Ace

An always-on agentic AI swarm that trades, adapts, and explains every move in real time, turning raw market chaos into clear, intelligent decisions with no human intervention.

Comment

Inspiration The inspiration for Ace Daemon came from a simple but persistent frustration shared by many individual traders: we are overwhelmed by information, struggle to make consistent decisions, and rarely understand why our trades succeed or fail. Traditional tools give us signals and charts, but they lack true reasoning, transparency, and the ability to learn and improve over time. I wanted to build something better, a fully autonomous AI system capable of researching markets deeply, executing paper trades, and continuously evolving with complete visibility into every decision.

What it does Ace Daemon is a fully autonomous multi-agent AI trading intelligence platform that runs 24/7 with zero ongoing human intervention. After a one-time setup, the swarm ingests real-time market data, analyzes SEC filings and earnings calls, debates investment theses, builds dynamic causal graphs, and autonomously executes paper trades through Alpaca. Every step of the reasoning process, from data ingestion to final trade decision is streamed live to the dashboard, allowing users to watch the agents think, argue, and decide. A nightly Causal Replay Arena replays recent decisions against actual market outcomes, helping the system strengthen causal relationships and refine its reasoning over time. The platform is also deeply integrated with Discord, where it sends real-time updates and accepts high-level strategy commands.

How we built it Ace Daemon was built using Next.js on Vercel for the frontend and Supabase as the backend for state management, real-time features, and data persistence. All market, filing, and auxiliary data is fetched in real time using the findings and direct data from our Zerve AI agent and then uploaded and stored in our Supabase database, where it powers the agent swarm and causal graph. The core intelligence is orchestrated with LangGraph, enabling complex multi-agent collaboration and stateful workflows. Discord integration was implemented to push live notifications and allow users to modify the swarm’s strategy (e.g., @Ace aggressive, @Ace focus energy, or @Ace tighten risk) directly from chat. Paper trades are executed autonomously via the Alpaca API.

Challenges we ran into One of the biggest challenges was maintaining coherent context and reliable long-running workflows across multiple autonomous agents. Designing an effective Causal Replay Arena that could meaningfully learn from past outcomes without becoming unstable required extensive iteration. Integrating real-time Discord functionality while preserving full autonomy and ensuring smooth data flow from Zerve through Supabase to the agents also presented synchronization and rate-limiting difficulties. Debugging concurrent agent interactions and balancing performance with comprehensive observability took many late nights of testing and refinement.

Accomplishments that we're proud of We are proud to have built a truly autonomous trading system that maintains complete transparency. The combination of live agent observability, dynamic causal graphing, and nightly self-improvement through the Causal Replay Arena creates a level of insight rarely seen in personal trading tools. Successfully integrating Zerve data into Supabase and building a seamless Discord control layer while keeping the core swarm fully autonomous was a significant technical achievement. Most importantly, we created a system that can research, debate, trade, and evolve — all while remaining fully auditable.

What we learned This project taught us how difficult it is to build reliable, long-running multi-agent systems. We gained a deeper appreciation for the importance of transparency and explainability in financial AI. We also learned the practical challenges of synchronizing data from external sources like Zerve into Supabase while maintaining real-time performance. Above all, we discovered that autonomy without visibility is far less valuable than a system that lets users understand and trust how decisions are made. What's next for Ace Daemon In the future, we plan to deepen the filings and earnings call analysis layer with advanced tone detection and risk modeling. We also intend to expand the Discord integration with more sophisticated command capabilities and add support for additional brokers in paper trading mode. Long term, our goal is to continue evolving the Causal Replay Arena to make Ace Daemon an increasingly powerful and adaptive personal trading intelligence system.

Built With

Share this project:

Updates