How AI Agents Are Becoming Self-Learning Money Managers

Decentralised News

Best crypto exchanges, wallet and trading tools: https://decentralised.news/ecosystem Read the complete guide: https://decentralised.news/how-ai-agents-are-becoming-self-learning-money-managers  The Future of Finance Has Arrived — Powered by Autonomous AI Agents

Transcript

00:00today we'll be talking about ai agents and how traditional automation typically follows fixed

00:06rules it doesn't actually improve the experience right so by contrast you've got reinforcement

00:13learning which treats markets as an environment where an agent takes actions that is either buy

00:18sell size or hedge and receives rewards which is a pnl or risk adjusted returns here and updates

00:26its policy to maximize a long-term reward so in finance these reinforcing reinforcement learning

00:35and continual online variants help agents to adjust regime changes right instead of overfitting to a

00:43particular single vectors for example and recent academic industry overviews kind of highlights

00:50rl sustainability for uh kind of a dynamic sizing inventory control execution and also risk in

00:57non-stationary markets as well so encrypted this evolution matters because execution accounting

01:03and also custody can all be automated on chain letting agents actually both reason and act without

01:09any uh human middleman right so venture and ecosystem essays now explicitly frame autonomous

01:16policy-bounded defi agents as the next ux leap and users express uh kind of intense agents translate

01:24them into multi-step on chain actions right so these three cases that we'll be looking at today

01:31offer these agentic finance uh ai agents uh starting with shekel which is part of the virtuals ecosystem

01:39right where it's a spin-up fund and then you know you can let the agent manage it's pretty much sits

01:45inside this virtual protocol landscape a base native platform for launching and co-owning ai agents

01:52so shekel's product is a permissionless fund that can be fully human-run or ai managed with staking that

02:01shares profits and fees from every fund under that brand's umbrella so the flagship agent rabbi shlomo

02:09executes uh autonomously on base it lands in a shekel style fund right agent loops uh kind of look like

02:18data ingestion signal allocation policy execution whether that's swapping lps staking then the

02:25realized reward the policy update and then the last step which is the continuous learning is kind of where

02:32these uh ai managed uh framing departs uh from that basic bots uh kind of how basic bots would be

02:39working policy weights which can be refreshed as new episodes as well to settle using this rl continuous

02:46learning concept uh it's something that you know is very interesting so shekel token itself is live on

02:51base uh you can uh swap it on dexes etc and banker is also another um kind of a ai agent which is a

03:01social chat driven ai agent that lets use a swap bridge research place orders by typing on x or

03:08farcaster uh via private or private terminal which is a dm to trade uh the project is positioned as a

03:15leading base edge a agent and also describe the kind of support for multi-chain actions and advanced

03:25auto types as well uh it matters because obviously it's intent plan executes uh pipeline

03:31exposes that agentic model uh to the mainstream uh for example user states something like uh buy

03:37300 usdc of eve dca weekly and then the agent translate that into a multi-state plan with

03:44safeguards and the team also navigated platform policy risk uh brief x suspension and reinstatement

03:50also i kind of find a scoring a real world constraint for social embedded agents there as well distribution

03:57can be throttled sometimes so a robust fallback on ui kill switch limits are all essential so the token

04:04the revenue itself bnkr is the token ticker for the uh banker ecosystem public explainers i note that

04:11you know obviously it powers a few fee and also a value capture for the agent uh and it's backed by

04:19base aligned venture as well so go check it out banker and then you got y combinator

04:24an accelerator for for agents uh kind of uh presented as an autonomous accelerator agent within

04:32the uh virtuals ecosystem as well designed to support early stage uh ai blockchain projects

04:38curating funding iterating agent ideas as well so think of it as an agent that kind of helps

04:43grow other agents so project trackers and profiles describe its role in its contract contracts on

04:50base as well uh it trades on base uh dexes as well and uh yeah so it's a very interesting one so under

04:57the hood i think these uh agents uh self-led you know what's interesting is obviously they they all

05:03vary depending on the teams that are making them obviously the state and the features on chain

05:08balances the fields the price paths the volatility liquidity depth impact of chain context and all that

05:15the policy the functions then the reward system and obviously the updates so it's a very interesting

05:21but obviously there's also some risks obviously with intent ux execution like banker law friction drives

05:27that usage so the agent handles that orchestration on chain track records as well shekel style funds the

05:34verifiable right pnl builds trust faster than opaque c5 products i think uh ecosystem flyables like why

05:42um discovery support iteration for the agents so top risks you should think about managing is

05:48overfitting and regime change distribution and path uh or the risk there with the platform and then

05:54obviously the security and the privilege so i think there's a lot to learn guys when it comes to these

05:58ai agents if you just go to our main site the centralized.news you'll see a full list of a lot of

06:04these ai projects uh being broken down so make sure you check those out and check out the links down

06:09below to trade on the best platforms peace

Category

Transcript

Be the first to comment

Recommended