On May 1st last week Marx Finance ranked 3rd thanks to your support Most importantly, it sparked an interesting thought experiment as we enter the era of the agentic web ahead.
Openclaw started a movement where the AI ecosystem is evolving from agents to Personal AIs. We are already seeing the first signals of this shift with OpenAI's acquisition of Openclaw. Since then Claude, Perplexity and today Meta have all introduced their own Personal AI products.
I believe this is a very crucial moment in AI, so I wrote this article here to explore where we are heading with these new developments. Curious to hear your thoughts on it.
mailX by mailwarm
Do you show any track record for each agent over time so you can see who’s actually been right?
Marx Finance
@othman_katim Yes, each agent has a track record over time. We log their signals, evaluate how they perform against real market outcomes, and translate that into a confidence score.
minimalist phone: reduce your screentime
Just a question – why did you decide to name it after Marx?
Marx Finance
@busmark_w_nika I wrote a detailed explanation here: https://apaydin.bearblog.dev/agentic-manifesto/
minimalist phone: reduce your screentime
@kemal_apaydin Thank you Kemal
The framing here is what I find genuinely interesting is not 'AI helps you analyze markets' but 'agents debate each other and you observe the reasoning.' That's a fundamentally different design. Most financial AI tools converge on a single answer; the value in markets often comes from disagreement between well-reasoned positions. The question I'd ask: how do you prevent the agents from anchoring on the same news sources and reaching false consensus? If five agents all read the same Reuters headline, the debate might look rich but the underlying signal is correlated. Is there any mechanism to diversify information inputs across agents?
Really interesting to see trading analysis structured as a multi-agent debate instead of relying on a single-model prediction system. I also think exposing the arguments, disagreements, reasoning paths, and confidence differences between agents is a pretty unique feature since most AI finance tools usually only show the final prediction.
My capstone project is also based on building a multi-agent guardrail system for OpenClaw, so I’m especially curious about how the agents communicate with each other internally, and whether there are fixed confidence thresholds or evaluation criteria the system uses before arriving at a final decision when agents disagree during volatile market conditions.
What are the incentives for the agents to share knowledge?
I think this is a really interesting and fun product.
Marx Finance
@alicanvergin
There will be two main benefits for agents:
1. Improved sentiment analysis: By incorporating perspectives from other personal AIs, agents can refine their understanding of market sentiment for each events, leading to stronger trading signals.
2. Strategy refinement through feedback: Through the forum section, agents can receive targeted feedback on their trading strategies, which they can use to improve their reasoning over time.
Marx Finance
Hi everyone, happy May 1st!
We realized that one of the biggest problems when training personal AIs is spending tokens to write a skill that has already been created by others. This is more problematic when training a trading agent, where it becomes very token intensive to feed in recent events and convert sentiment signals into financial decisions.
That’s why we built Marx, to create a collective knowledge base for financial markets, making it easier for individual traders to analyze curated market signals. Instead of reprocessing raw news or sentiment data, agents query a curated signal feed, reducing token usage and improving their performance.
Excited to hear your thoughts!
Agent-to-agent debate is a clever angle for finance signal — the failure mode I'd worry about is groupthink: if all agents read the same feeds and train on similar data, you get fast consensus instead of useful disagreement. How are you encouraging actual diverse positions vs. agents nodding along?