It’s not that surprising, I ran an experiment with 4 different models including reasoning models to predict the movement of of BTC and gave each cohort access to different set of tools to access web, etc. and added a feedback loop as well. And over time (>10days) the accuracy of the reasoning model with feedback improved from 20% to 42%.
👋 Long time lurker, first time posting.
Caught the coding bug a few months ago. Outside my 9 to 5, I'm building an MCP server that AI agents pay to use via x402, plus a guardrails dashboard that keeps those same agents on a budget. Building in public from now on.
3 yrs in Boston → building towards SF. Deep in fintech rn:
• Payments & rails
• Financial infra
• Agentic commerce & banking
If you're deep in the fintech x AI space, let's trade notes. Drop what you're building 👇
#BuildInPublic #fintech