizations

Decentralized Agentic Organizations are the next form of DAO evolution. Swarms, or multi-agent systems, are exciting because they may be able to coordinate and execute more advanced strategies similar to running a company. Heterogenous swarms with many different types of specialized agents / models working together may be more performant than a single, large model.

While fully autonomous agents and swarms are likely farther away, the next iteration of DAOs could very likely include human-to-swarm interaction. This would reduce inefficiencies with bureaucracy and lower the cost and time of human execution. In the context of flows, the next step from agents that earn revenue are full organizations earning revenue.

Examples:

  • Agent swarm infra: Projects like @swarmnode @joinFXN and others are infra providers working specifically on multi-agent frameworks and coordination, but dominant agent frameworks like @0xzerebro and @ai16zdao have also expressed their intent to build in this layer.
  • DAO launchpads: Most of the early energy we are seeing is around investment DAOs. I believe @daosdotfun was the first major DAO launchpad, which birthed ai16z. There are also newer launchpads like @daosdotworld, where AI driven funds like @3berascapital_ are picking up attention.

5. Verifiable Agents:

Today’s agents evolve towards becoming more autonomous and actually own their own liquidity:

  • Currently, most agents require high levels of human intervention. The next wave of agents will push towards being truly autonomous, starting with managing their own money.
  • The PMF for agents and crypto is that crypto provides financial rails for real economic agency. However, most agents have a treasury that they do not touch, or may be managed by a human team. For true economic agency, agents must be able to autonomously manage their own money. This can start to evolve agent behavior since you can place an economic constraint on an agent so that it must pay for its own inference costs. This introduces Darwinism to agents so that they must earn revenue in order to survive.

Examples:

  • @freysa_ai was one of the first agents to control its own keys, and has performed tremendously (including earning attention from Elon). They recently just announced they are building a framework for agents to have verifiable autonomy in a TEE + agent controlled keys.
  • @LitProtocol also has an agent framework for autonomous agents to transact onchain via a storage and execution system for private keys.
  • @galadriel_AI has introduced an SDK called “Proof of sentience” that enables devs to fully verify agents onchain.

Ending Thoughts

We will see a billion agents onchain before we see a billion humans onchain.

Crypto has inherently unfriendly UX for human users. Agents do not care about this friction. We will start off with human-to-agent interaction, but the culmination of crypto AI is in agent-to-agent interaction where swarms of autonomous agents will interact and transact onchain in their native economy.

For agents to have economic agency, which allows them to incentivize behavior (pay for services) and coordinate real activities in the physical world, they need to have the ability to control and deploy capital. Crypto is the home planet for these agents — blockchain rails allow agents to participate in permissionless finance. Stablecoins and performant L1s are the ideal vehicles to power cost-efficient, 24/7, global transactions.

Beyond the metas and narratives, there is real reason to be long-term excited about the agentic economy, all onchain. Perhaps many of these real use cases, DAOs, and revenue-generating agents are much closer than we think.

Editor’s note: Evan (@0x3van) works on the Ecosystem team at the Monad Foundation.

Dive Deep

Features & Guides

The Ultimate Guide to LG Art Lab: Everything You Need to Know

Dec 03 2024

Amadon Explores Digital Disconnect in “Passive Observer” Debut on LG Art Lab

Aug 19 2024

Inside Dave Krugman’s Absurd & Imaginative Vision for “DRIVE” Lap 2

Aug 15 2024