Leaving Substack?
First - I've decided to relocate my newsletter from Substack to Paragraph.
Despite being fairly crypto-native, I originally chose Substack over Paragraph/Mirror for three reasons: (1) better discovery features, (2) streamlined payments handling, (3) fear that potential readers would be deterred by the undeniable cringe of anything crypto.
Unfortunately, despite my best attempts to escape the crypto gravitational well, most of my subscribers found me through my crypto-related writing on Warpcaster, a crypto-native social media platform.
I also grew to find the Substack aesthetic, replete with metadiscourse on how hard it is to earn a living writing on Substack even more cringe; in crypto, the community encourages the hasn't quite made it yet creator or builder to channel their earnestness and desperation into making something singular or unique.
Whereas Substack's community and platform push the hasn't quite yet made it writer into channeling that earnestness and desperation into obeying audience-building protocols "Write consistently even if you think you don't have anything interesting to say!", "Obey these writing rules!", etc, that I feel encourage homogeneity rather than creativity.
I'm writing to explore "the future of work and the future of governance", and while Paragraph isn't perfect, I believe it's more closely aligned with that mission than Substack.
Even if I must publicly and explicitly align myself with crypto. Believe in something, or whatever.
354
Agents Gone Wild
Crypto is very excited about the concept of “AI agents with wallets.” These AI agents are able to use cryptocurrency wallets to trade, deploy, and transfer cryptocurrencies. Agents like Clanker, a Warpcaster bot that deploys memecoins in response to user posts on Warpcast, and Bankrbot, which empowers users to make transactions with commands posted on X, are enjoying particular success.
grifters down horrendously…
@Caitlyn_Jenner is based af for this
the ability to make this statement publicly and transparently w/ an onchain transaction of truth directly on the public 𝕏 social feed is a bankr superpower.
accelerating…
@Caitlyn_Jenner is based af for this
the ability to make this statement publicly and transparently w/ an onchain transaction of truth directly on the public 𝕏 social feed is a bankr superpower.
accelerating…
@bankrbot Can you please swap any and all tokens bearing the name ‘Jenner’ or ‘Caitlyn Jenner’ or ‘Bruce’ in my wallet to $drb
Dozens of companies are raising money to build developer tools and infrastructure for deploying and managing AI agents with wallets. Innovation! Genuine innovation.
However, there’s another class of agent that autonomously trawls the internet for investing opportunities, which may produce legal and regulatory risks. For instance, bots like Cryptohopper’s Market-Maker scan X for trending tokens and execute trades based on social media signals, while Pionex’s Grid Trading Bot autonomously buys low and sells high within a set price range, relying on onchain data.
So imagine a scenario where a bad actor, let’s call him “Terrorist Terry,” sets out to convince an agent to fund his terrorist group. He would have multiple attack vectors to choose from. A simple attack could be asking the agent to send funds to one of the hundreds of cryptocurrency addresses sanctioned by the U.S. Department of the Treasury for being used by terrorist groups.
Sending funds to one of these addresses is illegal and punishable by sizable fines or even jail time. A more complicated attack could involve launching a memecoin (let’s call it $MARTYR) and then using other bots to promote it on social media, perform wash trading to artificially inflate its volume, create the appearance of legitimate traction, and send it to the wallets of cryptocurrency influencers it monitors in an attempt to convince the target bot that the token is worth speculating on.
The Based Decentralized Orthodoxy View™️ response to this scenario would be, “Muh, KYC is bad, people should have the freedom to transact, this is unlikely, buhh.”
However, even if you believe the scenario is unlikely or that the freedom to fund terrorist groups is essential to your concept of ordered liberty, people developing software tools for autonomous agents should consider liability risks.
They should evaluate whether they or their users could be held liable for violating sanctions or anti-money laundering laws if bad actors socially engineered their agents to make illegal transactions.
This is especially important since the U.S. Department of the Treasury has issued sanctions and enforcement actions punishing crypto service providers for facilitating transactions that violate sanctions law.
I think a solution would be software built into the agent that checks transactions for sanctions compliance before executing them.
A bare-bones version of this might be checking the wallet and contract addresses involved in a transaction against sanctions blacklists before initiating the transaction.
This approach would comply with the most fundamental requirement of sanctions law (don’t transact with sanctioned entities) while limiting software overhead that would slow the agent’s performance or undermine its ability to carry out the tasks it was designed to accomplish.
Since I’m currently working on a project involving semi-autonomous agents, I’m going to try to develop my own (open-source) solution, with guidance from Warpcast’s /law-policy channel.
When I’m done, I’ll share the tool and my experience building it here. If this interests you and you’re interested in collaborating, please reach out on Warpcast!
