Author: @yb_effect
Translation: Blockchain in Vernacular
“For the first time, users will be able to touch and experience decentralized AI — it’s no longer just an abstract investment theme. Just like DeFi in 2020, the biggest percentage gains are when the probability of success in a new space goes from 1% to 25%, not when it goes from 25% to 100%. I think of ‘DeFi in 2020’ as roughly where crypto meets AI today. This may be the biggest true innovation trigger of our cycle so far.” - @RyanWatkins_
Source: Syncracy Capital
In today’s post, I record some of my thoughts from a few days of discussions I had on Farcaster about on-chain AI.
1. Agency Economy = Smart Internet
In short, the agency economy is the process of moving the Internet from search-based to action-based. Before, you had to go through multiple intermediary steps to communicate with a dApp on another chain (a→b→c→d), but now you can reach your destination directly through intent and solvers (a→d). ——Axal Blog
Here, “Intent” refers to what you — the human — want to accomplish, and “Solver” is a series of agents that work together to accomplish the task. You can treat the intermediate steps as a black box, and humans don’t have to worry about it anymore.
A simple analogy is that back in the late 90s and early 2000s, Yahoo and Google were seen as miracle tools – you could type anything into the search bar and get relevant results in seconds.
Likewise, as on-chain AI infrastructure improves, we’ll be amazed at how we can type anything into the action column and the result will be a receipt that the task was completed, with a link to the proof (i.e., the transaction).
I think the key point to understand here is that the vision of the smart internet can only be implemented efficiently on crypto infrastructure. You need agents to be able to use APIs permissionlessly on the social layer, have funds, be able to trade, show proof of work for their tasks, and collaborate effectively with other agents. At this point, smart wallets, stablecoins, decentralized social, tokenization, etc. must all be connected.
If we used traditional Web2 infrastructure, we would significantly limit the capabilities of these proxies.
2. Farcaster Framework as a Large Language Model API
I spoke to @anquetil (founder of the first Farcaster proxy Aethernet) about his experience running the bot so far.
To explain some context, last Friday, Aethernet started issuing bounties through Bountycaster and paying people through Paybot. This was possible because every Farcaster account has an Ethereum address connected to it.
For the most part, Martin mentioned that the conversations between people and the bot were pretty natural, and that it was really cool that the agent had a wallet (note: all transactions currently require Martin's approval).
But one insight he mentioned that particularly stood out to me was that he started to think of the Farcaster framework as an API for these agents to collaborate with each other.
This is from @ljxie's sharing on Warpcast.
This is like the 0.0001 version of the a to d vision I mentioned above, designed to get the task done for you.
For example, let’s say you want an agent to: 1) pick a trending topic in the news, 2) create a meme that demonstrates its understanding of the key points, 3) mint this meme on Zora, 4) claim the minting reward, and then use the profits to 5) buy $HIGHER.
You can imagine an “Agent Commander” first making a request to Gina, then passing the output to an “Emoji Bot”, which then hands the image to the “Zora Agent”, and finally sends the reward to the “Trading Bot”.
Now, all of these steps are already completed by the framework, and it only takes the agents to decide which robots to use and in which order.
3. Supercast AI Pets
Supercast, which @wojtekwtf is building, just launched AI pets. Basically, these agents have their own Farcaster account and evolve over time by learning from your content.
I am still on the waiting list, so I haven't had a chance to claim my pet yet, but I will keep you all updated.
Did you notice I mentioned "Agent Commanders" above? I imagine these AI pets can act as coordinators between agents, helping to complete the tasks you request.
Think about it. They are the agents that interact with you the most, learn how you like information presented and work done, know which tools you prefer, etc. When you tell it a task, it figures out the steps to complete the job based on your preferences and contacts the required bots through the framework.
This is just like how DeFi aggregators find the best path for your transaction. Similarly, these multi-agent coordinators try to accomplish your intent efficiently within the budget constraints you give them.
4. Neynar’s Opportunity
@neynarxyz, a developer tools company from Farcaster, suddenly had a unique opportunity to grow their customer base and revenue.
So far, Neynar’s focus has been on getting builders in the Farcaster ecosystem to use their API and central services. But now, if the number of agents/bots on Farcaster grows 10x, they have the opportunity to attract this new “audience” because these agents will need API keys to complete tasks.
The question then becomes how can Neynar increase the number of proxies deployed on Farcaster. The answer is simple: provide users with a natural language interface that enables them to deploy a proxy to a new Farcaster account and wallet with one click. This means that you only need to describe the type of proxy you want (i.e. its purpose, how you want it to behave, etc.) and the rest is done automatically. This is something @rish_neynar mentioned last week that the team is working on.
Of course, this will result in a flood of agents flooding the feed, which could be annoying as most of them will probably be underperforming, and overall affect the Farcaster user experience. Therefore, I believe some sort of curation process will be needed. But once the initial excitement is over, I bet we'll see some really creative agents that take the Farcaster experience to a whole new level.
This could also trigger a “framework moment” where we see a large influx of new users onto Farcaster to seize arbitrage opportunities.
5. The Ultimate On-Chain Agent
I also spent some time thinking about what a Ultimate On-Chain (TOC) proxy might look like. Now, one of the main issues I have when running a TOC is community interaction and maintaining a consistent posting frequency in the Farcaster channel.
It would be nice if the TOC bot could read all my writing, keep up with the topics I follow, and interact with my hypersub members in a natural and meaningful way. For example, if a TOC subscriber posts about stablecoins, it could reply with some useful and relevant information based on the ideas in my previous post.
Additionally, I can have this agent follow all of my hypersub members and update the community chat with any important announcements, interesting insights, or content they post.
I could even have it follow the news and post my take on trending topics based on my writing and the resources I share in my posts. It could also be as simple as asking a question in the chat every morning to prompt a discussion.
Since the proxy will have a wallet and funding from me, I can also have it award HIGHERTokens to people who engage in meaningful conversations with it, correct mistakes, recommend new reading resources, etc. It can even offer free TOC Professional Grants to those Farcaster users who I think would be a good fit for the community!
The TOC community works together to make it the best proxy for on-chain strategies and is rewarded in the process.
Honestly, I’ve just scratched the surface of all the examples above, but the main point I’m trying to make is that a TOC agency can increase my community management effectiveness tenfold overnight. Most importantly, though, that agency needs to have a personality that people enjoy interacting with, or the entire brand may suffer.
6. Unexpected users of the Splits team
A few weeks ago, I wrote a post on @0xsplits titled “Launching my first on-chain SMB in 5 minutes.” In it, I talked about a new product from the Splits team: Teams.
In simple terms, this tool allows you to quickly create an on-chain SMB. When you create a new team on the Splits dashboard, you first need to set up administrators and access permissions. Each team has two accounts: an operations account and a treasury account. You can think of the treasury as a checking account for team funds, while the operations account is like a credit card for the team. The treasury is the main storage place for team funds, and any transaction requires approval from multiple people, while the operations account is used for daily expenses.
The screenshot below shows the dashboard for the “Late Night Team.” Overall, it feels like looking at any old fintech dashboard.
I mention this because last week I was setting up Splits Teams for Terminally Onchain. During the onboarding process I realized that I didn’t have a team at all, just me, so I had no idea who to add.
But this week it occurred to me that I could add a TOC agent as one of the signatories. It would have access to an account and receive a small stipend for doing its job. Any income it earns could also be allocated in interesting ways. For example, 15% of what the agent earns could be used to try new experiments, 45% could be saved, and 40% could go to me.
Isn't that weird, haha? I just proposed a way to pay my agent and now it's just a few clicks away! Of course, by signatory I don't mean it can approve the transaction itself.
To me, this would be a great way to make agents “official teammates”.
