Antonio García Martínez pfp
Antonio García Martínez
@antonio
Given that crypto is programmable money and financial logic, and given advances in AI, why hasn't anyone really created agentic AI bots that essentially live on chain, doing whatever arbitrage (or speculative trading) is to be found, and generally assuming almost standalone financial sentience?
10 replies
2 recasts
11 reactions

Antonio García Martínez pfp
Antonio García Martínez
@antonio
You could create an AI agent that has its own financial life onchain and even shitposts about it on Warpcast, call it Ansem-as-a-service to give it a name, and yet seems not to have happened yet? That would be the ultimate (crypto) Turing test. Given it's all NFT avatars, how could you even tell?
1 reply
0 recast
5 reactions

cody pfp
cody
@codyb.eth
mentioned this last night but I think the answer is that this is being constantly tried but it's very difficult to earn $ consistently. The other X factor is that once you release an agent, how does the ecosystem change when 1000x of those agents can be copied and deployed?
0 reply
0 recast
0 reaction

kevin j pfp
kevin j
@entropybender
you could for sure. it's just an alt and you give your pk to an on device agent and you're basically good to go. ppl don't do it because it's not that much more profitable than simply sniping a new memecoin. you can do it as an experiment tho. i had an idea several months ago while doing ai crypt stuff to do this but for uma protocol. release several agents, bond UMA, see if they are just as good at resolution and indistinguishable from real humans
0 reply
0 recast
0 reaction

Marius Watz pfp
Marius Watz
@mariuswatz
Well, high-frequency trading won't work because of gas and the relative slowness of blockchains (compared to public stock markets.) I would assume there are bots out there, but it has to be a risky business given the currency volatility. A dollar usually stays a dollar, but 1 ETH can change a lot in 24 hours.
0 reply
0 recast
0 reaction

Brian Kim pfp
Brian Kim
@brianjckim
@entropybender
0 reply
0 recast
2 reactions

Jonny Mack pfp
Jonny Mack
@nonlinear.eth
jaredfromsubway.eth
1 reply
0 recast
1 reaction

Galen pfp
Galen
@galen
This is exactly what @koeppelmann.eth and the Gnosis AI team are building... https://www.gnosis.io/blog/meet-gnosis-ai
0 reply
0 recast
1 reaction

sha ↑ pfp
sha ↑
@sha
have to imagine it's already happening and we don't realize it yet
0 reply
0 recast
0 reaction

Galen pfp
Galen
@galen
@antonio bumping this with a short talk from Web3 Summit “Onchain agents and longtail predictions markets” https://streameth.org/web3summit/watch?session=66c1c52d6e7e09ffb2c8c6c8
0 reply
0 recast
0 reaction

Prithvir pfp
Prithvir
@prithvir
The hallucinations and stochastic nature of LLMs make generating consistent onchain profit infeasible. It’s possible to have narrow machine learning trained bots with like python scikit learn that do pretty well onchain. Whenever someone says “I’m building an AI agent to trade on chain” I know they’re an idea guy and not a builder.
0 reply
0 recast
0 reaction